Printable View
I've got my 256mb Pixelview 7900GT at 550:1800, but want to go further and I'm wondering how much temperatures will rise using the 1.55V considering with the Modded Silencer (See sig) how much temps will rise from 45C idle and 55C load, thanks guys. :)
I did the 1.5V (not 1.55V) mod and get very much similar temps for idle but around 64C for load.
Clocked to 690MHz:1800MHz. :D
Got it up to 690MHz(Plus 19MHz delta):1864MHz. :D
results for the 512mb sparkle calibre 7900gt: http://www.xtremesystems.org/forums/...d.php?t=101451
Heh, hardware pimp flytek. :p:
What's peoples opinions on me running my PCI-e frequency up trying to get 700MHz out of the core?
I only have a ATA HDD, so nothing connected to the SATA ports.
Thanks.
In im opinion bumping it up 10-15mhz from the stock 100mhz can stable out an OC but its different for everyone. Give it a shot, just dont go to high, its been known to bork your windows install needed a reformat to fix. Its happened to at least 2 people I know of.
Just a question... got my GT a few days back, is 567/784 (unmodded with NV Silencer 5 rev.3) with the 'autodetect' a bit reasonable?
I know it makes my 7800GT look like puke but just wondering...
It is not "a bit reasonable", it is fantastic (assuming that it is stable and artifact free).Quote:
Originally Posted by Garrett
Garret - i find the 256mb cards almost always detect core at about 550 and memory often at exactly 784.
since you have good cooling on the core you just need some extra safety active cooling on the memory vreg area, a small vgpu mod (ie 1.4vgpu), and you'll be running 650/800 24/7 at least :D
Jeepers....I had 2 BFG 7900GT's that did 580/840...they were RMA'd cos 2 of the Caps near the PCIe edge connector were hanging off. Upgraded to 2 Gainward 7900GT GS 512Mb cards...they only do 760 on the RAM but the cores bang away @ 635!....All Non modded btw :D
Must get & Play! :cool:
*Edited for Typo
OK I'll try modding it with silverglue... if that's removable with acetone... ;)Quote:
Originally Posted by t024484
I'll try on an old and dead 6800GT card
Sounds to me like the new revision PCB's with 1.45 factory set in 3D are now getting into retail supply. No way should you get those clocks on 1.2v. :toast:Quote:
Originally Posted by Kin Hell
Can you do a voltage check while running a 3D test? What cooler is on it, and what temps are you getting?
Right now m8, I've had several Jack Daniels & Cokes, so there aint no way I'm going in there with a Digital MM. The card is Stock Gainward Cooler & its Idle temps are 29 to 31 Degrees whilst popping ONLY 49 Degrees running this Beautiful Demo/Benchmark:Quote:
Originally Posted by xit
Eye Candy & Then Some! :cool:
I must add, I sent one of the Gainward 7900GT GS cards back cos it only did 575 core. Should have the replacement for my SLI Rig this Friday...along with this mofo:
The Mother of ALL PSU's
The FSP Sparkle 700W PSU I had, got it's Ass RMA'd cos it volted @ 4.87v on the 5v rail. :mad:
*Edit* the RMA'd 7900GT GS was also 10 Degrees hotter than the one in here on it's own & the HOT one was in the lower PCIe slot!
Did you only RMA it because it couldn't clock high?
If that's the case, I highly doubt they'll take it back.
No need....Nibitor lists only 1 Voltage setting....1.2vQuote:
Originally Posted by xit
I can Bench 3DMark @ 650/700, tho ATI tool does artifact slightly. At 630/746, there are no Artifacts & 3DMark runs fine. Idle temps are 29 Deg. 630/746 and 33 Degrees 650/700. Full loads are 46 & 50 Degress respectfully & this is on a Stock GW Cooler.
To reply to Liqu!d's earlier remark on his opinion of getting it swapped out because it doesn't overclock that far, please Consider this:
Maybe I am lucky to be able to deal with a rather nice chap at my Suppliers. But then, you wouldnt know the amount of hassle I've had building 2 Systems at a cost of £3,500, consisting of 5 Faulty BFG cards, 4 Faulty DFI nF4 SLI DR EXpert boards, Firmware issues with Maxtor hard drives on the nF4Chipset & Faulty 700W Sparkle PSU's. You can imagine the Costs having to RMA all that cr4p back to the supplier. Also, the card I sent back was 10 degrees hotter than it's SLI Brother sat here now. Considering these GW 7900GT GS cards are Hand-picked cores to run with the faster than Normal Ram, I'd say that HOT card was a dodgy one anyway!
And you dont think I will get it swapped out! :slap: ;)
The replacement Card will be here Mon/Tue Next week along with a Spanking new 1KW PC Power & Cooling Quad SLI PSU. Right now, I'm considering myself Lucky the Dear lady in my life isn't filing for Divorce & that both my knackers are still swinging in their original Places! :D
OHHH freaking YES!!Quote:
Originally Posted by Garrett
Acetone removes the silverglue juuuuuust perfect :thumbsup:
Now I'll mod my 7900GT card and let you guys know in a while just how it goes
-edit- after masking the area that didn't need the glue I managed to 'drown' the connecting points in glue LOL
But I just wiped it away with my lovely little bottle of acetone, and managed to successfully apply the 1.4v mod :)
664 core says autodetect, not that bad right? ;) (still 784 mem but there's no cooling on there whatsoever, that'll change later)
Well done mate, ramsinks or anything won't you net much more on the RAM.
