Yes they have :slapass:Quote:
Originally Posted by flytek
Printable View
Yes they have :slapass:Quote:
Originally Posted by flytek
card is officially dead ;(
replaced both mosfet's & tried to boot & the same mosfet went up in flames. When I mean flames, it burst into flames & melted my solder joints & fell onto my motherboard :eek: Was burning for about 5seconds with a nice yellow flame..
Nothing else was damaged luckily.
C-BuZz
Sorry to hear though you maybe could have made a fortune selling the video of the fireworks :eek: :D
heh, can always solder another in its place :)Quote:
Originally Posted by alpha0ne
C-BuZz
@graham - did you take that picture? maybe you could get the full codes of the largish elongated chip about an inch below the 0605 date stamp. that'll be the vgpu controller i would think.
the two 8 legged ones at the below that chip will probably be for the vmem. full codes would help there too.
thanks
flytek it is my card I'll try and get some more with a bit more detail
exellent. also it would be very interesting to find out what coolbit's nvidia driver autodetects the core/mem frequency overclock at.
something else of great interest would be stock frequecy clocks and a 3dmark06 run.
lastly the codes of one of the ram chips would be very nice to know. for all we know gainward changed that on these golden samples
Sorry to hear.Quote:
Originally Posted by C-BuZz
To find the reason is not easy, but for a mosfet to explode into flames, probably means that it was not fully turned on while transferring lots of current and because of that is dissipating huge amounts of heat.
The device responsible for controling this mosfet is the voltage regulator, so my guess is that your voltage regulator is no longer working properly.
Replacing this device is beyond the scope of a hobbyist, but chances are high that your memory also received a voltage that was too high and is damaged because of this.
You could try to test the board with an external power source (with both mosfets removed) to find out if it is worth investing any further.
how would I power the card with an external source?Quote:
Originally Posted by t024484
Im just guessing, but I dont think that the current flowing through the mosfet is supposed to be so high? Would have to be at least 12v to do damage like that? The chip it totally unrecognisable & basically turned to powder.
C-BuZz
Sluggo, your information is both wrong and misleading.Quote:
Originally Posted by sluggo
Wrong because the 32n03 has a total gate charge that is only half the gate charge of the 4420 that it replaced, wrong because a turn-off delay does not generate heat, and wrong again because the Infineon turn-off time is in fact shorter than the IR part.
Misleading because charging and discharging a capacitor does not generate any heat in that capacitor (or Mosfet), apart from the (mostly neglectable) loss factor.
And as a last remark, the switching frequency for Vmem is 660Khz and not 250Khz, and has never been overclocked by anyone as far as I now of.
you mean the main vpgu voltage regulator t024484? maybe the vgpu mods are staining the vreg in the wrong way and leading to the vmem mosfets burning out?
Nasty. Did you remove all mods including the pencil rub first?Quote:
Originally Posted by C-BuZz
You could always *try* to RA it and say "it just blew up!" lol
The mosfet that is burnt out is connected to the 12Volt, and your memory needs just 2 Volt. Suppose that 10Amp is flowing with a Drain to Source resistance of 1 Ohm, this will generate 100Watt, more than enough to blow your Mosfet to pieces.Quote:
Originally Posted by C-BuZz
What you need is an external power source capable of supplying 2Volt and preferably something between 15 and 20 Amp, and of course with a good short circuit protection. If you cannot borrow this somewhere, there are companies specialised in renting such electronic equipment.
No I am talking about the 6549 Voltage regulator for Vmem.Quote:
Originally Posted by flytek
This one works completely independant and is not influenced in any way by the 6568 voltage regulator for Vgpu.
it's a real pleasure to have you on the board t024484 ;)
He is a bit of a legend. Is that a crocodile embryo as your image? Or a Jurassic Park promo pic? :slap:Quote:
Originally Posted by flytek
Nail on the head. It is a Croco embryo.Quote:
Originally Posted by xit
Thought so....damn they're tasty. :stick:Quote:
Originally Posted by t024484
Anywho, back to the 7900's !!
I wasn't comparing the 32n03 to the 4420, I was comparing it to the IRF7821. The guy with the burnt chip was considering replacements for his fried 4420 and I was advising him on the pros and cons of the those two alternatives.Quote:
Originally Posted by t024484
I didn't say or imply that turn-off delay created heat. I mentioned it only as an illustrative consequence of the higher gate charge.
