8800GTS 512Mb (G92) vMods

[T_M]

Inspiration from largon's work on 8800GT :)

http://img.photobucket.com/albums/v310/T_M/G92-8800GTS-vMods.jpg

http://img.photobucket.com/albums/v310/T_M/G92-8800GTS-vMods2.jpg

[K404]

yup! same as GT :D

Let us know how you get on with clocking! :D

[T_M]

Note difference between the GT and GTS at the base of the 1K VR there...

[K404]

To account for GTS default voltage being higher? :)

[T_M]

Absolutely no idea, but i just noticed it was different to the photos of GT.
People have been asking about teh differences between the two, particularly the extra power phase etc.

[Achill3uS]

nice clean mods, thanks :)

[metro.cl]

Nice, can you send the BIOS i think there should be a vmod via BIOS :)

[fornowagain]

So stock voltage is 0.05v higher than the GT with bios mod, interesting.

[finalanrchy]

How does the memory overclock? Better than The GT?
I am thinking of getting one of these cards, but i might just get the standard GT if its not that much better.

[dinos22]

working OVP mod will be important here as well right tim?

excellent pics and best to see someone actually have the mods done as well in guides :)

[Trouffman]

Nice done :)

Nice pics ANd nice to have examples of value for VR and voltages !

[T_M]

Please note that after playing with these mods last night, i wish to revise my mod to a 500 Ohm VR for vGPU.
1K makes voltage jump up too quickly once you are about 1.35V+ (very small turn can go up too much volts).

[K404]

I got my card today- 300 ohms would work well too :)

[Achill3uS]

I got my card today- 300 ohms would work well too :)

push it Kenny push it :D

[K404]

Hey Bala :D I will, dont worry! The package is in the UK now ;) Hopefully get it tomorrow, and real benching will be on Monday :)

[largon]

The pic in my GT guide has some extra components that are not found on retail G92GT cards - components like the 8-legged IC under the Primarion vGPU controller that T_M referred to. Thus I believe the cards & mods are identical except for what comes to the missing phase.
btw, what GDDR chips does the GTS have? Are they Qimonda 1.0ns?

[Achill3uS]

Hey Bala :D I will, dont worry! The package is in the UK now ;) Hopefully get it tomorrow, and real benching will be on Monday :)

great news :D have fun with it ;)

i will do the mods and some benches around the weekend too :)

[T_M]

Havnet pulled off HSF yet but that's happening tonight or tomorrow.
Ive seen photos of the memory elsewhere though so look around.

[VETDRMS]

EVGA 8800GTS-512 doing pretty well.

vGPU: 1.275v
vMem: 1.936

Core: 864
Shader: 2106
Memory: 1134 (2268!)

3DMark'06: 15,156 with E6600 @ 3.666

36C loaded.

:cheers:

[K404]

http://img70.imageshack.us/img70/8018/g92gtsreadpointssk5.jpg (http://img70.imageshack.us/my.php?image=g92gtsreadpointssk5.jpg)

http://img512.imageshack.us/img512/785/g92gtsvoltmodsgd3.jpg (http://img512.imageshack.us/my.php?image=g92gtsvoltmodsgd3.jpg)

Theres mine :) I know the readpoint wiring is completely impractical, but its "just" for benching.

[Achill3uS]

:D :up:
very nice work!

[dinos22]

have you guys looked into that OVP mod that one of the guys in 8800GT thread figured out by unsoldering a couple of things around that Prix@#^@%#*^@ chip whatever the actual name is heheh

i also wonder what the actual OVP for this card is?
seems it is definitely higher to GT

[K404]

EDIT: removed for being utter nonsense

I hadnt looked at it until you pointed it out so...thanks for the heads up! Im hoping it wont be an issue, but every pushed back ceiling will help me.


@ Bala: Cheers bro! Are you all ready to bench as well?



big OT: I was gonna be benching tomorrow, but delayed it till Monday. I was hoping to be the first to publish frozen GTS clocking...I think il be beaten to it now! :(

[cowie]

have you guys looked into that OVP mod that one of the guys in 8800GT thread figured out by unsoldering a couple of things around that Prix@#^@%#*^@ chip whatever the actual name is heheh

i also wonder what the actual OVP for this card is?
seems it is definitely higher to GT

i will try on both my v modded 88gt and 88gts 92 this next few days
but air(zalman vf-1000) is my limitation.


and great work on the mods all:up: i love this kinda stuffs

[T_M]

Mine will be frozen Saturday and Sunday :)
Any OVP developments i'll update here.

K404, if you were planning to use insulation then you might want to revise your readpoint locaiton becasue that area needs foam on it.

[K404]

Yea, been thinking about that. Still to test the pot placement. Might need to add in some insulation for the back of the card. :down:

Good luck with the benching T_M ;) I hope you get the MHz and results i think this card of capable of ;)

[G-Man]

Is the OVP protection still kicking in at the same 1.35v as the GT? If not, What clocks are you all getting with 1.4v plus?

[Nosfer@tu]

I must have missed somthing, this is the first time i se 8880

[MorGo7h]

I must have missed somthing, this is the first time i se 8880

Prolly a mistake.

[T_M]

fixed

[syn7h37ic]

This is my result:
1.36 volt, GPU only.
http://img75.imageshack.us/img75/2772/gtsvmodeb7.jpg

[dinos22]

try running higher shaders as they impact performance a lot

[tranceaddict]

What is the stock voltage for the 8800 GTS/GT? Moreover, is there a BIOS volt mod for the GTS?

[Stelios]

What is the stock voltage for the 8800 GTS/GT? Moreover, is there a BIOS volt mod for the GTS?

1.15V stock voltage for GTS . No bios mod can give you higher vgpu.

[NiNe_iZe]

great work Tim !! :up:

Thanks for this mod :clap:

[s e t h]

dont need to unsolder them, shorting them out with small pieces of wire will work :) Electricity is lazy- if it has a choice of a resistive path (resistor) or non-resistive path (bit of wire) it will skip the resistor totally. Thats assuming the wire has no resistance at all, which is only true in theory, but its "close enough" here.

worst case scenario: the wire and the resistor form a parallel resistive network, but the total equivalent resistance will basically be the resistance of the wire.

[/theory]

I hadnt looked at it until you pointed it out so...thanks for the heads up! Im hoping it wont be an issue, but every pushed back ceiling will help me.


@ Bala: Cheers bro! Are you all ready to bench as well?



big OT: I was gonna be benching tomorrow, but delayed it till Monday. I was hoping to be the first to publish frozen GTS clocking...I think il be beaten to it now! :(

please explain this to me a little more carefully...i really don't like obviously voiding my warranty and if what you say is correct then a simple stripe of conductive paint accross the top the the relevant resistors would achieve the same removal of OCP.
to my thinking removing the resister will not be the same as bridging it...removed the circuit should be broken due to connections literally missing and therefor almost infinate resistence in those circuits disables the OCP...however bridging the resisters with wire or paint will result in almost no resistence being offered and the circuits will remain intact...

so how is bridging the resisters anything like removing them?

please excuse if i've misunderstood the intricacies involved here as i have no electronic background whatsoever..

if anyone can work out how to disable OCP without voiding the warranty i'll be very pleased
(at this point i'm also not sure if these GTS512mb cards have the same capability in terms of removing OCP will allow LN2 voltages without further issue)

[T_M]

I too was confused by that comment.

[cowie]

so how is bridging the resisters anything like removing them?
its not two different things
remove no current at all
short 100 current floqw
so im lost on that comment as well

[The Asgard]

Glad it just isn't me but like everyone else I just about know enough to get by.

Shame really since I have a degree in electrical and electronic engineering. :eek:

[jason4207]

I think that if you decrease resistance by soldering another resistor (or Cu wire) in parallel you are raising the OCP threshold. OCP won't kick in until much higher current flows. If you get the right value resistor in there you can raise OCP, but still allow some level of protection (probably a good thing). Just soldering a Cu wire or using a conductive pen will raise the OCP threshold to a level where you probably will never have to worry about it anymore.

Removing the resistor completely removes OCP.

Both end up giving you the same result, but do it in different ways.

This is what I've gathered from all my readings on here and other forums as well some PMs w/ some very knowledgeable people. I also have a B.S. in Computer Engineering. :p:

I don't know this to be fact, but it seems to make sense. In any case take it w/ a grain of salt. ;)

[Timbosan]

I think that if you decrease resistance by soldering another resistor (or Cu wire) in parallel you are raising the OCP threshold. OCP won't kick in until much higher current flows. If you get the right value resistor in there you can raise OCP, but still allow some level of protection (probably a good thing). Just soldering a Cu wire or using a conductive pen will raise the OCP threshold to a level where you probably will never have to worry about it anymore.

Removing the resistor completely removes OCP.

Both end up giving you the same result, but do it in different ways.

This is what I've gathered from all my readings on here and other forums as well some PMs w/ some very knowledgeable people. I also have a B.S. in Computer Engineering. :p:

I don't know this to be fact, but it seems to make sense. In any case take it w/ a grain of salt. ;)


No stop confusing people. If you remove the resistor, the circuit is broken, and the primarion contoller never gets a signal from that circuit to trigger OCP so OCP is removed. If you short out the resistors, you would increase the signal the primarion controller gets, by lowering the resistance, and most likely lowering the OCP point. The card may not even boot. Do not bridge the resistors.

[NightRaven]

well one thing i thought of then..

would be grounding the ocp/ovp resistors.

however, the primarion might take it as infinite resistance or bridging the resistors since the circuit isnt broken.. only one way to find out :D

[s e t h]

simple pencil modding the relevant OCP resisters will reveal if lowering resistence raises OCP kick in point...easy and usefull experiment.

i too wondered about grounding the OCP resisters but then though better of it :p

good luck ... maybe we see some even more serious clocks from these cores pretty soon...like 1200mhz or more :0

thanks for publishing your sub0 results with these NightRaven
it seems serious 3D benching just reached a lower price point...always a good thing :D

[jason4207]

No stop confusing people. If you remove the resistor, the circuit is broken, and the primarion contoller never gets a signal from that circuit to trigger OCP so OCP is removed. If you short out the resistors, you would increase the signal the primarion controller gets, by lowering the resistance, and most likely lowering the OCP point. The card may not even boot. Do not bridge the resistors.

I heard from a very good source that soldering resistors in parallel (lowering resistance) would raise the OCP threshold. He specifically mentioned 4.2-4.7k for the 2 locations on the 8800GT. I'm assuming that bridging the connector would just raise the OCP threshold higher, and since someone posted that it worked for them it seemed logical.

Re-reading his email it seems that bridging is a bad idea after all as the regulator is also using that signal to keep the current sharing b/n the 2 phases in balance.

Soldering resistors in parallel seems to be the best way to gain more voltage headroom, and is the only way recommended to me by my source.

I do agree that removing the resistor will remove OCP, and I said that in my previous post.

Again I haven't tried any of this...yet, but I know my source has, and he confirmed the results.

My source: Viper John (thanks buddy!)

[Timbosan]

i may have been a little harsh, i just dont wanna see a post saying "i just bridged the resistors and burnt my card". Viper John does know his stuff, no question. However are you sure you arent confusing OCP with the OVP mod? In the OVP mod we decrease the resistance of a single resistor, to ~4.7k, as this increases the signal to which the real signal is compared thereby raising the over-voltage trigger point.
for this OCP mod, we are trying to decrease the signal sent to the primarion controller, rather than alter the signal to which it is compared, as we dont know how that system works exactly.