Coolbits auto detects ~570 for my core, just like before the 1.4V mod.
Manually i can go all the way up till ~705MHz, not bad eh?
I think 664 on core with the autodetect is good, I got 584 right now, also with the 1,4Vmod. I can play and do a lot of things on 660MHz though. The best way to test this is with 3DMark I think.
I got 703 auto detect today on the core =)
Mem detected at 874 (goes higher manually though)
Ive had the 1.55v mod done for quite awhile now and last time I auto detected was around the time that I did it and it wouldnt auto higher then what it did when it was stock. But I tried today and it finally auto detected like it should with the mod.
Way to go m8y! :banana: :banana: :banana:Quote:
Originally Posted by smut
From what I understand the 1.45 is a hardware mod by the manufacturers, and may not be shown in the Bios. Pictures of newer replacement cards on this and other forums show extra components on the back of the card near the volt mod area when compared to original stock cards.Quote:
Originally Posted by Kin Hell
to see what i am talking about, have a look at http://www.hardforum.com/showpost.ph...&postcount=101
I am still interested if you can test the GPU voltage with a mulitmeter when it is under 3D load....thanks. :)
Quote:
Originally Posted by Kin Hell
thanks, its not like I just got those speeds though. Been running at 700/1980ish for awhile now. Just had coolbits auto detect very high speeds for the first time. When I did the 1.4 months ago everyone said just to auto detect and youll know if it worked. So did the 1.4 and kept auto detecting and it would stay the same as stock speeds so I kept redoing the mod thinking it didnt hold. FInally I used atitool to OC and indeed the 1.4 mod was working so I must have redid the mod over and over again thinking it didnt work. So anyway I said this in a post few weeks ago to not rely just on auto detect as it may not detect higher even if the mod worked. I did the 1.55 mod + mem 2.22 mod and got up to 700+/1984 up from 650/1600 with the 1.4mod and no memory mod. Lsat night my PC went down though, I think something on my motherboard fcked up, my monitor isnt getting a picture. The self test says its working though. when I turn on the PC, the monitor blinks for a minute and goes back to the screen it was on before which is a blue/red/green box saying its working correctly and to check the monitor cable or computer power. The graphics card is still geting power as tested with my multimeter and teh PSU still powers up all the hard drives and optical drives and stuff so that leaves the motherboard. Right before it stopped working I plugging the 24 pin conector on the mobo and forgot to turn the back switch off on the PSU and I think it blew something when I plugged in the 24 pin connector, I remember the monitor blinking when plugging in the 24 pin. It was a stupid mistake but I hooked up the floppy drive cable wrong because I needed to get the RAID drivers because i was reinstalling XP so to unplug and replug the floppy cable in the right way I have to take the 24pin connector out. Thats how the ASUS sli premium board is layed out, pisses me off. So I must have forgotten to hit the switch on the pSU. I always make sure to unplug the power to psu or turn the switch off but I guess this time I was in a rush and forgot to! So I think something with the 24pin plug on the motherboard is shorted out or something. I wish I could find out for sure but have no spare motherboard or anything around. THats the only explanation I can come up with because right before that it was working fine and right after I think something blew on the 24 pin on the PSU or the 24 pin on the motherboard. Right after that is when the monitor wouldnt get a picture anymore besides the self test it always does. Like I said the monitor blinks for a split second like its about to show the BIOS screen and stuff or whatever then it goes back to the self test. I think its not getting power or someting I dont know. But the core/mem on the graphics card are getting power, I checked with my DMM. SO i dont freakin know!?!? so it looks like I have to RMA the board which is going to be a pain. I got a watercooling setup cooling the CPU and GPU and that wont be fun taking it all apart and draining it. I just hope its the motherboard thats the problem, this sucks though. Im at the library right now as my pc is down obviously. Any suggestions, do you think its the motherboard?
Heck smut....I seen this happen on my old 9800XT. I had a black screen but the puter still booted. Do you know if yours is booting or not? Watch your hard drive light to take this out of the equation at least. If it does boot, you can still shut down "blind" by pressing the windows key once, pressing the Up cursor key once & then hit enter or space. Press TAB once & hit Enter or Space & your Rig will shut down.Quote:
Originally Posted by smut
Shame you dont have another board there to try, but with all the things you have done to your Rig, it would be hard to even start to guess where the trouble lies. It could be summat on the graphics card considering the mods & OC you have done, though my 9800XT was stock. I was lucky to be able to RMA 1 month after its Warranty Expiry & Hercules France repaired & returned my card FOC.