Discharging a capacitor most certainly does generate heat. A current has to remove the charge, and that current has to pass through a conductor with some finite resistance.
My mistake on the frequency, the Vgpu switches operate around 250kHz and that number stuck.
graham_h,Quote:
Originally Posted by graham_h
You certainly bought the 7900T/512 because you were looking for an improved performance.
Do you already have comparable results between the 7900GT and this new board, both with the same clock speeds.
That would be great to see.
Where are you guys getting your GW GTs from?
Hello guys,
I've roughly read through this lenthy discussions. Nice mod really!
I had a card vmod 1.4v, no vmem mods.
640/900... not a big overclocker but I am satisfied with it. :)
My question is about those MOSFETs.
In my case, they got extremely hot. Though the computer is quite stable even without heatsinks, I dare not put them.
After putting them on, the heatsinks still measure ~70 degrees celcius or up (well, with my fingertips... maybe not very accurate...)
so I managed to put a fan to blow some air onto the chip and the heatsink, and then... my motherboard temperature goes up about 5 degrees!!
How hot actually are those chips?
I am starting to think it is after all not a very stable mod...
I am curious about their functionality. Why are the two switching to each other? What do they actually do?
One chip is the voltage regulator for the memory voltage and the other one is a Mosfet connected to this regulator.Quote:
Originally Posted by timchen
Together with a third Mosfet they form a "switching regulator" to make 2Volt out of 12Volt with a relatively high conversion efficienty.
So the Vgpu modding you have done has no impact on the heat development of these components, but of course the memory overclocking to 900 has. Going from 650 to 900 let the current increase accordingly.
A heatsink is very good, and 60-70 degrees seems O.K.
The ventilator you are using is probably directing its flow to the measuring diode on your motherboard, thats why.
Where/how would I connect such a power supply to my board with no mosfets?Quote:
Originally Posted by t024484
I concur :toast:Quote:
Originally Posted by flytek
C-BuZz
You should connect it to where the Mosfet would have it's output pins...Quote:
Originally Posted by C-BuZz
The three source pins of the Mosfet that is blown away are directly connected to the 4 drain pins of the second mosfet. The 4 drain pins of the second Mosfet are soldered to this large area under the Mosfet.Quote:
Originally Posted by C-BuZz
This is where you have to connect your external powersupply.
Alright, do I just connect the +2v to the 4 x Mosfet's drain area like shown in the pic?Quote:
Originally Posted by t024484
Edit: UPDATE: I've been doing some research on the power requirements of the memory & found that they use 1.8v / 35mA
Could I use a Zener Diode Rated @ 70mA 2V to reduce the 5v on my psu? & would this setup be in the ballpark to power the card?
Thx
C-BuZz
I see. That must be the reason why the stock frequency of that ram is relatively low. Thanks a lot!Quote:
Originally Posted by t024484
I'm also interested in the fact that you've mentioned that on your board they measured no more than 50 degrees. Do you know why is that?
It's really nice to have some technical specialists like you here.
The memory chips are drawing more than 1Amp under full load, times 8 chips.Quote:
Originally Posted by C-BuZz
The least the power supply should therefore be able to supply is 10Amp.
The zener solution will not deliver.
Alrighty, & I thought I was getting the hang of electronics. Well I was right if I was powering 1 of the 8 x memory chips lol.Quote:
Originally Posted by t024484
So the memory uses 35mA * 8 = 280mA or 2.8A? This is unloaded situation. When loaded this figure can tripple or Quad? Or mabye more..
I have a power supply with 40A on 5v rail, & I have another with 15A on the 5V. Will I be able to use a Diode (or mabye trimpot? or somthing) to reduce the voltage to 2V & still be able to deliver the required Amperage? If so could you recommend a suitable Diode that would be up to the task?
Is my drawing the correct location for the input voltage? Branched across the 4 X drain pins?
As always, thanks for your wisdom & all the help you have provided. True Scolar & gentleman ;)
If there's a glimmer of hope for my poor 7900 i'm willing to try it (if it's cheap or free lol).