[jason4207]

I looked back again to confirm and he has the 4.2-4.7k resistors in parallel w/ the 2 brownish resistors b/n the 78x & the 15b resistors...not the 78x resistor itself. So you were good to call me on it, b/c otherwise mis-information may have spread.

This mod will reduce the amount of voltage drop the OCP sees by ~25%.

This is the OCP, not OVP mod that I'm referring to, but it is on the 8800GT, not the 8800GTS-512 mentioned in the thread title. I'm sure the mods will be similar, though. The GTS-512 has 3-phase power, so I'm not sure how that affects it yet. I'm still learning!

Sorry for the confusion!

[T_M]

can a mod please move this GT talk to another thread?

[jason4207]

Sorry if I seemed OT, but it seems to me that a mod for the GT will be relevant to finding a working mod for the GTS-512.

If soldering the 2 resistors in parallel as I described works on the GT, then it would seem logical to assume that doing something similar on the 3 resistors on the GTS-512 would work as well.

http://img248.imageshack.us/img248/4440/ocpxu7.jpg

The values on the GT were 4.2-4.7k, and I'd think similar values on the GTS-512 might work, but since the GTS-512 has a higher working voltage then a different value may be in order.

I'll leave it to you experts to figure out the values if I'm on the right track. Or maybe I'll hit up VJ and see if he's had time to play w/ this card yet.

Sorry if I pissed anyone off. I just got my GTS-512 yesterday, and I just wanted to see this card get tweaked before I took my soldering iron to it (this will be my 1st vmod on a gfx card). I thought I had some info that might help you guys reverse engineer this thing. I'm sorry if I mislead anyone or posted any mis-information.

[jenqho]

[QUOTE=jason4207;2638524]I looked back again to confirm and he has the 4.2-4.7k resistors in parallel w/ the 2 brownish resistors b/n the 78x & the 15b resistors...not the 78x resistor itself. So you were good to call me on it, b/c otherwise mis-information may have spread.

Are you sure the 2 brownsih components are "Resistors"?? They looks like SMT capacitors to me.

[jason4207]

Sorry again! Yes those components are SMC's, not SMR's.

[K404]

please explain this to me a little more carefully...i really don't like obviously voiding my warranty and if what you say is correct then a simple stripe of conductive paint accross the top the the relevant resistors would achieve the same removal of OCP.
to my thinking removing the resister will not be the same as bridging it...removed the circuit should be broken due to connections literally missing and therefor almost infinate resistence in those circuits disables the OCP...however bridging the resisters with wire or paint will result in almost no resistence being offered and the circuits will remain intact...

so how is bridging the resisters anything like removing them?

please excuse if i've misunderstood the intricacies involved here as i have no electronic background whatsoever..

if anyone can work out how to disable OCP without voiding the warranty i'll be very pleased
(at this point i'm also not sure if these GTS512mb cards have the same capability in terms of removing OCP will allow LN2 voltages without further issue)


I too was confused by that comment.


so how is bridging the resisters anything like removing them?
its not two different things
remove no current at all
short 100 current floqw
so im lost on that comment as well


I think that if you decrease resistance by soldering another resistor (or Cu wire) in parallel you are raising the OCP threshold. OCP won't kick in until much higher current flows. If you get the right value resistor in there you can raise OCP, but still allow some level of protection (probably a good thing). Just soldering a Cu wire or using a conductive pen will raise the OCP threshold to a level where you probably will never have to worry about it anymore.

Removing the resistor completely removes OCP.

Both end up giving you the same result, but do it in different ways.

This is what I've gathered from all my readings on here and other forums as well some PMs w/ some very knowledgeable people. I also have a B.S. in Computer Engineering. :p:

I don't know this to be fact, but it seems to make sense. In any case take it w/ a grain of salt. ;)


No stop confusing people. If you remove the resistor, the circuit is broken, and the primarion contoller never gets a signal from that circuit to trigger OCP so OCP is removed. If you short out the resistors, you would increase the signal the primarion controller gets, by lowering the resistance, and most likely lowering the OCP point. The card may not even boot. Do not bridge the resistors.


oh **** Im so sorry for that- do not short the resistors! I was utterly wrong with that information. I removed my comment. :shakes:

half-assed excuse: it was very early morning and I would have been tired.

[Timbosan]

Sorry if I seemed OT, but it seems to me that a mod for the GT will be relevant to finding a working mod for the GTS-512.

If soldering the 2 resistors in parallel as I described works on the GT, then it would seem logical to assume that doing something similar on the 3 resistors on the GTS-512 would work as well.


where is this thread where he has soldered the resistors in parallel with the caps? i came up with the idea of removing the caps altogether, but never tried it, and n1ck r1mer thought of soldering VR's across the caps, but hasnt done it either (though we were talking about it concerning the G80 GTS)

[jason4207]

It's not in a thread. John didn't want to post it online, so he sent it to me in an email.

[jason4207]

Does anyone know if increasing vRAM voltage has any effect on the GTS-512? I know it had no effect on most GT's.

Also, how high can you increase vGPU before the OCP kicks in? I'm assuming it's a little bit higher than on the GT...maybe .05v higher? AquaMark3 seems to be a good test for OCP.

[187(V)URD@]

Should be easy if you can overclock a graca like a cpu in bios.

[tomati]

EVGA 8800GTS-512 doing pretty well.

vGPU: 1.275v
vMem: 1.936

Core: 864
Shader: 2106
Memory: 1134 (2268!)

3DMark'06: 15,156 with E6600 @ 3.666

36C loaded.

:cheers:

Hi,

Are these the stock voltage of the evga 8800GTS 512mo ? or did you made the vmod for both ( gpu & mem) ?

[K404]

stock voltages are 1.16/20.1V

(at least...they are for me!)

[VETDRMS]

Hi,

Are these the stock voltage of the evga 8800GTS 512mo ? or did you made the vmod for both ( gpu & mem) ?

No, they are not stock voltages. I increased the stock vGPU and decreased the stock vMEM.

:up:

[tomati]

No, they are not stock voltages. I increased the stock vGPU and decreased the stock vMEM.

:up:

Ok thanks ,

do you remember what O/C do you hit on stock ?

I ordered one evga, hope to get it next week.

[VETDRMS]

Ok thanks ,

do you remember what O/C do you hit on stock ?

I ordered one evga, hope to get it next week.

I never had the card stock, I modded it the day it arrived before installing it.. :ROTF:

[jason4207]

Stock: I'm at 771/1944/1062 on my GTS-512 at 100% fan stable for at least 25mins in ATI Tool, and made it through 8 loops of Battle of Proxycon (3DMark03). I consider this stable.

16190-3DMArk06 w/ 450x8 (3.6GHz) x3210 Quad, P5K-Premium, 2GB RAM @ 1125MHz 5-5-4-10 --> (24/7 stable).

I haven't tried to test for stability at lower fan settings yet.

[ViperJohn]

Stock: I'm at 771/1944/1062 on my GTS-512 at 100% fan stable for at least 25mins in ATI Tool, and made it through 8 loops of Battle of Proxycon (3DMark03). I consider this stable.

16190-3DMArk06 w/ 450x8 (3.6GHz) x3210 Quad, P5K-Premium, 2GB RAM @ 1125MHz 5-5-4-10 --> (24/7 stable).

I haven't tried to test for stability at lower fan settings yet.

You should be fine with the GTS-512 at 1053 in 03 BoP. Also loop 3Dm05 Canyon Flight as well.
My GTS would do 1134 in 03 BoP but craps it's pants in Canyon Flight at anything over 1107
memory and in the same manner as the GT's crap out in BoP.

Viper

[Timbosan]

It's not in a thread. John didn't want to post it online, so he sent it to me in an email. I understand his concerns over in OCF

thanks ive got it sorted now, and similarly i wont be posting it lol. you can delete convo if you like.

:up:

I have found for the 320mb GTS, GT2 (firefly forset) in 05 & 06 to be the best test for ram artifacts. Also GT3 (canyon flight) in 05 is great for finding shader artifacts.
Reapeating nature in 03 is a good test for max gpu clocks, too high you will get lockups.
May also be the same for the G92 rev cards.

[jason4207]

You should be fine with the GTS-512 at 1053 in 03 BoP. Also loop 3Dm05 Canyon Flight as well.
My GTS would do 1134 in 03 BoP but craps it's pants in Canyon Flight at anything over 1107
memory and in the same manner as the GT's crap out in BoP.

Viper

I did 8 loops of CF in 3Dm05 w/ no issues at 771/1944/1062, so I guess that's one more test that is showing I'm good & stable at these clocks.

Now I just need to mod this thing, and see what it can really do!

[ViperJohn]

thanks ive got it sorted now, and similarly i wont be posting it lol. you can delete convo if you like.

:up:

I have found for the 320mb GTS, GT2 (firefly forset) in 05 & 06 to be the best test for ram artifacts. Also GT3 (canyon flight) in 05 is great for finding shader artifacts.
Reapeating nature in 03 is a good test for max gpu clocks, too high you will get lockups.
May also be the same for the G92 rev cards.

Were the SMR's we talked about on the front of the GTS 320???

Viper

[Timbosan]

Were the SMR's we talked about on the front of the GTS 320???

Viper

not that i could find, directly behind (front side) the primarion controller are solid capacitors then a blank space until the edge of the heatsink obscures the board, and the power mosfets start. i cannot find them as yet. I do however feel the r/c circuit will work the same way as the G92 GTS & GT, the resistance across the 3 caps in question is 1.29k & the adjacent resistor is 78 ohms. im taking the card out tomorrow (been so busy havent got a chance) and probing around fully to see if i can find the smr's. i cant remove the HS though, as i dont have any thermal paste left, but im almost certain they would not be that far away. maybe the design of the px3540 differs from the px3544 in this regard, and there is only one resistor & one cap for each phase.

[VoRtAn_MaDgE]

I modded my 8800gts 512 from gigabyte this morning.
Still missing vmem (does it really helps?)
Don't have the correct pot at my place.

Well...added a switch for on/off.

http://img143.imageshack.us/img143/6718/26806464og8.jpg
http://img148.imageshack.us/img148/1324/85440098cj9.jpg
http://img143.imageshack.us/img143/122/94285739tw0.jpg

Results later in the correct post for the efect.

Cya guys! ;)

[Pacha]

edited

[scotty1991]

What does the little power connector type wire do on these voltage mods?

[jason4207]

What does the little power connector type wire do on these voltage mods?

That's just there to give a place to measure the voltage you are applying.

[VoRtAn_MaDgE]

What does the little power connector type wire do on these voltage mods?

Correct.
The switch is just for on/off.

[scotty1991]

That's just there to give a place to measure the voltage you are applying.

Fair enough, just looked like a molex or pci-e connector to me and got me confused :D

[4Qman]

MSI 730/1944 OC edition

Stock Voltages
Core = 1.13
VMem = 1.94

I am ATI tool stable at 800/1190 With these voltages and 3Dmark 06 stable 820/1200 STOCK AIR!

I think this is a decent clocking card or will be once Vmoded. :D

[T_M]

My GTS would do 1134

Interesting, got screeny?

[aussie-revhead]

What are you guys doing for ram sinks ? Does anyone have a pic ? What aftermarket air coolers are you guys using ? I will be using a mcw60 , but am curious as to my alternatives .

Good work T_M and congrats to team-eva and team-au.

:up:

[T_M]

team-eva doesnt exist??
We are very much a family at i4memory, not a one man band.

Dont bother with ram sinks for this card

[aussie-revhead]

No sinks at all ? Not extra chipset etc ?

I look forward to buying one of these .

:up:

[finalanrchy]

Does anyone know what the jumper looking thing at the back of the card near the power connector is for?

I am just to tempted to place a jumper on it :p: .