Best of luck hammering this one out, my gut churns for ya! :(
Yeah the system boots, can hear the hard drives booting and the lights show it as well. Its like everything is working except not receiving the picture on the monitor. This gives me a chance to RMA my g.skill ddr500 2GB of ram though that never ran stable at DDR500. So im going to RMA the ram and the motherboard and hopefully ill be back to normal. I just hope ASUS doesnt take forever with RMAs but for some reason I remember some1 saying that they do. And like I said it was working fine until I plugged in some of the 24pin connector and it seemed like something got power that shouldnt have at that point because the PSU switch was on meaning althought it wasnt on it had power. After that, no picture. Teh psu boots the board up, hd's, optical drives and the graphics card has 1.55v on the core and the memory has power as well so its gotta be working. I checked the monitor cable, switched out the dvi/vga adaptor to no avail. Ive done everything I can think of, its going to suck taking everything apart though. The water cooling draining, the cpu block removal, gpu block removal, taking out everything, removing the mobo. Going to take quite a few hours. At least I can get my RAM replaced, I havent sent it away yet because I didnt want to be without my computer, its how I make money right now selling stuff on forums and ebay but I have no choice now. newegg is much better at RMA speed then asus so I should have the RAM back way before the motherboard. If its still not getting picture when the mobo gets back then will try to figure out which I should RMA first, the PSU or graphics card. I doubt the monitor is the problem, it says its working on the self test. It blinks for a second totally black when i turn on the PC then switches back to the blue/red/green box. Its like it got power for a second then stopped. Ahh, this blows. just hope the motherboard fixes it, the onboard audio doesnt work on the mobo either so thats getting fixed too.Quote:
Originally Posted by Kin Hell
Yeah xit, sorry for not getting back sooner, but I have had my hands full getting things together at my end. Unfortunately, my Gainward 7900 GT GS cards are not performing correctly in SLI. I am still waitng to hear from my supplier RE the issues I have & consequently will not be making any atempt to mod or poke around the cards. My PSU is now out of the equation without any doubt. I have just fitted a PC P&C 1KW Quad SLI PSU, but my rig just will not run these cards in SLI without some crazy anomalies. I promise to write up a full report as soon as I know where I am going, but RMA time looms ahead with these Gainwards, refusing to SLI at default clock speeds. I am not happy having a £540 investment in 2 cards to see my rig lock up, crash with "nf4_disp" errors or BSOD's with "nF4_disp" Error codes whilst trying to game. On top of that, I get corrupted 2-D windows while running a 3-D stuff in a window. So many installs, so many different System hardware drivers verified & tried...my patience is wearing thin. Go have a read of these links at some of the graphic issues I have been seeing:Quote:
Originally Posted by xit
http://forums.nvidia.com/index.php?a...r&f=25&t=14044 :eek:
http://enthusiast.hardocp.com/articl...50aHVzaWFzdA==
Sorry to here of your woes Kin, I reckon its not just the Gainward but all the 512MB cards that use the non reference PCB, just my opinion :)
You may well be right alpha, it really is a GS card with the Bridge Chip still in place. I heard GW had to do this originally to get the card to work with with 512Mb. One way or another, I feel somewhat "Fudged" as both cards perform without issues as single cards. The Ram of these cards is really cr4p for OC'ing tho & the card become "unpredictably unstable". SLI 2 of these cards & they do the same!Quote:
Originally Posted by alpha0ne
Edited Double Post
It's heat issues...that easy...1 card idles at 33, the other at 41. With the Hot card in the top slot (its the fastest of the 2), the heat from the bottom card just makes the top card too damned hot & thats why the the graphics are locking & BSOD's. This is happening at the Cards Default speeds btw! :eek:
To proove it, I swapped em round so the slower but Cooler card is in the top slot. This then allows the SLI'd cards to clock to 625/740, but the side panel has to be off. Put the side panel on, the cards heat has nowhere to go but upto the PWM & Northbridge Area pushing case Temps from 30 DegC on CPU, 40 on PWM & 43 on chipset to 45, 49 & 58...ffs! 2 x 256Mb BFG's never did this & those cards vented sideways through it's slim cooler. These GW's have a Bigger Cooler & Ram Sinks but it's blowing all the Heat from the cooler over the Power area of the card for cooling, which really is 4rse about face on Industry Standard case Air Flows. What are the graphics card manufacturers thinking eh!
My Rig is fine with one card & the side panel on, tho the hotter of the 2 cards on it's own will still start to throttle back after sustained gaming periods.
No wonder so many people are RMA-ing 4 cards to end up with the same issues & sustained use with these kind of conditions will just start blowing components on the cards. This is probably why my m8's GTX blew itself to bits. He had candy pink splodges, wierd textured artifacts & seriously reduced FPS numbers in 3DMark after his nuked itself.
You can't get much better Air flow as I have on my Lian Li PC75 Server Case, but I still can't believe the card manufacturers are cooling these cards 4rse about face to conventional case Air Flows, which really are an Industry Standard. Then again, it's a bit like Lian Li & their Upside down cases such as the V1000, 2000 etc. Quite an interesting design in some respects, and so nice to see your Graphics card & PCI devices the right way round, but then the Mobo manufacturers start putting Heat Pipes on the Chipsets & they DONT work upside down! You gotta laugh, it sure beats the hell outta crying but ffs at it all. The designers need their 4rses Spanking seriously hard! :slapass:
Addendum: You can tell when the cards are going to throttle back when it's got too hot, you'll see the running Graphics Pause for a split second & then carry on. If you know what you're looking at, you see HUGE differences in the numbers....Like finishing Aquamark at 54 intead of 103 or seeing a best of 160 odd in 3DMark03's elevator when it should be 380+....Sheesh! :rolleyes:
I made vmod with pencil on my Palit gf7900 gt 512 MB card, and now i would like to measure voltage of the core with multimeter. I read forums but i cant find explanation how to do that. Could someone paste a picture with marked points where i need to connect multimeter to measure volts ?
Hows things looking for ya Smut?Quote:
Originally Posted by smut
Have you been able to draw any Conclusions yet?
I RMA'd the 2 x GW 512Mb Gt's and took an XFX XXX GX2....Core hits as high as 74 DegC, 86 with the Side Panel on! :eek: ....no lock ups tho & no BSOD's. I spect if Quad SLI ever becomes available to DIY Consumers, there will still be heat issues with the Lower SLI Card acting as a Blast furnace for the Upper SLI card! |Gunno have a look at increasing the Gap on the cards. I'll know more for splitting em!