C-BuZz
what about a vsense mod on the 3.3 volt rail on the external psu to give you the 2volts?
if it deliveres enough amps that is
Each memory chip draws more than 1 Amp, times 8 gives 10 Amp.Quote:
Originally Posted by C-BuZz
If you would use Diodes to lower your 5Volt supply, each Diode will carry 0.6Volt. This voltage-drop times 10 Amp gives 6 Watt for one Diode.
You will need al least 5 Diodes to drop from 5Volt to 2 Volt, so you have to get rid of 30 Watt of heat generated in the Diodes. You will need heavy Diodes for that purpose and a large cooling body.
Trimpots are out of the question.
You are right, every attempt is worth trying as long as it is cheap.
Now you probably start to understand why a switching regulator is so well suited for the job, it is extremely small and super efficient.
there's a schematic of a "ddr booster" in the motherboard mod section.
really simple (the second schema at the end of the first page)
able to deliver big Amps using a mosfet.
It may be usable for u, have a look ;)
for the topic :
http://www.xtremesystems.org/forums/...ad.php?t=64453
and for the post with the good schem :
http://www.xtremesystems.org/forums/...2&postcount=25
Don't know if it can go below 2.5V, have asked ;)
It does not go below 2.5Volt.Quote:
Originally Posted by bouc
But for the test, it will do to have 6 diodes, submerged in a glass with water and ice cubes (Nvidia on the rocks).
You will now fast enough if the memory is still OK.
This is the easiest and cheapest solution. Any diode capable of handling 15Amp or more will do.
Thanks for your help, but I now have the required power supply.Quote:
Originally Posted by bouc
Please confirm that my drawing is correct. Also where should I ground this circuit?Quote:
Originally Posted by t024484
Thanks
C-BuZz
Your picture is not so clear to me, but the large metal area under the unburnt Mosfet is where you should apply the 2 Volt.Quote:
Originally Posted by C-BuZz
There are several 3mm holes in the board with large areas around those holes that are all connected to ground
Good luck
With card out of the system sitting on the table running exactly 1.85v through the 4 drain pins I get a vmem reading of 1.85v & the mem begins to warm up ever so slightly, which I suspect is normal. However, when I plug the card into the mobo the volts drop to 0v??Quote:
Originally Posted by t024484
When I power the motherboard up I get the same flashing light with my mobo, searching for video...
Cheers
C-BuZz
Your added power supply will have a + terminal and a ground terminal. The ground terminal from the power supply should be connected (securely) to one of the ground points near the mosfet. The closer to the mosfet the better. You can't simply ground the supply to the chassis.Quote:
Originally Posted by C-BuZz
I've grounded it at the closest point near the mosfet. Ground is running back through my custom power supply.Quote:
Originally Posted by sluggo
With the card out of the system the current flows no problem, & I can get good readings on various points of the card. The second I plug it into the pci-e slot the volts drop to 0v :confused:
Edit: Ok this cant be good. When I power the card up & take mem reading its in the negatives !!!
Cheers
C-BuZz
Pulled some of the PCB and solder of form the 0.2 Vgpu stop :nono:
I solder the whires on, but it came of twice because I was reluctand to use glue, the PCB got pretty hot.
I Marked a littel Blue Cirkel where my whire pulled some of the PCB of.
I decidet to use the alternate soldering point marked with Yellow on the attached picture.
Everything worked Perfect, Card booted, And 0.05 and 0.1 Vcore was addet and resulted in 1.35 Vgpu :)
But 0.2 Volt did not work :(:(
QUESTION.
1. Did I kill the connection to 0.2 Vgpu when I "pulled" some of the PC and solder of the card??
2. How Can I acheave more then 1.5 Vgpu now when 0.2 Vgpu incresment
dosent work :(??
Thx for your help everybody :) Will poste some oictures of the card tonigth.
Ít works now :D
Just dident solder it the correct placew, was VERY tired after 16 hour values seminar at work
I put a single rectangular Thermaltake Copper heatsink over those 3 mosfets that get hot, and it still gets very warm. However I can now (just) keep a finger on it whereas without heatsink 2 of those 3 chips literally burnt flesh. Is this what other are experiencing?
The rectangular heatink covers all of the large mosfet, and 70% of the smaller 2 mosfets surface area. Does it matter that the 2 smaller ones are not 'completely' covered by the heatsink?