[eva2000]

haha no team-eva it's Team AU = http://teamau.i4memory.com/ and http://teamau.i4memory.com/1 :)

[aussie-revhead]

Nah mate I meant i4 in general not team-au :rolleyes:

ie your group = i4 = team-eva

:D

As soon as money comes through from my GTX sale I will be off to grab a 512 GTS . Then off to Jaycar to get some more VR's ...

:up:

[T_M]

No sinks, and there is no extra chipset either.

[VoRtAn_MaDgE]

Yes I am using a hard mod to reduce vmem.

1.4ns timings are tighter than the stock 1.0ns timings. I am going to tighten the timings further to a theoretical 1.5ns spec and see what the results are. Give me 10 min and I'll post a loose/tight timing 3Dmark06/05 score.

I won't hi-jack this thread, but tighter timings netted a very minimal increase in 3DMark05. 23244 v 23173. 3DMark05 would not complete at 1144 with the default timings, so yeah, tightened timings --> more mhz.

I've checked the qimonda 1.4ns datasheet but i can't find all the values that nibitor is showing me.

quimonda 1.4ns
http://img407.imageshack.us/img407/4949/quimonda14nspage1ds3.jpg
__________________________________________________ _______________________________________

http://img139.imageshack.us/img139/5797/quimonda14nspage2xi8.jpg

Gigabyte bios 8800gts g92 that nibitor is showing me...
http://img293.imageshack.us/img293/5296/nibitorg92oq6.jpg

I've got to change all the timming sets ???

Could you share the vmem mod for lowering the vmem value ?

[VETDRMS]

No, only change the value in timing set 1.

1.4ns
---------
tRC: 30
tRFC: 52
tRAS: 18
tRP: 12
tRCDRD: 11
tRCDWR: 6
tRRD: 6

[Mr. K6]

Hey all,

I must have read this thread 4 times and am very interested in vmodding my G92 8800GTS, but there's only one problem: I've never done a Vmod before :eek:! I have a 15W iron and some thin rosin core solder. I also have some 22 gauge solid copper wire, but I think I might use some fan wire (seems like its thinner, 24 gauge+) instead. I did some practice work on an old PCI 56K modem just to get some PCB work under my belt, but I'm still a little wary of the exact procedure. I also ordered some 200, 500, and 1K ohm trimmers that look very similar to the ones T_M used in the first post. Anyway, can anyone throw some hints/tips for a newbie my way, or are there any good guides around? Thanks for all the information already and the great read :up:

[VoRtAn_MaDgE]

No, only change the value in timing set 1.

1.4ns
---------
tRC: 30
tRFC: 52
tRAS: 18
tRP: 12
tRCDRD: 11
tRCDWR: 6
tRRD: 6

Changed timmingset 0 or 1 ?

[Stelios]

Changed timmingset 0 or 1 ?

If you click the "autoselect timingset" button , youl'll get a message with the timingset the bios uses.

[K404]

team-eva doesnt exist??
We are very much a family at i4memory, not a one man band.

Dont bother with ram sinks for this card

One of my RAMsinks feel off due to heat and the card artifacted and locked up instantly? I wouldnt try running without RAmsinks...no way.


___________________________

Max benching clocks on water: 880MHz core (1.375V (idle) 1100MHz RAM if I edged it up slowly. (default voltage)

[VoRtAn_MaDgE]

How about in one post pics of all the mods?
Vgpu, vmem, ocp, ovp ???
Lots of information in different thread... 8800gts, 8800gt, g80 :confused:

Thks!

[gOtVoltage]

On my EVGA the Vgpu mod yields a max of 1.52v on AIR with stock modded heatsink....I will not do the memory the Qimonda dosnt run like my samsung did on the 320...The samsung would run 2300mhz 1.210 no prob! the Q onlydoes 2250max @50c ...It not worth the risk untill they decide to use better memory...Also @ 918//2268//2250 the bandwidth sucks....256bit bus shows little improvement when overclocked past 808//2000//2250...

Note:Ovp kicks in @1.53v on this card with Air

808//2000//2250 = 3Dmark06 13090 points @3.48ghzX2
918//2268//2250 = 3Dmark06 13200 points @3.48ghzX2

Overall the Max SM 2.0 =6150 Overclocked
Overall the Max SM 3.0 =6700 Overclocked

The GPU and Shader helped the SM 3.0 score by 1000points over stock score of 5600!

Having a high Oc on the memory is a must,, This is what helps the SM 2.0 score the most along with the Extra 10bw gained in the band width department:shakes: ..

SM 2.0 Stock was 5300 and Overclocked gained 850 points..Since the The memory dosnt do anything more over 2250mhz im not messing with the Vmod for it...Ive ran the memory @2300mhz but it would fail half way through any test...Temps were 50c max load running anything!

In short the Vmod will get you 918mhz GPU and 2300mhz Shader easy..But with 1.5v-1.525 before the Prtection kicks in on Air!

If some one is using water id like to see results of this with out Chilled Water..

I may throw on a water block and go further ,but am still testing this thing on Air...

This thread has been Very help full,, Solder points are the easiest by far.
Keep up the good work :up:

On a Quad this would easily get 2000-3000points more @3.7-4.0ghz So i guess im happy...No X2's come close to 16000 even @3.5-3.7ghz:down:

http://i168.photobucket.com/albums/u192/darrellsierra/OcVmodBench.jpg

http://i168.photobucket.com/albums/u192/darrellsierra/MaxRunonAir.jpg

[STEvil]

where is this thread where he has soldered the resistors in parallel with the caps? i came up with the idea of removing the caps altogether, but never tried it, and n1ck r1mer thought of soldering VR's across the caps, but hasnt done it either (though we were talking about it concerning the G80 GTS)

http://www.xtremesystems.org/forums/showthread.php?t=171016

[T_M]

Ovp kicks in @1.53v on this card with Air

What did you use to test this?
If it was OVP, then if you turn up voltage while idle the system will crash at 1.52V.
If it was OCP, then if you turn up voltage while idle, then teh system will crash during a benchmark and will be far higher than 1.52V (more likely 1.6V).

Please can you test by running GT2 in 3DMark'06 (Firefly) and tell me at what idle voltage does your OCP kick in?

[gOtVoltage]

What did you use to test this?
If it was OVP, then if you turn up voltage while idle the system will crash at 1.52V.
If it was OCP, then if you turn up voltage while idle, then teh system will crash during a benchmark and will be far higher than 1.52V (more likely 1.6V).

Please can you test by running GT2 in 3DMark'06 (Firefly) and tell me at what idle voltage does your OCP kick in?

Im sorry ,,I think when it had crashed the first time i was ajusting during bench and went too much:D The Volts jumped past 1.6v when it had shut down...Its safe to say That OVP is 1.6v or higher! My card is freezing the screen @ 1.576v loaded..So it may be heat related..Which sucks because Riva reports a max of 51c when i bench this high! Idle is @37c when cranked to 920mhz! If you look in my Screen ,, my temp was about 42c max here then about a few minutes later i aborted from a freezing screen...

I cant complain for Air,,but i really think this thing has to be runing htter than wha tit says or this is my max core @920 and Air?

so this is what i just did...I loaded3Dmark06 started bench @907mhz

Idle 1.52v

load 1.576v

it benched!

So then

Idle 1.53v

Load 1.59v-1.60v

Froze screen...So i may be shutting down due to Core?

i remeber that i had a Full Pc reboot @1.525v during a bench last night...i was shure that was from the OVP...

Hope this helps..was done with memory @ stock clocks too!

http://i168.photobucket.com/albums/u192/darrellsierra/1525vIdle1576vload.jpg

[VoRtAn_MaDgE]

If you click the "autoselect timingset" button , youl'll get a message with the timingset the bios uses.

I'm on vista64 and the autodetect on vista64 doesn't works...already mencioned on mvktech forums...

The bios i used was this one - http://www.mvktech.net/component/option,com_remository/Itemid,26/func,download/filecatid,2333/chk,c976623c899c2e022d5ed81d90c5a735/

I edited the timmingset 1 but if someone could test the autodetect to just make sure :up:

Still looking for mod to lower vmem

Thks!

[aussie-revhead]

HAPPY NEW YEAR everyone !!!!

(well its 1.20am 1st January here)

Have you guys found a particular nv driver works best with these ? I found 169.09 worked best with my GTx but this version doesnt recognise my GTS . I also had a similar problem with the newest 169.28 drivers , what are you guys using ? I got my rig to 16,050 at 3.9ghz , but it should be capable of more . Is there much to be gained by altering timings with nibitor ?

What sort of core clocks can I expect with volt mod on water ?

Any help would be great.

:up:

[VoRtAn_MaDgE]

Well seems to me that the vmem on my 8800gts 512 should be different than the generic.
Why?

Look at the ic at the back of my card ...
http://img108.imageshack.us/img108/825/pc310092ny8.th.jpg (http://img108.imageshack.us/my.php?image=pc310092ny8.jpg)

Already googled but no luck :(

Any info would be great ;)

Got to discover if my mem's are from qimonda too, to see if the mod should be lowering the vmem or not :cool: .
But i guess they should be qimonda...anyone with a 8800gts 512 from gigabyte to confirm this ?

found one guy with a 7100gs with that ic but no luck too i guess -http://forums.vr-zone.com/showthread.php?t=183215

[Stelios]

I'm on vista64 and the autodetect on vista64 doesn't works...already mencioned on mvktech forums...

The bios i used was this one - http://www.mvktech.net/component/option,com_remository/Itemid,26/func,download/filecatid,2333/chk,c976623c899c2e022d5ed81d90c5a735/

I edited the timmingset 1 but if someone could test the autodetect to just make sure :up:

Still looking for mod to lower vmem

Thks!
I have also the same gigabyte board and the timingset is No 1 .

[VoRtAn_MaDgE]

I have also the same gigabyte board and the timingset is No 1 .

Thks Stelios.

Do you have same ic on the back of the card in the place of the standard vmem ic from intersil ?

Is your card modded ?

[Stelios]

Thks Stelios.

Do you have same ic on the back of the card in the place of the standard vmem ic from intersil ?

Is your card modded ?
To be honest , i don't remember . I didn't bother to check it ... and now it's difficult to see it cause i have to remove the card .
I've done the vmem mods as showed in the 1st post of this thread and work great. We also have the same bios in the card . Your picture shows the same IC as in the first post in this thread .

What is the "standard vmem ic" ??? Do you mean the ISL6549CBZ ??? All G92 cards are supposed to have the same IC . Maybe it's relabelled .

[VoRtAn_MaDgE]

To be honest , i don't remember . I didn't bother to check it ... and now it's difficult to see it cause i have to remove the card .
I've done the vmem mods as showed in the 1st post of this thread and work great. We also have the same bios in the card . Your picture shows the same IC as in the first post in this thread .

What is the "standard vmem ic" ??? Do you mean the ISL6549CBZ ??? All G92 cards are supposed to have the same IC . Maybe it's relabelled .

Increasing vmem give you higher mhz on your gigabyte ?
Or you lower vmem? The shown in 1st post of this thread is the standard.

Maybe relabelled cuzz the reference is completely different as i shown 4 posts back.

Your card is just Vgpu/vmem modded ?
Which clock are you reaching? Cooling?

Thks again ;)

[Stelios]

Increasing vmem give you higher mhz on your gigabyte ?
Or you lower vmem? The shown in 1st post of this thread is the standard.

Maybe relabelled cuzz the reference is completely different as i shown 4 posts back.

Your card is just Vgpu/vmem modded ?
Which clock are you reaching? Cooling?