Best of luck with your probs anyways! :)
Hows things looking for ya Smut?Quote:
Originally Posted by smut
Have you been able to draw any Conclusions yet?
I RMA'd the 2 x GW 512Mb Gt's and took an XFX XXX GX2....Core hits as high as 74 DegC, 86 with the Side Panel on! :eek: ....no lock ups tho & no BSOD's.
I do have several 110 Event ID's saying the HW Clock has throttled due to Heat. I spect if Quad SLI ever becomes available to DIY Consumers, there will still be heat issues with the Lower SLI Card acting as a Blast furnace for the Upper SLI card!
Gunno have a look at increasing the Gap on the cards. I'll know more for splitting em!
Best of luck with your probs anyways! :)
After having looked again at the voltmod area, I found a third way to do your voltmodding.
This third alternative is the easiest of them all. You don't have to connect to ground anymore, and the mods can be done independently.
Red gives you +0.05Volt
Yellow gives you +0.1 Volt
Purple gives +0.2Volt
See picture below, on the lower row.
Attachment 48507
If you want just 1.4 Volt, make the purple connection, if you want 1.5 Volt use purple + yellow, etc.
If you have changed the 5 Kohm resistor into 50 Kohm, you start with 1.5 Volt instead of 1.2 volt, but the voltage steps remain the same with 0.05, 0.1 and 0.2 Volt.
updated mod @ 1.7v max
1.6v 24/7 water gpu 720(+40 delta)
1.7v bench gpu 750(+40 delta)
p.s photo from se w900 :)
Heh, mine kept overheating and artifacting at 1.5V, probably because my system is with all silent fans.
Sorry if this is remedial but if I want 1.4v can I chose to connect whatever set of purple pins that I want to or do I need to connect all three purple sets.Quote:
Originally Posted by t024484
The answer to both questions is a clear NO.Quote:
Originally Posted by jmsandrsn
Read again, and you will find your answers.
Only this new third option gives you 1.4Volt by connecting the purple pins.
Is there a way to get 1.45v, or is it only 1.4, 1.5 or 1.55 ?Quote:
Originally Posted by t024484
Thank you again t024484 for your wisdom. :clap:
Quote:
Is there a way to get 1.45v, or is it only 1.4, 1.5 or 1.55 ?
you can combine (the color) this mod, so you can have 0.05v increments...Quote:
Red gives you +0.05Volt
Yellow gives you +0.1 Volt
Purple gives +0.2Volt
That´s itQuote:
Originally Posted by andre X_X
red 1.25
yellow 1.30
red + yellow 1.35
Purple 1.40
purple + red 1.45
purple + yellow 1.50
purple + red + yellow 1.55
its the easiest mod ever.... doesnt void the warranty, an as easy as drawing a line... :) thx to t024484
I have a question. If I were to solder that 50k ohm resistor for a 1.5v base, would it be possible to reverse the process sufficiently if I needed an RMA? I'm not afraid i'd fry my core, Im @ like 39C on load, im not sure my memory wasnt defective, and if it should eventually give out noticably I'd want a new card.
Currently @ 760/975 3d06 stable @ 1.54v/2.17v, 770/1030+ in AM3 (meaning, was atable here, havent gone higher)
If you solder a 50Kohm resistor to replace the existing 5Kohm resistor, there is no way back for an RMA.Quote:
Originally Posted by CompGeek
Keep your Voltage Regulator Mosfet cooled with a heatsink, and your memory will last forever.
And as as a last advice: do not use Coolbits. The best thing to do ist to set the 3D frequencies in your BIOS with Nibitor.
Owe this all to Quijonsith @ EOC
http://img135.imageshack.us/img135/36/image0257gs.jpg
:toast:
Why don't you use a SMD resistor like the one you took off. It is so much more stable, and less vulnerable.Quote:
Originally Posted by CompGeek
because i didnt find any at radioshack
you can find it at farnell.com
anyway, if u decide to do the 50kohm mod, i suggest you to do this mod also. its a gread mod....
good advice, but the problem is we dont need the speed for daily using.... its only for bench..Quote:
The best thing to do ist to set the 3D frequencies in your BIOS with Nibitor.
im getting 2 xfx 7900gts in a few weeks (the black ones to match my case), i plan on ocing it on stock volts, then jumping to 1.4, and eventually 1.5v they will both be watercooled. I was wondering if anyone could post a pic showing witch mosfet chips need to be heatsinked so they dont overheat? thanks.
Quote:
Originally Posted by illuminati-hwt
Where exactly does one perform this mod? Is there a photo posted somewhere someone can refer me to which shows this pencil mod? In the photo t024484 posted (see below) which shows the place circled in lite blue for a 47Kohm resistor mod; is this the same place to do that pencil mod? Thanks in advance expecially to t024484 for all the work he's put into this!
http://www.xtremesystems.org/forums/...7&d=1148894066
Well, after looking through this thread, though not reading all the posts, I noticed that the v-mod using the chip is missing. The chip I'm referring to is the one in this picture.
I know for a fact it used to work on my previous eVGA card, but that was when the 7900 GTs had just come out. Now that I have a new eVGA 7900GT that mod no longer seems to work. I used a pencil both times so don't tell me to use a conductive pen. I'm just wondering if they did something to where this mod no longer works, or if it's changed in some way.
Thanks
EDIT: nvm, the mod does work. For some reason the stupid pins weren't connected even with a crap load of graphite on it. Anyways, I got it working. And here is a picture of the pencil mod:
http://img145.imageshack.us/img145/6716/vgpu26zs.jpg
@Alexstarfire,
The pins aren't stupid! :slapass: It's best to use conductive ink or paint (silver based is probably the best) when soldering is to difficult...