Also, do you guys think that Copper or Aluminium heatsinks are better for this application?
Im running at 601/800 with 1.4 penmod btw.
Ideally cover surface 100%, and its the old story about alu v copper, alu doesnt conduct heat as well as cooper but it dissapates it faster, so its swings and roundabouts, I feel for that purpose you are doing copper is better, but the alu ones should be ok as it was stock with none.
I have some small thermaltake copper sinks for the back ICs that get very hot. The thing is that both sinks are bigger then the ICs that I want to put them on and especially the real small one. The sink is basically "hanging" over on both sides, the longer IC isnt as bad as the real small one. So there is quite a big amount of sink on both sides not touching anything basically. Is that okay? I dont really have the tools to cut the sink in half or something so that isnt really an option as this point and I dont want to buy different sinks as these low profile TT sinks are some of the smallest I've seen and I have a brand new full package of them (8). So will that still be okay if the sink is bigger then the IC? Will it impact the benefit of the sink? Any comments/suggestions are welcome. Thanks!
When the voltage goes to zero, what is your 5Volt supply doing?Quote:
Originally Posted by C-BuZz
Diodes have a voltage drop that does not go to far beyond 0.7Volt, so if you measure 0 Volt, it can only mean that your 5Volt supply is going down.
The difference with the graphic board outside the computer is that there is no clock on the mem chips, and that no activities are taking place within the memory.
Once in the computer, the memory is adressed from the GPU and starts drawing real current.
If your 5Volt 15A power supply cannot deliver the current, it can only mean that one or more memory chips are ruined.
I am sorry, but it was worth trying.
The bigger the sink the better the cooling, as simple as that.Quote:
Originally Posted by smut
Be sure the sink is not touching any other component to avoid short cicuits.
t024484, would you also vote for copper over aluminium heatsinks for these mosfets?Quote:
Originally Posted by t024484
Thanks for the reply, I was thinking the same thing about the bigger the better but I wasnt sure if I was right. Thanks for the reply. And yes I will make sure they arent touching anything else. BTW i'm up to 700 on the core now since doing the 1.55v mod. I was at around 660 with 1.4v.Quote:
Originally Posted by t024484
I was thinking about putting sinks on some more of the ICs on the back of the card because I will have 6 sinks left over after taking care of the 2 lower ICs. Do you think I should go ahead and sink some of the others? I know it wouldnt hurt but just wanted your opinion on it. I also put a more powerful 90mm fan on my side window panel last night because the original thermaltake fan that came with the case (its a TT Tsunami) was shyt and hardly moved any air! I put a better fan in to push alot more air over the back of the card and to also move alot more air over my RAM since theres really no active cooling for them. So if I were to put sinks on the back of the card they would be getting ample airflow. What do you think? If I dont use them on the card I plan on finding some spots on my mobo (asus SLI prem.) to put them on.
OMG we have the same case, vid card and heatsinks.Quote:
Originally Posted by smut
What brand side fan did you upgrade to? I found the fan to be neither standard 80mm or 120mm.
Its a 90mm fan bro, thats why you found it to NOT be a 80mm or 120mm :D.Quote:
Originally Posted by xit
But anyway I had an older 90mm TT fan laying around from back in the day when Newegg made a mistake and sent me a 90mm fan instead of a 80mm fan. They didnt want it back so I kept it, I remembered I had it last night and decided to switch it out since the side panel air flow is almost non existant. I never noticed until I was working inside the PC with the power on and I could hardly feel the air blowing from the side fan. I dust it every day just about because it get can dust on it rather easily. I just wipe it off with my hand and it gets the dust off of there. But yeah anyway find a 90mm fan and replace the stock fan if you'd like to get better airflow blowing on your card/mobo! Althought the stock fan is near silent and the more powerful tt fan I replaced it with is louder the trade off is worth it. Its blowing ALOT more air now which can only help!
edit: the link to the Thermaltake 90mm fan I replaced the old fan with. IM sure you can get it at newegg! Link to TT site here with fan model!
soldering the right point does help :D When you solder these boards use a 15W soldering iron. It doesn't take much heat for these connections and using too much can damage your board, as you found out.Quote:
Originally Posted by Nosfer@tu
using 3.3v rail with 30A.Quote:
Originally Posted by t024484
The voltage remains constant @ 1.9v with my power supply, the voltage actaully running through the video card drops to 0v when plugged into the motherboard (motherboard not even powerd up). Now, when I apply power to the motherboard this gives me a negative voltage reading. Again my power supply voltage is constant @ 1.9v.