Thks again ;)
I have only tried to lower the vmem . Right now it's at 1.92v .I got the mod from jason4207 via PM , but i haven't asked him if i can publish it . The best way is to send him a PM for information . I don't either like this "secret" mod , but it's not my choice since it wasn't me who found it .

Right now i have the card watercooled with vgpu & reverse vmem mod . Vgpu is at 1.40V idle / 1.45v full load . I can bench at 950/1080Mhz everything . I haven't tried much , cause my proper benchmarks will be with both CPU & GPU in my singlestages( when i get some time though)
Yesterday i put some heatsinks at the voltage regulators (they were very very hot) ... no time to check for gains .

[VoRtAn_MaDgE]

I have only tried to lower the vmem . Right now it's at 1.92v .I got the mod from jason4207 via PM , but i haven't asked him if i can publish it . The best way is to send him a PM for information . I don't either like this "secret" mod , but it's not my choice since it wasn't me who found it .

Right now i have the card watercooled with vgpu & reverse vmem mod . Vgpu is at 1.40V idle / 1.45v full load . I can bench at 950/1080Mhz everything . I haven't tried much , cause my proper benchmarks will be with both CPU & GPU in my singlestages( when i get some time though)
Yesterday i put some heatsinks at the voltage regulators (they were very very hot) ... no time to check for gains .

I guess i've already found it on my own... :up:
Digged almost 1h30 through pdf and multimeter reading but i guess i'm correct now!

Gonna take a picture of my card and publish it here...of course is untested but i guess it's ok to put it here...if someone that already have it wants to confirm...be my guest!

Next...shutting off pc, taking my card off, picture, photoshop and upload.

Be right back

[jason4207]

The "vmem buck" (or "reverse vmem mod") is not "my" mod. The credit goes to Viper John. Credit also goes to him for the OCP mod on the GT.

I have simply forwarded the info John gave me to others via PM. I believe John has a link in the bottom of his posts if you need help on any of these mods. He's the one that really understands this stuff well. I can understand it once he explains it to me, but until then I'm still very much a novice. I am learning, though, and I have a BS in Computer Engineering, so I'm not completely in the dark.

I really don't like all this secret stuff either, but it's not my mod.

I'm waiting to get my GT back from RMA, and then I plan on doing the vGPU, vMEM buck, and OCP mod to it and putting the HR03GT on it. I might post up a nice thread on it once everything is up and running.

I'm debating modding my GTS-512. I paid $360 for it, and I might save it for step-up in a couple months. We'll see.

[Stelios]

The "vmem buck" (or "reverse vmem mod") is not "my" mod. The credit goes to Viper John. Credit also goes to him for the OCP mod on the GT.

My mistake i forgot it ... Viper John found these mods !:up:

[aussie-revhead]

team-eva doesnt exist??
We are very much a family at i4memory, not a one man band.


T_M , ..... rant removed .... sure thing buddy .


Dont bother with ram sinks for this card

This pic shows an area where the factory heatsink contacts some other chips , you dont worry about sinks on these ? With you running LN2 you probably freeze the whole card , but on air or water , and particularly with volt mods in place should these be heat sinked ? Or dont they get that hot ?

I just read a review and saw that the factory oc cards are running at 804/1080 , is this right ? I dont think I have seen a factory oc card that comes so close to the cards actual limits , usually even a factory oc card has a decent amount of headroom , but these dont seem to . I have also read of some guys that have problems with stability at clocks similar to the factory oc , does that mean these might be binned ?

Second pic shows a card from this forum with heatsinks fitted ...



:up:

[4Qman]

Im sorry but im a little confused.

I am at 915 core 1.39v IDLE 1.47v LOADED on stock air low ambient. BUT my memory craps out at 1040 and its reading 1.94v

SO......
Do i need to do Vmem mod and increase voltage to improve or lower to get better memory speeds?

Thanks

[T_M]

gotvoltage, 1.6V max sounds about right for max volts loaded.
Mine is 1.505V overvolting to 1.6V, and more than that is shutdown.

[aussie-revhead]

This pic shows an area where the factory heatsink contacts some other chips , you dont worry about sinks on these ? With you running LN2 you probably freeze the whole card , but on air or water , and particularly with volt mods in place should these be heat sinked ? Or dont they get that hot ?

T_M , would you be able to answer this for me please ?

;)

[T_M]

I have not sinked anything at all on the card, but when on DICE/LN2 i have a strong delta blowing onto the power circuitry

[Zeus]

Are there any components that need extra cooling if you're going to push some more Vgpu?

I will put an EK fullcoverblock on it so i think the core-side of the card will be cooled properly.

[Trust]

Owners of 8800GTS (G92), please try to help me: http://www.xtremesystems.org/forums/showthread.php?t=171215

[VoRtAn_MaDgE]

Are there any components that need extra cooling if you're going to push some more Vgpu?

I will put an EK fullcoverblock on it so i think the core-side of the card will be cooled properly.

Mem's, power circuit and there's also a ic on the back of the card that gets hotter than the rest.

Negative temps have a major impact on pcb temp...that's why users with that kind of cooling don't worry with that issue.

[Zeus]

VM, thanks, i will poke around with my indexfinger on the back feeling what needs some extra cooling.

Just did the Vmod for GPU with a 1K pot i had laying around, works fine but it's hard to get it stable when temps go over 50ºC.

Had 850 core 2000 shaders with 1.3V (load- so 3d voltage) but got some yellow dots in ATI tool, dang fan was at 100%, loud as hell. :D

I'm sure watercooling will take care of that.

What do you get with the stockcooler?

[VoRtAn_MaDgE]

VM, thanks, i will poke around with my indexfinger on the back feeling what needs some extra cooling.

Just did the Vmod for GPU with a 1K pot i had laying around, works fine but it's hard to get it stable when temps go over 50ºC.

Had 850 core 2000 shaders with 1.3V (load- so 3d voltage) but got some yellow dots in ATI tool, dang fan was at 100%, loud as hell. :D

I'm sure watercooling will take care of that.

What do you get with the stockcooler?

When you go above 1,35v gpu take care using 1kohm...little bit more difficult to jump volts in small steps...take more carefull turning the pot.

On air (stock cooled) 900 (almost reaching 910) is possible with a good core and 1.45 (loaded) or more...depends on each vga of course!

above 52, 53ºC it's hard to stabilize the vga aircooled...wc gives a much big advantage playing around with this card cuzz the difference for stock cooling loaded is huge....

There are also aftermarket good aircoolers for those babies...but wc it's a must when it comes to loaded temps!!!

I'm considering getting an external radbox system with watercooling just for single vga/sli, crossfire system... just need to update waterblock in the future ;).

Don't need it internal cuzz i don't use any case... like to use the hole system + prommy on the desk :D

[aussie-revhead]

Can I use a 500 ohm VR as T_M suggested ? Will this make adjustment easier ? I used it on my GTX but the voltages were way too high so I had to remove it , so I have a spare laying around .

:up:

[Zeus]

500ohm VR will make it a little less "jumpy" when adjusting over ~1.35V.

When you're using 1K VR you have to be more carefull turning the pot cause a little dial means more voltage than with a 500ohm over ~1.35V.

I did set my 1K VR to 500ohm before soldering so a 500ohm VR is even better than 1K actually.

[aussie-revhead]

How much core clock do you get with ~1.35v ? At what point do I need OVP mod ? Im about to search for it but could you point me to the OVP mod ?

;)

[Zeus]

I think it depends on your cooling, anything over ~50ºC isn't stable for me, i'm waiting for my waterblock to arrive so i can get temps to decent levels, mine idles at 58-60ºC with the stock cooler, stock volts. :eek:

In a cold ambient with 1.35V you should see around 850 core i reckon.

Overvolt protection kicks in around 1.6V so unless you're planning to go xtreme there's no need to mod for that.

[aussie-revhead]

Cool !!

I wasnt planning to go that high as I dont really want to fry it up , but a little vmod fun wont hurt :D

I am planning a two stage mod with a small jump for 24/7 use and a large one for benching . I had this on my last GTX and it worked well . My G92 games fine at 800 core so far and benches up to 820 , so I hope I get more than 850 at 1.35v.

:up:

[T_M]

To avoid some possible missunderstanding:

OVP = Over voltage protection (typically if you reach this point the system will freeze/reboot even with no load)
OCP = Over current protection (typically if you reach this point the system will freeze/reboot only under load)

On G92 8800GTS, what we see is likely (i havent tested) to be OCP, which kicks in around 1.6V and high clocks under load.
This is not the idle voltage however, due to overvoltage under load, so you actually can only max out your voltage to about 1.5-1.525V idle.

[Zeus]

So here is 850 core 2100 shaders with 1.35V and stockcooler fan at 100%but becomes unstable over >50ºC:

[nickless]

If there is allready a pencil Vmod, can someone post a pics?!
Thanks!

[aussie-revhead]

Thanks T_M , that explains it perfectly .

:up:

Kako si Nickless - Im pretty sure no pencil mod exists , not yet anyway .

;)

I finished my volt mod last night but that vgpu point is hard to see even with the magnifying glass ... The EK block will go on today .....

:D

[Zeus]

Thanks T_M , that explains it perfectly .

:up:

Kako si Nickless - Im pretty sure no pencil mod exists , not yet anyway .

;)

I finished my volt mod last night but that vgpu point is hard to see even with the magnifying glass ... The EK block will go on today .....

:D

Stop!!!

If you dailed that VR to 500ohm, it will be 0ohm on the other end, meaning putting the switch in the wrong position results in 0ohm and way to much volt on the feedbackpin of that chip which could possibly kill your card!!

Why did you use both "variable" legs on the VR?

I always cut of one of te outer legs of the VR so i won't make a mistake.

I hope it's not too late.

That block is lovely, i ordered it december 19th but still haven't got it. :(

To anyone who knows: Is there any difference in overclocking between the lot of bios versions that are out there, or does a different bios number just mean a different vendor? :shrug:

edit: on a closer look it seems like you are using 2 VR's stuck onto each other?

[jason4207]

It looks like the VR has 2 legs at each of the 3 positions. And it looks to me like he can switch the mod on and off. Middle is 'off', and either direction results in 'on'.

[Zeus]

It looks like the VR has 2 legs at each of the 3 positions. And it looks to me like he can switch the mod on and off. Middle is 'off', and either direction results in 'on'.

My guess is that it's 2 VR stuck together so that he can set either position of the switch to a diffent value.

For example, switch to the left is 1.35V and to the right is 1.5V, the middle is off.

[VoRtAn_MaDgE]

Middle off.
left or right depends on which pot he wants to use!

These vga's above 52/53 aircooled seems to have an issue

I measured temps using rivatuner on non-sucessfull 3dmark06 runs... all of them (the non-sucessfull's) are due to driver restart...then i check the temp on rivatuner...and there it is...53ºC

Zeus...stock bios from each manufacters are the same...only differences on the clock's (overclocked versions) and the name of the vga manufacter... already look deep into 4 (gigabyte, xfx, gainward and bfg) all same!
Mem timmings are just exactly the same!

Can anyone express in terms of mhz the gains reducing vmem?
I'm not seeing any in my case!

[Zeus]

Middle off.
left or right depends on which pot he wants to use!

These vga's above 52/53 aircooled seems to have an issue

I measured temps using rivatuner on non-sucessfull 3dmark06 runs... all of them (the non-sucessfull's) are due to driver restart...then i check the temp on rivatuner...and there it is...53ºC

Zeus...stock bios from each manufacters are the same...only differences on the clock's (overclocked versions) and the name of the vga manufacter... already look deep into 4 (gigabyte, xfx, gainward and bfg) all same!
Mem timmings are just exactly the same!