Now will someone please respond to my previous post? Thanks!
@ Zeeky... on resistor K... doh
lol, unless you mean something else.
posted in wrong thread, sorry guys!
Do you mean cover the K resistor with pencil lead (better then to use conductive ink/paint) from end to end? And if the answer to my question is yes then why perform a 47K ohm resistor mod when a simple pencil mod will do essentially the same thing?Quote:
Originally Posted by CompGeek
Edit:
I PM'd Haltech on the matter and this was his response:
"I penciled right over it. If i were to do it again, i would either solder a VR or not even touch it. My memory clock went down when i did it. I had to blast the area with chemtronics circuit cleaner to get it off."
Well I now know how to perform this pencil mod but evidently it doesn't work? Perhaps the 47K ohm resistor mod is the only solution if one has the memory overclocking problem t024484 made mention of?? Thanks in advance!
You are mixing two different issues.Quote:
Originally Posted by Z33ky
1) Resistor K is for increasing the oscillator frequency. You can do a pencil mod on top of resistor K, to bring down the resistance to 45Kohm.
With conductive ink you will bring the resistance down to 0 Ohm, which is a no no.
BUT, pencil moddings are not that stable, and lots of debris are produced which could upset some other parts. Much better is to solder a 100Kohm resistor on top of resistor K.
2) The 47 Kohm resistor is to lowering Vref, to get more room for noise on the VGPU before the OVP shuts down. This is all described on the bottom of page 20.
Not to difficult to find with your fingerQuote:
Originally Posted by adamryer
Attachment 48671
Anyone have any idea why, since I did this mod, my cards clock higher, but my scores are way down?
Im getting about 18k in 05 @ the same speeds with which I previously got 21k.
How far is it safe to take Resistor K down to? Decreasing it to 38Kohm has GREATLY increased my stability @ 1.7v, will going down to 30Kohm be an issue? Will i risk killing something?
I think you can go without risk to 30 Kohm. As long as the mosfets are not getting boiling hot, everything is O.K.Quote:
Originally Posted by CompGeek
I never tried 1.7 Volt, so it is good to know that K has to be decreased even further.
If you make your system more stable with various mods, you come at a point where everything seems to be stable yet, but scores are going down.Quote:
Originally Posted by CompGeek
Just a little decrease in Mem or GPU frequency is enough to get your scores like they should be.
Nah, that didnt help, it was the above 38Kohm that was NEEDED, and I think 30 will help even more :)Quote:
Originally Posted by t024484
Hey I havent had time to read thru this whole section but I got a quick question
I have an evga 7900GT CO SC stock at 580/780 well any way I want to do the 1.55 mod but I heard you need watercooling for that.
I have a vf900 installed and my idle temp is 35 C.
If 1.55 is too much, is 1.5 resonable?
I already know what to do , its just I googled evreything looking for a heat dissapation chart on the voltages.
thanks :D
@AnthraX
The extra cooling won't prevent frying your card! Check out the following posts:
http://www.xtremesystems.org/forums/...d.php?t=102275
well I just did the 1.4 mod...
temps idle at 39 ish 40.
put the core right up to 685 no problems...
But I think I may have done the 1.5 by accedint..see I messed up so many times, that I think when I wiped it off with acetone, I couldnt get all of it and yeah...
one I get my digital multimeter Il check it out.
Also can someone please show me a picture of where to put the heatsinks on the IC's??
I would appreciate it.
also is it normal to hear a high picthed whining sound when I run intensive 3D apps?
:)
1) Heatsink: See earlier this pageQuote:
Originally Posted by Lord AnthraX
2) 1.55 Volt: Is still an open ended discussion. In general, the more volts, the more likely that lifetime is shorted. But unmodified boards with 1.2 Volt are dying, and others are seemingly doing O.K. with 1.7 Volt. With the VF900, you do not need watercooling for 1.55Volt.
Do not use Coolbits. Switching from 2D to 3D with Coolbits aggrevates the problems with the GPU.
3) Whining: Search this forum and you will find how to cure this.
t024484,
In the above post you mentioned to not use "Coolbits" because the switching between both 2d and 3d can aggravate the problems with GPU. Now I have done minimal searching on this topic both here at extremesystems and also googled it without much specific information. I am curious because I have heard it mentioned in the past.
With that said, is the problem caused by "Coolbits" itself or is it just the actual switching of 2d and 3d clockspeeds that aggrevates the issues? Despite where the problem stems from is there a safer way to change clockspeeds between 2d and 3d modes, maybe by using another program like RivaTuner, ATI Tool, etc...? Thanks for you help and I'm sorry if indeed this question has already been beaten to death, lol!
LordAnthraX check out post number 567 as it will answer your last question.
Ok yeah I got it.
I put some Heatsinks on the IC's and it stoped
Im glad I only heard the noise once I dont want to damage them.
So does 1.5 seam reasonable? in terms of heat output?
In another thread, I have reported my experiences when I thought my board was dying, because of flashing screens, freezes and strange artefacts independent of frequency, and the fact that it was no longer possible to run Deep Freeze in 3DMark06.Quote:
Originally Posted by Roger_D25
Then Flytek and Humeboy both advised me to deinstall Coolbits and try again.
Because everything functioned perfectly before the problems started with Coolbits installed, I had little believe in deinstalling Coolbits.