Somthing must be cutting power to the card when it is inserted into the motherboard, as my power supply is supplying the correct voltage :confused:
Mabye I should just retire the old girl.
Cheers
C-BuZz
Mate, grab another AU$500 and get another one. This time don't go over 1.55v core (with Zalman VF900) and don't mod the memory at all! Also heatsink those evil little mosfets.Quote:
Originally Posted by C-BuZz
I think that while you were very ambitious, you wouldn't try to drive your car at 100mph in 2nd gear would you? :nono: I did once (accidentally changed down) and blew all sorts of parts.
Thanks for that. Did you notice the fan speed on the specs on that page?Quote:
Originally Posted by smut
285010% RPM LOL !
Thanks for your words of wisdom. But alas you are right. I definately wont be going that slow in second gear.. I'd want at least 150mph out of second. If it dies, it dies.. I can live with that.Quote:
Originally Posted by xit
@t024484, thx for your efforts, but I now fear the card has suffered long enough :) Time to put her back into the box she came in...
C-BuZz
I dare you to try to RA it....just clean it up a bit and say "it just blew up! And I'm mad coz it could have taken out my entire PC!"Quote:
Originally Posted by C-BuZz
:banana:
with 2 missing mosfets/wires hanging off it/diodes stuck it. Also 50K vgpu resistor soldered in place? No heatsink/dielectric greas all over it/insulation stuck to it front & back..Quote:
Originally Posted by xit
Hmm, wonder what the'd say?
C-BuZz
Guys, bios mod with Nibitor 2.8 worked for 7900gt xfx extreme edition 520/1500 ? i need change volt 1.2 for 1.4 , work??
[]´s Krause
Krause: if you have a question such as this (IE, off topic) start a new thread. bios mod is not the same thing as volt mod. Also, don't post your question in multiple threads, again start a new thread.
I am puzzled by your mail. So you did not use a 5V supply with 5 diodes, but a 3.3Vsupply (probably a computer PSU) turned down to 1.9Volt.Quote:
Originally Posted by C-BuZz
Is this a separate power supply, or are you using the same supply that is feeding your computer?
How did you turn the power supply down to 1.9Volt?
How can on one end the power supply generate 1.9Volt, and have 0Volt on your graphic board with only two wires connecting the two?
Are you using very thin wires with a high resistance?
Anyhow, what I understand is that the board measures 1.9Volt outside the motherboard, and 0 Volt plugged on the motherboard, with no power on the motherboard. This means that there should be a connection between your 1.9Volt and something outside your graphic card, which is normally not the case.
I see three possibilities:
1) you have connected your 1.9Volt on the wrong position, it should be the large area under the unburnt Mosfet.
2) the Burnt mosfet area has a short circuit between the 3 source pins and the large drain area underneath connected to the 12Volt, or
3) somewhere else on your board is a short circuit because of the burnout between the 1.9Volt Vmem and either the 5Volt or 12Volt supply.
1) and 2) can be corrected, 3) is the end of this experiment
lol thats funny! I didnt notice that. Heres the real RPM of the packaging it came in, its 2850RPM at 35 dba.Quote:
Originally Posted by xit
After a lengthy discussion with my local electronics store about my particular project, they recommended that I use the 3.3v/30A rail on a computer power supply. So I used 30A Bridge rectifier with 4 X .4v diodes inside it. Which drop the voltage .4v per diode. So total voltage drop across the 4 diodes is 1.2v. Leaving 2.1v remaining. They also recommended that I put a load on the circuit as the bridge rectifier does not actually begin working until there is a small load on the circuit. So I use 5W/1A resistor to put a small load on the circuit. After installing the resistor my power supply voltage was now 1.9v exactly.Quote:
Originally Posted by t024484
I will take pictures of this experiment as it's quite difficult to explain.
Pics to follow.