Can anyone express in terms of mhz the gains reducing vmem?
I'm not seeing any in my case!

Not sure on that mate, i've seen al lot of different versions, so far i've seen:

62.92.16.00.09
62.92.16.00.25
62.92.16.00.26
62.92.16.00.39
62.92.16.00.43

I'm just wondering if there's any difference.


What do you mean by driver restart?

My card gets unstable at over 50ºC, yellow dots appear in ATItool.

I'm awaiting my EK fullcoverwaterblock so i can finally max the card out, with the stockcooler it seems impossible because it reaches 50ºC rather quick.

[VoRtAn_MaDgE]

Not sure on that mate, i've seen al lot of different versions, so far i've seen:

62.92.16.00.09
62.92.16.00.25
62.92.16.00.26
62.92.16.00.39
62.92.16.00.43

I'm just wondering if there's any difference.


What do you mean by driver restart?

My card gets unstable at over 50ºC, yellow dots appear in ATItool.


I'm awaiting my EK fullcoverwaterblock so i can finally max the card out, with the stockcooler it seems impossible because it reaches 50ºC rather quick.

As i said before, from the bios models that i already tried the differences are the clocks and the manufacter.
Driver restart is when the vga driver sef-restart caused by a fault (overclock for example)
You'll get that information on your traybar when it happens!

[Zeus]

OK thanks, i was just wondering.

Makes it easier for me so i don't have to flash the cards bios. :)

I've never had a driver restart happen to me, if overclocked too far it just locks in 3d mark.

How easy is the Vmem reduction mod, is it soldering a pot or is it just a pencil mod?

[VoRtAn_MaDgE]

How easy is the Vmem reduction mod, is it soldering a pot or is it just a pencil mod?

Hard mod...no pencil.

[aussie-revhead]

To avoid some possible missunderstanding:

OVP = Over voltage protection (typically if you reach this point the system will freeze/reboot even with no load)
OCP = Over current protection (typically if you reach this point the system will freeze/reboot only under load)

On G92 8800GTS, what we see is likely (i havent tested) to be OCP, which kicks in around 1.6V and high clocks under load.
This is not the idle voltage however, due to overvoltage under load, so you actually can only max out your voltage to about 1.5-1.525V idle.

Hi guys , T_M - in regards to the information you showed me above , I have volt modded and water cooled my G92 GTS and it has responded very nicely , but as soon as I push vgpu above 1.45v I get gpu lockup in 3D06 , I can escape out of it most of the time and it will recover but it doesnt respond to more voltage , if anything it gets more unstable , is this OCP kicking in ?

I can bench up to 900 core 2250 shader and 1150 mem at 1.4vgpu , is this good ? I managed to score 17887 in 3D06 with my main goal to break into the 18k range but this issue seems to have halted my progress , unless I go back and try to get some more out of my cpu but its not a special one and will only bench to 9x435=3915mhz .

Any input would be great.

:up:

[dinos22]

that's awesome man

you are doing well with the OCs for sure :D

nice modding too.....very neat

only thing i could suggest is try backing off on RAM OC and see if your core/shader go higher
also try backing off on core a bit and see if you can get some bigger jumps in shader

bench 3DMARK05 for stability first as it's really tough on these cards :)

[Zeus]

Hi guys , T_M - in regards to the information you showed me above , I have volt modded and water cooled my G92 GTS and it has responded very nicely , but as soon as I push vgpu above 1.45v I get gpu lockup in 3D06 , I can escape out of it most of the time and it will recover but it doesnt respond to more voltage , if anything it gets more unstable , is this OCP kicking in ?

I can bench up to 900 core 2250 shader and 1150 mem at 1.4vgpu , is this good ? I managed to score 17887 in 3D06 with my main goal to break into the 18k range but this issue seems to have halted my progress , unless I go back and try to get some more out of my cpu but its not a special one and will only bench to 9x435=3915mhz .

Any input would be great.

:up:


Do you mean 1.4Vgpu 2d or 3d? Cause 1.4V 2d is 1.45~1.47V 3d.

What clocks did you get prior to the Vmod and what temps do you get now with watercooling?

I think 900/2250 is very nice, most top out around 900-915 core with voltmod and around 2200 shaders.

Also 1150 on the mem is very very good. :up:

I hope to reach the same when i can finally strap my EK blok onto the card.

[dinos22]

that is definitely highest RAM OC i have ever seen no matter the cooling considering how hard he's pushing the core/shader

[aussie-revhead]

Thanks guys , but seeing these guys benching with 45nm at 4.5-5 GHz is making my wallet hurt with anticipation :rofl:

With stock volts the gpu temps top out at 43 degrees under load in a fairly warm case (currently 40 degrees). With 1.3v it tops out at 49 degrees , 1.4 pushes load temps to 50 and 1.5 topped out at 52 degrees , but then I get the above mentioned graphical lockup .

I will add a pic when I get them out of the other pc .(its busy benching)

Dino - the mem benched happily at 1100 before I modded anything too , with stock cooler , I dont think I pushed it any higher .


Do you mean 1.4Vgpu 2d or 3d? Cause 1.4V 2d is 1.45~1.47V 3d.

Measuring with DMM set in bios or idle , it does rise a little under load but not that much iirc.

:up:

below is a pic - its a poor angle but you can just see my toggle switch for dual vgpu settings and 2x VR's . Then there are just some pics showing how good the EK block looks .


:D

[pumbertot]

bloody awesome mate, good to see an 'amateur' o/c compete with the top clockers. ;)

[Zeus]

that is definitely highest RAM OC i have ever seen no matter the cooling considering how hard he's pushing the core/shader

Yeah, that ram clock is pretty awesome, mine does 1130 unmodded with the stock cooler, i hope to reach 1150 with my EK block. :)

[T_M]

My guess is you are topping out due to heat.
Independently clock your core, then your shader, to find out which one is limited at those volts if any.

[aussie-revhead]

T_M , I have seen what you and others have said about temps but how is this possible when the stock cooler generates much more heat without failure ? The difference in my loaded temps between 1.3-1.4-1.5v is very small . Also , clocks that will pass at 1.4v fail at 1.5v without making more heat , in fast even 800 fails when vgpu is set to 1.5 . Based on these I assumed it sounds very much like OCP kicking in . Another point is - at 1.3 or 1.4v I have not found a gpu clock that locks up , only scores start to drop again , where at 1.5 it goes berzerk regardless of clocks .

:shrug:

Hey Paul - thanks for the encouragement .

:rolleyes:

[T_M]

Well, have you tested what the loaded vGPU is when it crashes?

[Zeus]

Don't think it's the ocp kicking in, when that happens the card shuts down and the system freezes or crashes/reboots i think.

I have a feeling the vreg circuit can't take it due to noise or something.

In the 8800GT voltmod thread ther was a picture of a card with added caps, maybe that will help yours to clock higher at higher voltage?

I just recieved an email that my EK block will be send on wednesday (finally!!), as soon as i have it mounted i will do some testing with mine, Vmod is already done. ;)

[STEvil]

Oh how I love macro....

Easy breezy to see where you can and can not solder ;)

edit

darn this is hard to do with a gas powered iron and huge tip lol ;)

well, kinda hard to make it stick to that point really is all.

[dinos22]

lovely macro man
same camera or did you get a different one

[STEvil]

Still the Canon SX100IS

edit

added after-mod pic

[T_M]

Did you intend to solder to almost every component/pad in a 1 meter radius of the mod point? :P

[STEvil]

Yeah pretty much

[Timbosan]

Oh how I love macro....

Easy breezy to see where you can and can not solder ;)

edit

darn this is hard to do with a gas powered iron and huge tip lol ;)

well, kinda hard to make it stick to that point really is all.

wow dude...... maybe its time to invest in a normal iron lol.

[dinos22]

wow dude...... maybe its time to invest in a normal iron lol.

or pair of steadier hands :D
heheh
hides under rock

[STEvil]

I actually had it on the normal spot fine (that little dot) but I didnt like it.. didnt want it to rip off too easily.

Which is funny because I packaged the card back up after doing the mods and took it to the PC it was going to be installed in (different house several miles away) and since i've installed the card I have found that the mod for measuring the memory voltage came off (due to putting it back in the original packaging)... didnt look really close to see if it took the cap with it though.. dont want to look lol (edit - I looked, cap is still there so i'll blame a poor soldering job as part of it).

Easy fix really, but no iron here..

Also, since this is still about volt modding, my card (XFX 8800GTS 512mb Alpha Dog Edition) is running 1.18v idle and UP TO 1.225v under load... average about 1.20v under load (at stock)

vmem is exactly 2.00v.

Also, this might sound like a dumb question... but what the heck do I use to overclock this thing.....?? Rivatuner wont give me any tabs to work with like it normally does... lol....

[T_M]

You need to edit the .cfg file in the Rivatuner folder.

http://www.xtremesystems.org/forums/showpost.php?p=2617965&postcount=64

[STEvil]

Thanks T_M

[aussie-revhead]

T_M , I cant test it now but iirc it was pretty close to that which I set it at , maybe a little higher . As I was saying I can bench at 910 at 1.4v with 51 degrees temp but cant complete a test at 850 / 1.5v 52 degrees , then again it probably doesnt even get that hot as it doesnt even take a load , or maybe for 1-2 seconds.

:up:

[T_M]

Load volts are higher than idle, by as much as 0.1V.
Thats why im asking what are your load volts when the card freezes.

[aussie-revhead]

I will test it when I can , but at 1.3v in bios I get 1.33v loaded so it rises but not 0.1v at this voltage .

Yes I get that driver reset thing when it crashes - nv failed to repond - or something to that effect - please restart pc to regain full function . This is sometimes accompanied by display resolution setting itself to 640x480 .

How can I futher reduce loaded temps ? Im using a dedicated gpu loop with 1x360 rad , EK full block and a little DB-1 pump . I was told a bigger pump wont reduce temps .

Where can I find the OCP mod for these ? Is it hard to do / worth doing ?

:up:

[Zeus]

I will test it when I can , but at 1.3v in bios I get 1.33v loaded so it rises but not 0.1v at this voltage .

Yes I get that driver reset thing when it crashes - nv failed to repond - or something to that effect - please restart pc to regain full function . This is sometimes accompanied by display resolution setting itself to 640x480 .

How can I futher reduce loaded temps ? Im using a dedicated gpu loop with 1x360 rad , EK full block and a little DB-1 pump . I was told a bigger pump wont reduce temps .

Where can I find the OCP mod for these ? Is it hard to do / worth doing ?

:up:

I don't know that pump but if it's very weak it might help to replace it.

You can try some more powerfull fans on the rad.

How's the contact with the block and what thermal compound did you use?

Also, a cooler ambient helps. :D

Today i ran 3d mark 01 and 06 at 1.35Vgpu with the stock cooler at 881 core 2125 shaders and 1130 memory, so that's looking good.

Aiming for 900/2200/1150 with watercooloing though.

[cole2109]

I have only tried to lower the vmem . Right now it's at 1.92v .I got the mod from jason4207 via PM , but i haven't asked him if i can publish it . The best way is to send him a PM for information . I don't either like this "secret" mod , but it's not my choice since it wasn't me who found it .

Right now i have the card watercooled with vgpu & reverse vmem mod . Vgpu is at 1.40V idle / 1.45v full load . I can bench at 950/1080Mhz everything . I haven't tried much , cause my proper benchmarks will be with both CPU & GPU in my singlestages( when i get some time though)
Yesterday i put some heatsinks at the voltage regulators (they were very very hot) ... no time to check for gains .