But much to my surprise, after having removed Coolbits, everything functioned troublefree. I have then tried ATItool and RivaTuner instead and both where O.K. but at the end I have simply changed the 3D frequencies in BIOS with Nibitor, and my board is still completely troublefree, the question is of course " for how long"??
I can see a big difference when starting a 3D program like 3DMark06 with and without Coolbits.
With Coolbits I see a large black area on my screen flashing on and of shortly before 3DMark06 starts loading.
Without Coolbits the transition from 2D to 3D goes smoothly and without this large black area.
Maybe this is the moment where something in the GPU is getting damaged/disturbed when this has happened several times.
It seems that Coolbits is using a different switching mechanism or switching order that pushes the GPU over an edge that should not be there in the first place.
Quote:
Originally Posted by t024484
Hmm... i dont notice that kind of anomality on my card, i have a stock clocked Inno3d 7900gt 256mb, did it completely eliminate your problem after removing coolbits? ( im also using coolbits to overclock )
Also does putting cooling on those hot IC's really help ? coz i placed a total of 3 zalman vf ramsinks ( same with ones that came with the vf900 ) on those spots. and have an active fan blowing on them... first i tried w/o an active fan the heatsinks pretty much got extremely hot.
t024484,
I just barely got your above message, thanks for the reply! I'm glad to say I've had the same results as you did with my SS card. Just like you I would get this "Screen" with either lots of horizontal/verticle multi-colors lines or just a total black blinking screen when first starting up games, benchmarks, or anything 3D because of the initial 2d to 3d switch.
This only happened a few times because I had done a driver version switch late and night and didn't have time to reinstall RT afterwards (I just loaded up coolbits). When I install a new driver I usually uninstall/reinstall RT as well despite the fact I might not have to? Anyway after installing coolbits I started getting the "Screen" before starting games/benchmarks/or any 3d applications. Fortunelty I knew it might be because of coolbits (had heard of this happening over at guru3d) so I uninstalled driver again, reinstalled along with RT and the "Screen" went away, problem solved! Thanks for your experience.
that's been happening to me when loading eq2. I'll try the latest drivers fresh install and see if it fixes it.
Hello, I recently did the 1.55 volt mod on my 7900GT's and I'm loving the 700Mhz core action, to think that even with 1.9v on the 7800's those speeds were still unthinkable, but on to the point. I have a pair of the EVGA 7900GT KO's with the Samsung memory, my only beef is that the memory just wont overclock! I've read that with higher core clock speeds the memory will go up, but I only gained 10Mhz from stock Vcore to 1.55. I just cannot get past 1600MHz. Is this normal? From what I've seen from other users, 1700-1800Mhz is average, I was just wondering what you guys thought about this. I would volt mod the memory, but I've read of all the horror storys that happend days later. If I use some high quality pots is it safe? Just wondering what to do at this point, I'm fairly certain my cards are severely bandwidth starved. Thanks for any input, cheers!
Update: I just ran 3DMark06 and I get the artifacts on Deep Freeze... I guess It's time for an RMA.
When you drive the GPU frequency to its maximum, there is no more room left for increasing the speed of your memory.Quote:
Originally Posted by TehVince
Bring back your GPU to 650 Mhz, and then try to increase the speed of your mem. If you have found a maximum for your Mem at gpu at 650 Mhz, then make small steps with both in either direction to find the optimum.
Be prepared that too high a mem frequency will lower your scores with 3Dmark06 instead of increasing it.
Also cool the Vmem Mosfet properly, and do not use Coolbits.
A fixed resistor is more stable than a Pot to increase Vmem.
The fact that your card artÃ:banana:facts in 3DMark06 can just as well be caused by a GPU frequency that is too high.
Could I add that too much "Heat" could also cause Artifacting. I had 2 x Gainward 512Mb Cards that artifacted at default clock speeds in SLI. One card was cooking the other. Blow lots of Air over the cards & the artifacting cleared up. Running the cards seperately gave no artifacts at all & the PSU being a PC P&C 1Kw Quad SLI affair just wasn't even in the equation.Quote:
Originally Posted by t024484
3DMark 06 is a much heavier load on your System components, compared to earlier versions of the Bench.
Thanks guys, I'll try turning down the clocks and running again. They just looked like memory artifacts to me. My temps are fine, and I'll be sure to get a sink on there A.S.A.P. I've got two 120mm's blowing over the cards, but I'll do the finger test. Does anyone have an opinion on my PSU? It's an Antec 550W True Power II SLi. Do you guys think this is enough power with this volt mod? I haven't had any random shut downs and my rails are stable. I've been looking at a 700W GameXstream OCZ PSU, any thoughts? Thanks in advance.
if you had GTs in SLI the Antec is a little low, I had one and moved to the GameXStream 700, they needed it.
If you have just one GT the antec is enough imo.
Spot on! :)Quote:
Originally Posted by CompGeek
aha, wait, he said "the cards", meaning he has two.
The two GTs will thank you for a GameXStream 700
I just ran the bench mark again at 650/1520(stock) with zero artifacts, tried again at 700/1520 and once again, no artifacts. I'm just not satisfied with my memory clock, I can't even get 1600 with any core clock. But I'm buying the PSU right now. Thanks for your input.
it was the same here, id get flashing screens etc with the slightest mem OC until i bought the OCZ GXS
Oh wow! Very cool, Now I'm excited, haha! Thanks guys.