C-BuZz
ok. i finally hit the limit on my stock coolers. i got to 1.5v mod but temps become an issue after that.
one of my cards is artifact free at 661/498(1992mhz) soo close to 2ghz.
and the other doesnt like me at the moment with the ram, but the core is better?!? 668/440 (1760)
i have lowere the load temp by 10*c by lapping the stock coolers and shaving a mil or so off of each leg. i also knoticed that the air from the fan is being blowen through the hole in the back of the stock cooler, so i taped it up, and its helped by another 5*c :D
i havent got any hot chips, which is strange, but i have to 2x80mm, 1x 120mm and 1x 120mm delta fans blowing at the gfx cards. :D and i still cannot here them over my chiller. lol
Pics.
Bridge rectifier used: (imageshack)
http://img60.imageshack.us/img60/527/34nd.jpg
Red wire +1.9V soldered to 4 x drain pins where mosfet used to be. If this is the wrong place, please draw where it should be connected. Black wire = ground.
http://img343.imageshack.us/img343/4403/18uq.jpg
With my power supply switched on I can test the vmem voltage & get good readings from many different points on the video card:
http://img343.imageshack.us/img343/9384/dscn04044ex.jpg
Here is my custom made power supply with 5W/1A resistor installed. This power supply is being used to power the card only. The motherboard has a separate power supply:
http://img53.imageshack.us/img53/5889/dscn04058fk.jpg
Now, when I plug the video card into the motherboard (motherboard not powered on) & I try to take a voltage reading from the card it reads 0v :confused:
But here's somthing to make you want to pull your hair out. My 1.9v (RED WIRE) in second pic above, if I take a reading at the solder point (where current is going into the card) it reads 1.9v!!!!!!!!!!!! Yet 0V @ vmem Measure point or any other point on the card.
I cant figure out what is going on, & why it is reading 0v when plugged into the motherboard & 1.9v when sitting on my table :confused:
C-BuZz
what heatinks are you using there? I'm planning on getting the OCZ BGA (aluminum OR copper) ramsinks for the IC's by the vmem measure point, so the ones you pointed out that they get boiling hot.Quote:
Originally Posted by coolingbug
OR you guys think TSOP heatinks will be better? and where can i buy them online US retailer?
You have placed a bridge rectifier behind the 3.3Volt. In that case you use only 2 diodes giving a voltagedrop of 0.7Volt each, resulting in 1.9Volt.Quote:
Originally Posted by C-BuZz
So far so good.
But, the red wire is connected to the large metal are under the BURNED mosfet (this is connected to the 12V) and not, as instructed, to the large metal area under the UNBURNED mosfet tho the left (Q501).
Your Black wire is connected to some funny point, but certainly not to ground. You should as instructed use the large round metal area around one of the 3mm holes, like the one to the upper left corner of the board.
Ok, so the red & black are in the wrong place :slapass:Quote:
Originally Posted by t024484
Im not to sure what you mean by the large metal area under the BURNED mosfet. There is no large metal area under the mosfet. There is only 6 X Contact points under the mosfet. As you can see in the picture below. Please confim that I should attatch the 1.9v to the 6 x contact points circled in ORANGE.
Thanks
C-BuZz
Did the mod on the 2. Card now.
Accualy it came out quite good :)
But it was very hard to remove resistor J :(
Did it before on ic7 Max3 and the ic7 max2 aka Ic7-G.
But here it is quiet hard :(
The 6 contact points that you mention are solder dots on a large metal area.Quote:
Originally Posted by C-BuZz
You connected to the metal area under Mosfet Q502 (the burned Mosfet), it should be the metal area, or as you prefer to call it ""6 contact points" under Q501. I have nothing at my disposal here to make drawings, but what I say should be crystal clear.
The memory voltage regulators on my card are crapping out and I am most likely going to have to go through with an RMA. If I get one of the new XFX 7900GT EE cards with the Black PCB, will I still be able to do the gpu volt mods? Thanks
Yes you will, but it will crap out on you as well...mine did..it's the second one I might add.Quote:
Originally Posted by tranceaddict
They definately will have to deal with this issue, doesn't matter if you mod or not...the mem voltage reg craps out sooner or later.:mad:
I just ordered a 80mm Zalman thin fan to put on the back of the card and blow on all the regulators and MOSFETs. Hopefully this will prevent it from happening again.Quote:
Originally Posted by Clint
i've done the 1.4 ink vmod
is there a 1.45-1.5vmod with ink ?