Hmmm ok now i don't understand :confused:
Why do u need to do Vmem mod and lower voltage?

[STEvil]

zues - stock cooling?

I was running 850/2125/1000 (havent worked on mem a lot yet) with 1.35v idle on my XFX 8800GTS but got random game crashes after periods of time. I could run the shaders up to 2250 at that voltage but got artifacts ;)

[Zeus]

zues - stock cooling?

I was running 850/2125/1000 (havent worked on mem a lot yet) with 1.35v idle on my XFX 8800GTS but got random game crashes after periods of time. I could run the shaders up to 2250 at that voltage but got artifacts ;)


Yup, stockcooler but fan @ full throttle, loud as h3ll. :down:

[gOtVoltage]

I don't know that pump but if it's very weak it might help to replace it.

You can try some more powerfull fans on the rad.

How's the contact with the block and what thermal compound did you use?

Also, a cooler ambient helps. :D

Today i ran 3d mark 01 and 06 at 1.35Vgpu with the stock cooler at 881 core 2125 shaders and 1130 memory, so that's looking good.

Aiming for 900/2200/1150 with watercooloing though.

Im testing another new GTS..after selling my 907mhz Aircooled card to nieghbor!

This one does 950//2302//2250mhz so far with normal water and 1.44Vgpu..:D

Working out bugs with my M2N32 deluxe ...The Bios seems to be bugged with HT and Multi problems..But i managed to get this new 1 Day old card up to par:up:

http://i168.photobucket.com/albums/u192/darrellsierra/WoWater950.jpg

[aussie-revhead]

Nice , much better than mine , shame your AMD will hold back your score , what was your score by the way ? ( you covered it )

:cool:

[dinos22]

hey revhead
man you gotta do some mods for me :D i like your work there :D

[aussie-revhead]

Thanks Dino , I like to make things and make them nice .

Do you want my GTS ? Im bored with it already .... I need a better cpu . Im looking forward to trying a q9450 and a G100 , but thats down the track a bit.

:up:

[dinos22]

Thanks Dino , I like to make things and make them nice .

Do you want my GTS ? Im bored with it already .... I need a better cpu . Im looking forward to trying a q9450 and a G100 , but thats down the track a bit.

:up:

PM me da price

[Zeus]

OMG, 950/2300 and even good mem clocks! :shocked: Did you manage to finish 3d mark at those speeds?

What does your coolong look like?

[cole2109]

GPU 1.3V (load) game stable for now...
Accelero + 120mm

http://shrani.si/t/3v/fP/1zcs4Rew/sss.jpg (http://shrani.si/?3v/fP/1zcs4Rew/sss.jpg)

[gOtVoltage]

OMG, 950/2300 and even good mem clocks! :shocked: Did you manage to finish 3d mark at those speeds?

What does your coolong look like?

My 3Dmark06 suffers from my X2...I got stuck on the 1603bios which hampered my results by 1000+ points and even with stock clocks on the Card it will do this...

So i may get a Phenom pretty soon, although im holding out for a B3version!

Im still tweaking Mobo ...My best 3dmark06 was 13128+ with 2800HT on the 1201bios @907//2160//2250,, but im now limited to 2500 and high Multis with the 1603bios.

My bench with 950:yepp: The FPS are Nice ..My CPU used to score 2700+ but new bios has me under 2650! The bios really afects the score bad..Trust me its not the High Oc on card..It scores lower than the 1201 even with stock card clocks...

http://i168.photobucket.com/albums/u192/darrellsierra/33cloadBios1603beta.jpg

[T_M]

For anyone that missed it, yantronic over at VRZone has posted the OCP mod.
I have not tested it but will soon.

"To castrate the overcurrent protection, get rid of the three resistors marked with "78X" on them"

http://sg.vr-zone.com/articles/XFX_8800GTS_Alpha_Dog_Overclocking_Adventure/5462-3.html

[Zeus]

For anyone that missed it, yantronic over at VRZone has posted the OCP mod.
I have not tested it but will soon.

"To castrate the overcurrent protection, get rid of the three resistors marked with "78X" on them"

http://sg.vr-zone.com/articles/XFX_8800GTS_Alpha_Dog_Overclocking_Adventure/5462-3.html

Thanks for the heads up, did you try it yet?

That added cap looks huge! :shocked: :D

[Lastviking]

Nice scores...

It will be funy to se how much my old 939 with bh5 ddr1 will score with 8800 gts when i get it on monday :)

[aussie-revhead]

What does this mean guys ? I made a large change to my fan arrangement on my radiators but I still cant lean on it with any more than 1.4vgpu is it either has a graphical lockup or gives the error shown in the screenie :down:

Im still nudging 17800 area but I want to crack an 18k score , then I can game in peace .

:clap:

[aussie-revhead]

I don't know that pump but if it's very weak it might help to replace it.

You can try some more powerfull fans on the rad.

How's the contact with the block and what thermal compound did you use?

Also, a cooler ambient helps. :D

Today i ran 3d mark 01 and 06 at 1.35Vgpu with the stock cooler at 881 core 2125 shaders and 1130 memory, so that's looking good.

Aiming for 900/2200/1150 with watercooloing though.

I did a test run on the EK block to test contact and it seemed nice to me . I used Arctic Silver. I changed from one huge fan to 9x120 fans , but its still pretty much the same . The case is stuck under my desk and it tends to build up heat under there even with the a/c on , so Im sure this would help too .

T_M I tested vgpu-rise under load and at 1.3 static I get 1.35 under load , while 1.5 tops out at 1.58v . I think the dreaded 50 degree limit is costing me progress.

:(

[VoRtAn_MaDgE]

T_M I tested vgpu-rise under load and at 1.3 static I get 1.35 under load , while 1.5 tops out at 1.58v . I think the dreaded 50 degree limit is costing me progress.

:(

I'm having the same issue...

Almost almost making a 06 run at 910 core but temp won't allow me...
very temp sensitive the core and the vrm's...improve cooling
Did you already noticed at what voltage you get the error (ocp maybe?)

[aussie-revhead]

I asked T_M the same thing but he sounds pretty sure its not OCP related. It seems that as soon as temp hits 50 degrees it dies , but why when it runs much hotter than this with the standard cooler and much lower clocks ?

Anyway its about at its limit in my current configuration . I Will have to wait and see what some extra cpu score does in a few weeks time .

:up:

[VoRtAn_MaDgE]

I asked T_M the same thing but he sounds pretty sure its not OCP related. It seems that as soon as temp hits 50 degrees it dies , but why when it runs much hotter than this with the standard cooler and much lower clocks ?

Anyway its about at its limit in my current configuration . I Will have to wait and see what some extra cpu score does in a few weeks time .

:up:

Dunno if it's all relative to the temp cuzz i can play with card oc'ed above 55ºC for example with 0 problems!

The load must be bigger in bench than in games so maybe that's the cause...

[jason4207]

For anyone that missed it, yantronic over at VRZone has posted the OCP mod.
I have not tested it but will soon.

"To castrate the overcurrent protection, get rid of the three resistors marked with "78X" on them"

http://sg.vr-zone.com/articles/XFX_8800GTS_Alpha_Dog_Overclocking_Adventure/5462-3.html

I was under the impression that doing the OCP mod on the GTS was kinda pointless unless you are going for very high volts (requiring exotic cooling). Does anyone have any info to the contrary?

Also, if you are going to do the OCP mod you should solder resistors in parallel to the caps, not remove resistors. I know removing them works, but it also upsets the phase balancing. Someone on here already killed a MOSFET doing it that way. Comments?

They are also increasing vmem in the article. I know that decreasing vmem works better on the GT, and I assume the same to be true on the GTS. Comments?

I do like the caps they added, though!

[cole2109]

Not bad for dual core on 3700MHz :cool:

http://shrani.si/t/2r/zB/3ovysIGs/3dm-153k.jpg (http://shrani.si/?2r/zB/3ovysIGs/3dm-153k.jpg)

Need good Quad :eek:

[NightRaven]

I was under the impression that doing the OCP mod on the GTS was kinda pointless unless you are going for very high volts (requiring exotic cooling). Does anyone have any info to the contrary?

Also, if you are going to do the OCP mod you should solder resistors in parallel to the caps, not remove resistors. I know removing them works, but it also upsets the phase balancing. Someone on here already killed a MOSFET doing it that way. Comments?

They are also increasing vmem in the article. I know that decreasing vmem works better on the GT, and I assume the same to be true on the GTS. Comments?

I do like the caps they added, though!


well that is also true.. but yan used 1.5vgpu on air to be exact..

hmmm well sometimes engineers from respective manufacturers do tell us to remove certain resistors to get certain stuff.. so i will suppose it is relatively safe but not recommended at all. perhaps our friend was just unlucky.

to be honest i have had no gain with increasing volts for mem. havent tried decreasing but it may work. personally rather just leave it.

[VoRtAn_MaDgE]

to be honest i have had no gain with increasing volts for mem. havent tried decreasing but it may work. personally rather just leave it.


Lowering vmem for me gave me no beneficts at all...same as default.
Gonna try more volts to see what happens :rolleyes:

[NightRaven]

haha all right... i'm guessing no benefit too but let us know :)

[gOtVoltage]

What does this mean guys ? I made a large change to my fan arrangement on my radiators but I still cant lean on it with any more than 1.4vgpu is it either has a graphical lockup or gives the error shown in the screenie :down:

Im still nudging 17800 area but I want to crack an 18k score , then I can game in peace .

:clap:

Check your water temps and or Nvidia Driver..

Driver Error..Ive had that before with unstable driver..try installing newer Beta or roll back one driver..Ive noticed some drivers dont like to overclock past acertain point!

i just updated to 169.28beta for XP pro x64bit..Gets same clocks as the 169.21whql but is stable...The 169.21 would crash my pc sometimes while using 3Dmark06..169.28beta seems 100%stable at least for me..

Up date on My Vgpu modded GTS Evga512mb

3Dmark06 with 3.5ghz 1603beta bios and 1000mhz @4-4-4-12-2T TRC21/2.3vmem HT//ram setting sucked balls!


950//2322//2250 @1.528Vgpu and stock memory Vmem!Before finding the optimal memory and HT tweaking in 1603bios..1603bios really makes you start from scatch on the memory and HT settings:yepp:

http://i168.photobucket.com/albums/u192/darrellsierra/33cloadBios1603beta.jpg

Ive since done some tweaking and now my card sings @3.5 with optimized memory @only 388mhz:D 1T timing 4-4-4-12-TRC21 1T @2.3vmem..."Im Cpu limeted"..Though id never say that line with a X2 6000+:DStill tweaking with settings ,,Theres more too get outa this thing yet:D

http://i168.photobucket.com/albums/u192/darrellsierra/950gpu13165.jpg

[aussie-revhead]

Yes it seems that I hit a driver wall with overclocks , I have 169.28 but I had an issue with that too , so I will go back to it and try again , it just feels like the gpu is capable of so much more ....

:up:

Have you seen or tried this driver ?

http://downloads.guru3d.com/download.php?det=1818

Title says 169.28 WHQL but on the page it says 169.29 .

:shrug:

[andrew2004gto]

:confused: Hey i have a bfg 8800gts 512 well i did a volt mod and i cant get more than 1.3 volts top clock is 833/2052/1110 did a 14600 but anyway when i try to turn more volts the screen goes blank or freezes thought maybe ocp but i removed that nothing, could it be my powersupply that is only a 585, 2 rail at 20watt, with 6 fans, 2x2gb at 1052, 680i lite, e4300@3550@1.55volts with 1.4v

[aussie-revhead]

Carefully check your solder points mate , that gpu point is tiny and very close to other components. My card clocks to 820 before vmod and at 1.3 will pass at 900 . It wont be ocp at that voltage . What are your temps like ?