Edit: Newegg's out of stock :(
Double Edit: I was just messing around with the meter, Measured my top card at 1.99v so that would explain the memory clock issues, but when I pulled out the second card and re measured the voltage it was still 1.99. Is my power supply that bad? To be fair I didn't change the SLi jumpers because I'm lazy, and I know alot of the juice comes from the PCI-E slot, but with no second card it shouldn't make a difference right? Or did EVGA cut the volts on their cheapest card. I also noticed my core voltages were a bit off too. 1.548 and with the 1.4 mod, 1.398. Any ideas?
I think the PSU is your problem in all this, lower voltes, memory clock etc., get your OCZ 700W from Tiger direct if newegg's out of stock, they too have good prices...
Yeah.....check the Antec's Amperage on the 12v Side of things. Then look at the OCZ's Amperage. The OCZ should floor it. Another thing to do is check your PSU with a PSU Tester. They only cost about 10 bucks. PC P&C chuked one in with my 1KW Quad SLI PSU....how nice of them! :)Quote:
Originally Posted by TehVince
Another thing worth checking is to make sure you have all the necessary Connectors to the Board. ASUS & DFI have power connectors on the board which MUST be connected with power when Running SLI. The ASUS boards use a Molex Connector whilst the DFI Expert boards use a Floppy Connector. The mobo manuals insist on these Connectons when running SLI.
Best of luck! :)
*edit* Antec are not the best of PSU's
I was in luck, shortly after my edit they were back in stock at Newegg, woot! And you were right Kin Hell, the Antec had two 12v rails both rated for 19A, I have no idea how the six pin PCI-E plugs are spread on this PSU, but the OCZ is 18A accross all FOUR 12v rails, YAY! Not to mention CPU and GPU's are all on they're own rail, very cool. I appreciate all the help guys, I'll post when I get my power supply. Cheers!
Edit: I was just wondering what compononents I should have heatsinks on in the attached picture, I've read that one of these chips doesn't like to have a heatsink on it. I just want these cards to last as long as possible, and pending on the results I get with the new PSU, I want to be ready to do the memory volt mod. Are there any other chips I should have sinked? Thanks in advance guys!
It was posted on page 23, but for some reason the picture is no longer there.Quote:
Originally Posted by TehVince
Attachment 49208
Thanks t024484! I'll make sure that chip is sinked A.S.A.P. I'll post when I get my PSU, Thanks a ton guys!
do any one know if i use a 7900GT KO SC bios on my 7900GT CO will i get the 1..45volt on it too or will it still be set at 1.2
The higher Voltage can only be enabled if the hardware switches (D509,D512,Q517 etc) are there.Quote:
Originally Posted by ruff97
As far as I know, the CO board does not have these Switches onboard.
Bad news guys, I threw in my new and beautiful PSU and nothing's changed! I've lost my LCD multi meter for the moment, so I busted out the Micronta range doubler (22-204A) from the 80's and I'm getting the same volts (I think). It's a little hard to get an exact measurement and I think it might be a bit off, but it looks to be around 1.99 again. Needless to say, my overclock didn't change either. I've plugged all the molex's into the board, everything's hooked up correctly, what am I missing here? I'm going to try and get a hold of one of those EVGA bios with the timing tweaks and see if that helps at all. On a different note, I am capable of doing the memory volt mod, but I've read all over the net that doing this is just asking for trouble. Is this true? I've got the component that t024484 marked sinked, a good solder iron, and a good hand. If I use high quality resistors and don't get greedy with the volts will my card's stay alive for a couple years? I'm a bit confused here and looking for good ideas, thanks a ton.
No supplier is telling us the reason why the 7900 is having problems.Quote:
Originally Posted by TehVince
Although EVGA changed their memory chips from Samsung to Infineon, I do not think that this will be the solution for the majority of the problems.
When I use Coolbits, it looks as if my board is almost dead, although initially everything worked fine with Coolbits. So it is for sure that something is damaged and working worse than initially.
Without Coolbits everything is working as it should, already for almost a month.
Coolbits is seeming to use a function that others do not use in the same way.
With all this information, nobody can convince me that the memory is the cause of this. I have never heard of memory that does not like Coolbits.
So because the Infineon memory seems to use less power, it should help to reduce the heat problem of that single Mosfet, but it does not solve the rest of the problems.
So coming to your question, I think that you can safely increase the voltage on your Memory, as long as you keep the Mosfet cool. How long your board will last depends on the GPU.
Quote:
Originally Posted by t024484
I don't see the picture for this mod, is it gone? Would someone kindly repost it or point me to where it is?
Thanks.
Hey kin! Thanks for the PM while I was gone although I just not received it. It was nice to see that you were curious as to what happened. Well the good news is that I have the motherboard back from RMA and it turned out to be the motherboard was bad like I thought. Everything else is fine, although I did RMA my RAM while the PC was down. I needed to do it anyway but never wanted to have downtime (but in hindsight it only took newegg 4 days to have the ram replaced and back at my doorstep which was lighting fast compared to ASUS) so it was the perfect shot to take care of it while the PC was down anyway. The onboard sound wasnt working right before the black screen issue so it needed to be RMA'd anyway but since I have a dedicated audio card it wasnt a big issue and I wasnt planning on worrying about getting it RMAed until I was ready to upgrade the whole system. But then the black screen issue came up (no signal to the monitor). Im pretty sure something must have gotten fried when that power surge hit the mobo, a buddy said the same thing happened to him before. His was a burnt out cap that wasnt providing a signal to the monitor so I believe mine may have been the same issue. I'll never know exactly I guess, I just know it was something with the motherboard.