:up:

[andrew2004gto]

My top temp that i recorded using rivatuner was 68 while play cryis for an hour, i tried to set my computer back to stock so it would take less power nothing, trying increasing pcie nothing, when i try to increase vols and using atitool the whole thing freezes and when i lower a bit it artifacting like crazy but than about another half turn to lower volts its fine again at 2052shader but wont go high without artifacting like crazy, and what u mean about check my soldern, like what do u want me to look for?

[aussie-revhead]

The solder point is tiny , are you sure you didnt accidentally bridge to a neighbouring component ? I soldered mine under a magnifying glass and it was still hard to see. Most people are reporting problems with any temps over 50 , so Im not sure if this is part of your issue.

Did it run OK before you modded it ? What did it clock like before mod ?

:up:

[andrew2004gto]

:shrug: the thing is i only solder was the the negitive the one that gets close to the little chip i just let it touch it and i super glued it on the side but anyway would this help u if i told u i had a 8800gt evga and i had the same problem but just at 770/1900/1040, but had to rma cuz the ram went to :banana::banana::banana::banana: and artifact like that one screen shown in the 8800gt vmod fourm of him playing bioshock, so i didnt want to take a chance and get another 8800gt so i got the bfg, but i i get 28.5fps in crysis bench on very-high and at 1680x1050, i guess its not to bad and yea about over-heating wouldnt u think the chips it cools would get really hot since this whole thing is conect

[aussie-revhead]

:shrug: the thing is i only solder was the the negitive the one that gets close to the little chip i just let it touch it and i super glued it on the side ...


WTF dude ? I seriously doubt you will get a good connection like that ... attend to this immediately .

:shocked:

[andrew2004gto]

no that cant be it it wouldnt freeze up on me like that, it rases volts it freezes about 1.32 on load i try to keep it under i think is getting to hot

[jason4207]

Your connection is probably bad due to the poor conductive properties of super glue, and that is adding resistance (why you can't get your vcore any higher). It also may be micro arcing and causing stability issues.

Use a soldering iron, or leave it alone.

[Zeus]

Glueing wires? omg. :down: Nice clocks once again, gOtVoltage, can you show a pic of those clocks running ATI tool artifact scanning?

[aussie-revhead]

Does anyone else have an EVGA G92 ? Can you try to match the clocks of gotvoltage ? Mine is a GALAXY budget card so maybe thats why mine doesnt clock as well .

;)

[STEvil]

820/2000 isnt that bad, but should be doable at ~1.3v idle for 100% stability

Mine does 800 on stock volts but will crash eventually. 1.25v idle gives me stability at 800 (probably a bit more, havent tested higher at 1.25).

[cowie]

these boards are so much fun to v-mod
must do better on cpu/mem clocking for better scores but i love this card.
evga,stock volt overclocks 825/2100/1100
with only 1.33 at load ,mem stock 1.97
918/2250/1100 runs all 3dmark benches
lapped stock heatsink
http://aycu35.webshots.com/image/38314/2002571134449035372_rs.jpg (http://allyoucanupload.webshots.com/v/2002571134449035372)
http://aycu34.webshots.com/image/41753/2006144690999678700_rs.jpg (http://allyoucanupload.webshots.com/v/2006144690999678700)
first 88 cards to run 900+ air you have to love it

[aussie-revhead]

Cowie - where is your 06 score ? Why is your idle temp so low ? Is it snowing there ? :rolleyes:

Why is it the G92's with aftermarket cooling wont run over 50 degrees yet the stock cooler does ? Could they be set up to monitor fan function and set a lower gpu-temp threshold when fan speed is low ?

:up:

[cowie]

i not try to show my scores but they come in screen shot:p:
.....well cold air helps so i make cold the air ;)
why would i want put volts in card if you no keep cool?
this is extreme right? so extreme cold i make air for stock heatsink.
right now amb temp is 67f in room fan 100%
http://aycu21.webshots.com/image/40700/2004469967555614440_rs.jpg (http://allyoucanupload.webshots.com/v/2004469967555614440)
i not just jump to higher volts ok?
i rub it just right:)
\stock fan is very good
lenth of fins and cover are its 'tricks'
might be stock fan but its modded EXTREME:)

[Zeus]

Finally got my EK fullcoverblock in, (have to update my sig :D ) it was a bit hard to get perfect contact but eventually i managed.

So, with a rediculous amount of bubbles in the loop i started rasing voltage and never saw over 43ºC full load.

The air has to bleed out which takes several days and the thermal compound (MXII) has to settle a bit so temps will go down and i have to say ambient was pretty high at the time of testing.

I was testing the floorheatingsystem so i reckon ambient was up to ~23+ºC, usually it's ~20ºC here.

So far at 912 core 2205 shader and 1128 memory ATItool ran without errors but after 2 minutes some yellow dots started to appear.
This was with 1.47V idle 1.51V load.

I could go a little higher on the core but 928 locked up in 3D, shaders could go as high as 2250 but started artifacting there.

All in all, only a brief test but i'm quite happy. :)

[gOtVoltage]

Finally got my EK fullcoverblock in, (have to update my sig :D ) it was a bit hard to get perfect contact but eventually i managed.

So, with a rediculous amount of bubbles in the loop i started rasing voltage and never saw over 43ºC full load.

The air has to bleed out which takes several days and the thermal compound (MXII) has to settle a bit so temps will go down and i have to say ambient was pretty high at the time of testing.

I was testing the floorheatingsystem so i reckon ambient was up to ~23+ºC, usually it's ~20ºC here.

So far at 912 core 2205 shader and 1128 memory ATItool ran without errors but after 2 minutes some yellow dots started to appear.
This was with 1.47V idle 1.51V load.

I could go a little higher on the core but 928 locked up in 3D, shaders could go as high as 2250 but started artifacting there.

All in all, only a brief test but i'm quite happy. :)

Dont go above 2250 on the memory...Ive tested 5 cards in arow and they all get funny after that...

Ive seen the GT is now getting samsung memory which can handle 2300mhz @1.0ns.. Id like to get a new GTS with the Samsung then this card will come alive...bandwidth kills it...If it had 90b/w instead of 72.b/w it like having two totaly different cards..

I try ATi tomorrow ..need to download it for XPx64bit:up: Just redid my waterloop for CPU after modding my D-Tek..To 7 mm jet:D

:up: That Ek is real nice:up: ...:yepp: the oc is even better! :yepp: Im using MCW60 with ramsinks...I may make custom copper sinks with 3/8copper tube after testing one more card...I braze 3/8 tubes onto flat copper and flare the ends to adapt to water!Then ill tie the memory t into my MCW60 loop:D

[Zeus]

Oh really? Funny like what? Artifacting or something?

Just benched 06 at 900/2200/2300, 1.41V idle so around 1.46V load:

http://img.photobucket.com/albums/v620/Z3us/0615935.jpg

If only i had a quadcore. :)

[T_M]

I was under the impression that doing the OCP mod on the GTS was kinda pointless unless you are going for very high volts (requiring exotic cooling). Does anyone have any info to the contrary?

Exotic cooling is all mine runs


Also, if you are going to do the OCP mod you should solder resistors in parallel to the caps, not remove resistors. I know removing them works, but it also upsets the phase balancing. Someone on here already killed a MOSFET doing it that way. Comments?

Can you draw what you mean on a photo please.

[VoRtAn_MaDgE]

Can you draw what you mean on a photo please.

Sorry about using your pic...too laisy to pic my digicam and shut off my system :D
Something like this T_M,
check blue rectangles
http://i4.tinypic.com/82xun2u.jpg

[largon]

It's the same principle as for G80GTS:

This pic is of a GT so there's no SMD for 3rd phase...
http://img228.imageshack.us/img228/603/g92ocpyz8.jpg

[andrew2004gto]

:mad: I didnt super glue any wires i super glues the 1k on the side and the steel pin on the 1k was touching the spot where to solder, anyway i re did it and solder it and still the same problem, and im getting freezes all the time in crysis that i have to reboot,

What power supplys does everyone have, cuz i had to when i did the volt mod i could smell my powersupply burning, so i had an extra one around but i didnt like it cuz some reason my wiress mouse doesnt work as good, could be alack of power in my computer? +3.3v@36watt,+55 36A, two rails at 19A and the other 20A and the max is 585, i have 2 harddrives, 1 dvd, 5 fans, 2x2gb@1052@2.3v and e4300@3550@1.535,680 xfx lite board, and the bfg2 830/2000/1080, could it be freezein cuz the lack of power?

[cole2109]

http://shrani.si/t/2X/A4/35lV7yAi/900core-2160-shaders.jpg (http://shrani.si/?2X/A4/35lV7yAi/900core-2160-shaders.jpg)

Idle 1.33V
Load 1.38V

:up:

[T_M]

It's the same principle as for G80GTS:

This pic is of a GT so there's no SMD for 3rd phase...
http://img228.imageshack.us/img228/603/g92ocpyz8.jpg

Thanks vortan and largon.
Do the addition resitors help even without performing the OCP mod?

[gOtVoltage]

Oh really? Funny like what? Artifacting or something?

Just benched 06 at 900/2200/2300, 1.41V idle so around 1.46V load:

http://img.photobucket.com/albums/v620/Z3us/0615935.jpg

If only i had a quadcore. :)

Funny as in with the 256bus and memory Qimonda will act up and the shader will start erroring with Squares....Not pixels....After you run it enuff and get a pixeling ( this means the memory cant hadle the OC...) Lines on the screen and its usually all over ,It will slowly get worse....2300mhz on water is good but the Qimonda cant handle it for to long even when the card dosnt go above 41c @ 2250//2300mhz..Im still testing this and after a while i can get the memory to fail @ 2300mhz but not 2250mhz. The Max 2250mhz is common for this series of Qimonda ..Dosnt matter how high a GPU//Shader speed this seems to be the Max Safe limit before memory burn out:yepp: ...

Ive seen that Samsung is finaly working its way onto GTS512's...I may buy afew with some friends this weekend and see if this is true...If it is ,,it will be by far a better stable Oc using Samsung memory @1.0ns@2300..It will be far more stable than the qimonda...people with some new Gt's are already doing 2300mhz with the new samsung...Some have posted at [H] and here too in the GT section..

[T_M]

Keep us updated mate cos ill jump out and grab some straight away

[jason4207]

Exotic cooling is all mine runs



Can you draw what you mean on a photo please.

When does the OCP kick in on this card? How much voltage do you pump in w/ exotic cooling?

Others have already provided the pics. ;)

[T_M]

My card, running on LN2 at -80C to -100C, will top out at 1.505V idle (approx 1.6V max load) with clocks @ approx 1130.
This votlage was the same max for DICE, and is very clearly triggering OCP/OVP

[jason4207]

My card, running on LN2 at -80C to -100C, will top out at 1.505V idle (approx 1.6V max load) with clocks @ approx 1130.
This votlage was the same max for DICE, and is very clearly triggering OCP/OVP

1130...on the core?

So you're not sure exactly when OCP kicks in, but it's definitely getting in your way?

For us guys on air or water cooling looking for 24/7 clocks I think the OCP is irrelevant, but for extreme cooling it looks like the mod definitely has it's place. :up:

[T_M]

OCP definately kicks in for me on GT2 every time when above 1.505v set idle (~1.6V load)

http://img.photobucket.com/albums/v310/T_M/22586.jpg

[dinos22]

shame you don't have a 5.6 quad

[aussie-revhead]

1130...on the core?

So you're not sure exactly when OCP kicks in, but it's definitely getting in your way?

For us guys on air or water cooling looking for 24/7 clocks I think the OCP is irrelevant, but for extreme cooling it looks like the mod definitely has it's place. :up:


T_M do you think OCP would kick in at those volts regardless of cooling method ? If so then it does matter to us on water and air . Could you describe the symptom you get when OCP kicks in ?

:up:

[Zeus]

Funny as in with the 256bus and memory Qimonda will act up and the shader will start erroring with Squares....Not pixels....After you run it enuff and get a pixeling ( this means the memory cant hadle the OC...) Lines on the screen and its usually all over ,It will slowly get worse....2300mhz on water is good but the Qimonda cant handle it for to long even when the card dosnt go above 41c @ 2250//2300mhz..Im still testing this and after a while i can get the memory to fail @ 2300mhz but not 2250mhz. The Max 2250mhz is common for this series of Qimonda ..Dosnt matter how high a GPU//Shader speed this seems to be the Max Safe limit before memory burn out:yepp: ...

Ive seen that Samsung is finaly working its way onto GTS512's...I may buy afew with some friends this weekend and see if this is true...If it is ,,it will be by far a better stable Oc using Samsung memory @1.0ns@2300..It will be far more stable than the qimonda...people with some new Gt's are already doing 2300mhz with the new samsung...Some have posted at [H] and here too in the GT section..

Thanks bro, i'll keep that in mind, even though i'm only running that high while benching.


OCP definately kicks in for me on GT2 every time when above 1.505v set idle (~1.6V load)


Lovely score and overclock. :up:

How do you notice ocp kicking in?
Driver restart or screen freezing?

[T_M]

Black screen restart

[dinos22]

Black screen restart

can you tell us what different restarts there are and what symptoms normally occur also including what happens to GPU volts during different crashes

[aussie-revhead]

I wasnt game to ask a complicated question like that , so Im glad you did .

:D

[cole2109]

http://shrani.si/t//rw/2f2obnDW/940.jpg (http://shrani.si/?/rw/2f2obnDW/940.jpg)

[aussie-revhead]

GPU 1.3V (load) game stable for now...
Accelero + 120mm


Those are great clocks , is that still with the excellero ? What are your temps like ?

:up:

[cole2109]

No...
Accelero@850Mhz temp. 55c on Crysis
Water@900MHz 1.33V (Idle) temp. 43c on Cryisis, 940MHz 1.45V (Idle) temp. 44c AtiTool

[Zeus]

No...
Accelero@850Mhz temp. 55c on Crysis
Water@900MHz 1.33V (Idle) temp. 43c on Cryisis, 940MHz 1.45V (Idle) temp. 44c AtiTool


Excellent clocks there, 940 at 1.45V idle is nice, what were your shader clocks?

[yiancar]

hello , can i use any other value for the gt card exept 1.2k? tnx yiangos:)

[jason4207]

hello , can i use any other value for the gt card exept 1.2k? tnx yiangos:)

I was told to use 3.8-4.2k. I think lower R values will raise OCP detection even higher, and might be necessary if you plan on going DI/LN2, but for water/air I think 3.8-4.2k is probably best.

Hopefully, someone w/ more 1st hand knowledge can chime in, and correct me if I'm wrong.




T_M - WOW! :eek: Nice clocks man! :clap: I knew going cold would yield better results, but that is more than I was expecting. Too bad the vRAM doesn't like to play like that!

[yiancar]

electrician gurus help the noooobs :D

[ViperJohn]

hello , can i use any other value for the gt card exept 1.2k? tnx yiangos:)

4.2k SMR's will shelve the Isen signal down about 30% lower that stock. A 1.2K will shelve it down 50%. What counts is all SMR's added (2 for a GT and 3 for a G92 GTS) are the same value.

Viper

[yiancar]

4.2k SMR's will shelve the Isen signal down about 30% lower that stock. A 1.2K will shelve it down 50%. What counts is all SMR's added (2 for a GT and 3 for a G92 GTS) are the same value.

Viper

so i put a 1.1k or smthing like that it will go less?

[ViperJohn]

so i put a 1.1k or something like that it will go less?

Use 2 (GT) or 3 (GTS) SMR's of the same value between 1.2K and 4.3 K. On a G92 GTS
you shouldn't to need OCP mod at all unless you are just going nuts with the Vcore.

Viper

[T_M]

Thanks for your contributions John

[jason4207]

so i put a 1.1k or smthing like that it will go less?

For the GT I think you'll be better w/ something closer to 4k. You might give yourself more headroom w/ 1.1, but it's probably better to keep that OCP protection a little tighter...unless you're breaking out the LN2/DI, of course!

[yiancar]

last question m8s :P in how meny volts the gpu + ram dies? i am planing Dry ice so i have to know ... thnx yiangos :)

[Zeus]

last question m8s :P in how meny volts the gpu + ram dies? i am planing Dry ice so i have to know ... thnx yiangos :)


Go easy on the ram, increasing voltage doesn't even help in most cases.

As far as the gpu goes, if you go to high on Vgpu, it will blackscreen.

You went too high. :D

[Lastviking]

I got my Gigabyte 8800 gts card this week, i think i will be hapy with it abit low gpu mhz but the shaders and mems are good :)

Stock fan 100% and no vmodd yet.

3DMark2006:

http://service.futuremark.com/compare?3dm06=4724788

Core:783,Shader:2052,Mem:2214

To bad my cpu holding back the scores abit..

What mem does the GTS cards have?

[Zeus]

Backing down on shaders might help you to get higher core clocks.

GTS uses 1.0ns Qimonda but perhaps the latest have Samsung memory, i know the latest 8800GT have.

Are you gonna Vmod yours?

[Lastviking]

Backing down on shaders might help you to get higher core clocks.

GTS uses 1.0ns Qimonda but perhaps the latest have Samsung memory, i know the latest 8800GT have.

Are you gonna Vmod yours?

Lower shaders clocks didnt help my core clocks. Are it anyway to see what memory are used on the card without removing the fan?, i´m thinking about doing the vmodd....its looks abit tricky becuse the stuff is so small and i dont think my gain of the mod will not be so much, becuse my cpu is holding my stuff back.

The memory mod looks much easyer but i dont think it will be any gain in clocks if its not are samsung chips.

[cole2109]

How much is max voltage for 24/7?
900/2130/1090 1.36V idle is this to much? Max temp. on water is 46C.

[ViperJohn]

How much is max voltage for 24/7?
900/2130/1090 1.36V idle is this to much? Max temp. on water is 46C.

1.36 if fine for 24/7 on WC. 46C loaded is a few C high but it all depends on your
water loop components and what all they are cooling too.

Viper

[ryba1]

I got my Gigabyte 8800 gts card this week, i think i will be hapy with it abit low gpu mhz but the shaders and mems are good :)

Stock fan 100% and no vmodd yet.

3DMark2006:



I`ve got also a Gigabyte 8800GTS..:-( and max.Core is 856...
vMod
http://images27.fotosik.pl/143/f41af5c579ad943a.jpg (http://www.fotosik.pl)

vGPU:1,4v idle, vMem:2,03v..and max.Core:856/Shaders:2079/Memory:1104V

3Dmark06..17,484
http://images23.fotosik.pl/143/1908da88a5cf4c09m.jpg (http://www.fotosik.pl/showFullSize.php?id=1908da88a5cf4c09)

temp.38`C idle, max.47`C load on wc
http://images29.fotosik.pl/144/b190c9aeab82eaee.jpg (http://www.fotosik.pl)

[Lastviking]

I`ve got also a Gigabyte 8800GTS..:-( and max.Core is 856...
vMod


vGPU:1,4v idle, vMem:2,03v..and max.Core:856/Shaders:2079/Memory:1104V

3Dmark06..17,484
Temp.38`C idle, max.47`C load on wc

Did you look what memory the card did use?

Nice image :)

[VoRtAn_MaDgE]

Gigabyte 8800gts g92 bring qimonda mem's

[ryba1]

Gigabyte 8800gts g92 bring qimonda mem's

Yes, the same .. Qimonda HYB18H512321BF-10 (HVV09010)

[MuffinFlavored]

http://images27.fotosik.pl/143/f41af5c579ad943a.jpg


How do you take such good close up pictures? If I turn "flash off" on my camera, I have to be incredibly still (which I never am), or else the photo results in a big blur.

[Lastviking]

I`ve got also a Gigabyte 8800GTS..:-( and max.Core is 856...
vMod


vGPU:1,4v idle, vMem:2,03v..and max.Core:856/Shaders:2079/Memory:1104V

3Dmark06..17,484


temp.38`C idle, max.47`C load on wc

Looks like my Gigabyte card are beter than yours ;)

With vMod:

3DMark 2006

846,2160,1107(2214) (Real clocks)
vgpu:1.317(idle) Max load temp in 3Dmark2006 ~64c Stock fan @ 100%

[ryba1]

How do you take such good close up pictures? If I turn "flash off" on my camera, I have to be incredibly still (which I never am), or else the photo results in a big blur.

..a regular lamp and tripod, I always make use of time switch on camera for these pictures...


Looks like my Gigabyte card are beter than yours ;)

With vMod:

3DMark 2006

846,2160,1107(2214) (Real clocks)
vgpu:1.317(idle) Max load temp in 3Dmark2006 ~64c Stock fan @ 100%

that is rather well :)
a new max. 875/1100(2200)/2125 vgpu:1,45(idle) max load temp.47c on my old wc
http://www.fotosik.pl/pokaz_obrazek/b190c9aeab82eaee.html
3Dmark06 17669

[K404]

Yans mods suggest removing the 78X resistors to deal with OCP. I'd be happier adding in resistors in parallel. 2K ohm would probs do it to give a higher current ceiling?

Is OVP dealt with separately?

Im prepping for dice (finally) and dont want voltage limits- only silicon and cooling limits :p:

[neouser]

who can I ask for the lower vmem mod? thank you

my bfg can't go higher than 1060x2 (2120) in atitool. Is it normal? 1100 crashes everywhere..

[K404]

welcome to XS :) Very few people are getting 1100 stable for 24/7 or gaming, your clocks are pretty much maxxed.

What voltage is your GDDR getting?

[neouser]

currently I am at stock cooling. The gpu is 780 2k6 stable and 760(756) atitool stable with fan@70%.

The card is stock voltage. mem with tester os 2.05, tryed simple pencil mod and obviously More problems and less frequencies due to the vddq. so now all stock.

will mod card soon and place on it a mcw60 + sinks.

cheers mate:)

[Lastviking]

Update....i think this is max with indoor temp 23c

http://service.futuremark.com/compare?3dm06=4833220

Opty 165 @ 3017,50mhz 2x512Mb bh5 @ 274,32mhz 2-2-2-5 T1
Gigabyte 8800GTS @ 846,2214,1134(2268) 1.375v idle Stock fan @ 100%

The gpu core is very hard to get any higher...and thats with more vgpu, i can run with 1.317v also @ 846mhz but more vgpu dont give any more clocks...so i think the heat is the problem. But the shader did go up abit with more vgpu.

[initialised]

welcome to XS :) Very few people are getting 1100 stable for 24/7 or gaming, your clocks are pretty much maxxed.I wouldn't say I'm 24/7 stable at 1100 but it benches and games OK. Will see how far I can get it once its liquified.

Set up is below:
http://farm3.static.flickr.com/2242/2209303471_539cdd9b2e_o.jpg

[neouser]

who can I ask for the lower vmem mod? thank you

[T_M]

K404, i will not be trying to removal of resistors mod, instead will try the soldering in parallel method because of what VJ has said.