So I had to take apart the whole system, luckily I didnt have to drain my whole water setup, I was able to take out the motherboard by just removing the blocks from the CPU and GPU and carefully take it out so that saved alot of time and frustration. I actually waited a couple weeks before I even did anything, didnt really have the time (or so I thought it would take awhile but it didnt because of the water setup as mentioned above) but I eventually RMA'd the board and ASUS took longer then the "10 days" they say an RMA will take. That was something I feared because I remember hearing ASUS sometime can take awhile with RMAs. But anyway they received the board on the 19th, they don't include weekends or holidays as part of the 10 days so technically it should have taken 2 weeks (14 days - 4 weekend days = 10) but the 4th of July holiday rolled around too so that added an extra 2 days turning into 16 days it technically should have taken but it took about 23 when it was all said and done, not including travel time back to me which was an extra 2 days, I got lucky though, fedex dropped it off a day earlier then the expected delivery so that was nice. But ASUS took forever just to replace the board (whenever I called to check the status they said it was in repair status but once the board showed up I found out it was actually replaced so the long delay had me even more pissed that it took that long just to replace a board but oh well). I finally got the board back on the 13th (yesterday). 6 days shy of a full month, 10 days my ass :mad: lol.
Oh well, i'm just glad to be back up and running. I have alot of gaming to catch up on, I got HL2: EP1 and Titan Quest during the downtime so I finally get to play those. I had to use the net at the library to keep up on my forum sales and stuff (I sell at Hard Forum, anandtech and PCPER and will here once I reach the 100 posts required) and I kinda forgot about XS since I wasnt selling here and most of my limited time was spent on the forums I sell at. Most of my time was consumed with updating threads, keeping buyers updated and all of the other stuff that comes along with internet trading heh. But yeah it definetely sucked waiting for that board to get back, I watched a ton of movies during the hiatus and luckily a friend let me borrow his XBOX so I was able to keep busy with some console games as well. I played alot of games though, I finally got to finish Jade Empire which I never had the chance to. I used to have an xbox a year or so ago but it was stolen before I got to enjoy Jade Empire. I bought the game and guide and got to play for a few hours before it got taken. Anyway, its a great game and will be coming to PC soon so if anyone missed out on it because you didnt have an XBOX or whatever I highly recommend it, its an awesome game as everyone already knows. Im just glad to be up and running again, woot!
When I find that damn meter I'm going to get right on it. Thank's for all the help t024484 and everyone else! I wonder if I left it at work, hmm...
P.S. I haven't been using coolbits since I bought these cards, I was actually just waiting for you guys to break the 1.3v limit before I ditched my 7800GT and I stumbled accross the coolbits issues. Once again, thanks a ton everyone! Oh and, this PSU is really loud...
Nice 1 Smut...Asus...rofl...I chap I know waited 5 weeks for a replacement Asus X1900 XTX....The supplier let him have a tested GX2 XFX XXX card as a gesture of goodwill. Good service is Like Owning a Plextor Drive! ;)Quote:
Originally Posted by smut
Well, after about half an hour of searching, the little devil appeared in dirty clothes basket. I'm not sure why or how it got there, but I found it. Anyways, I quickly checked the voltages of my cards and to my surprise, the memory still registered 1.997v, so I checked the core and I was very happy to see the 1.550 that I wanted. I'm going to do some penciling and see if my overclock increases, if it does I'll put on either a fixed resistor or a pot. Any insight to why the memory is undervolted? When I bought these cards they were the cheapest newegg had, could this be EVGA's way to justify the 7900GT SC badge on more expensive cards? Thanks in advance.
Edit: With my pencil I'm running 2.22v on both cards and I still cant overclock my memory. I'm not to happy about this and am debating what to do from here.
Quote:
Originally Posted by t024484
Is the one marked in red the only chip needed to be heatsinked? or would the one directly above and on the rightside also need cooling ?
The one to be cooled is of course the red guy, why else should only one component be indicated.Quote:
Originally Posted by Demo
But, you can attach the same heatsink to the other two also to get a mechanically more stable platform.
The memory is not undervolted, but in fact already somewhat factory overvolted, because the nominal value is 1.8 Volt.Quote:
Originally Posted by TehVince
Nvidia has set the voltage at 2.0 Volt, which is exactly what you measure.
But the maximum allowable voltage is 2.5 Volt.
Going to 2.1 or even to 2.2 Volt on a continuous base does not seem to be too much of a problem.
My board is running on 2.1 Volt for almost two months at 1850 Mhz.
One word of advise, be carefull with pencil modding, it is not a very stable mod, and you are producing quite some debris that can cause trouble.
Another word of advice is to be carefull with overclocking the memory any further once its starts artifacting.
The memory or whatever else does not like this, and you can cause permanent damage if you go beyond the artifacting limit too far.
Use AtiTool, and stop when AtiTool visually starts Artifacting with yellow points or lines. That's all.
The reason you cannot overclock the memory, could be because you have the GPU at its limit. Just lower the GPU frequency and see what happens.
Be aware that there is a ratio of 2:1 in Mem vs GPU frequency, meaning that 2Mhz increase on the Mem has almost the same effect as 1Mhz on the GPU.
So if you have to drop the GPU 20Mhz to achieve a 100 Mhz increase in Mem, you have made quite a step forward.
Can you be a bit more specific. I have posted several picture. Which is the one you are referring to. Give me a link to the page where I can find it.Quote:
Originally Posted by InTheFlow
he means that pictures on page 1 of this thread are no longer seeable, I too can't see them anymore...