Did you read the thread dude? It shows how to doi t probably 5-6 times lolQuote:
Originally Posted by Supertim0r
Kind of an odd question for an overclocking forum, I know, but what if one was interested in undervolting a 7900GT? All the overvolting mods seem to be VID based, I was wondering if resistor mods were possible for this controller, or if there were any issues with this and that's why none are being posted.
Could I take the non-inverting input of the internal error amp (pin is called FB) high through a VR (set at max resistance, decrease resistance to increase voltage on the pin and reduce output voltage)? The only problem I'd forsee is according to the PDF for the controller undervoltage threshold is 82% of VID, which would give me space to ~0.95V. Doubt the GPU could even go this low, but if it can and this does end up being a problem I could just change to a lower VID code (say 1.1V) to offset the threshold down to ~0.85V, or basically as low as I need with the right VID.
Thoughts, suggestions, tips? Sorry I know this isn't terribly exciting for you overclocking guys, but I like tinkering with efficiency and low noise computing, so it's exciting for me. Any help would be appreciated. :)
I would prefer a resistor mod as it would easily allow me to undervolt until I encountered instability. With VID codes this would be very difficult, as I'm looking for the lowest undervolt possible but have no idea how low I could go. I guess it would be possible to rig up some kind of five switch system to pull each VID pin low or high, this would be a lot of work (almost to the point of not being worth it I think).
Or you could just flash the bios to 1.1v. The bios doesnt support OVer-volting but it Does support under-volting.
Is this possible? I opened up the BIOS in NiBiTor and found an unused VID code in the tables. Assuming it was 1.1V, I selected it and flashed my BIOS, but unfortunately it didn't seem to do anything (no temp change and measured with a multimeter, Vcore remained at 1.2V). I figured since 7900GT is default 1.2V in 2D and 3D, reference design doesn't bother including the circuitry to allow other voltages to be selected in BIOS.
But maybe I did something wrong? Please help if you know how. :)
BTW, my card is an XFX 7900GT Extreme (520/1500) if it helps to know.
EDIT: Also, could soft-start cause any problems? Accoring to the PDF:
So how high could I pull FB before causing problems. What is the internal ramping reference voltage? I actually did an undervolt mod for an X800GTO previously, and ran into a problem where below a certain voltage the card would run fine if it was already running, but when the computer was shut off and restarted, the voltage regulator would not turn on until Vcore was raised to this level. My best guess was that the problem was related to soft start, which this IC had as well, but I wasn't completely sure. It wasn't too hard to work around, just lower voltage while the card was on to find the lowest undervolt, then use VID codes to permanently set that voltage and remove the resistor mod. Was kind of a pain in the butt though.Quote:
The ISL6568 also has the ability to start up into a precharged output, without causing any unnecessary disturbance. The FB pin is monitored during soft-start, and should it be higher than the equivalent internal ramping reference voltage, the output drivers hold both MOSFETs off.
To get around something like this could one pull the inverting input of the error amplifier low, as opposed to pulling the non-inverting input high?
Sorry for so many questions, and thanks to any with the knowledge and patience to read my posts and explain this stuff to me. I'm actually going to college to be an electrical engineer, but am still getting the physics, calculus, chemistry, etc. out of the way -- haven't got to any of the real interesting classes yet. :)
Can anyone help me out here... I am currently leak testing my h20 loop as I re-did the tubing layout and since I still have the vmem mod and vcore and vmem read points left to mod on the card (ill post my pics up once I am done) I just wanted to ask. Are there any vcore and vmem readpoints that are closer to the edge of the card away from the motherboard? Preferably somewhere closer to the middle of the card because I can no longer us the standard readpoint as shown in the vr-zone mod as my vcore mod is rather big (full 0.5v adjustment from stock to as high as I want)
I soldered 2 wires for the Vcore read points. Look athe Pictures of 7900GT mod thread. You will see it on the last page towards the bottom.
wow this masscool VF1-plus rocks. With the 1.5 vold mot, 2.23vmem mod and switching frequency mod i was idling 42-44 C with my zalman vf-700cu and now i idle 35-36 and my load is 49-51. My load use to be 69-71 while running atitool and other similiar programs o_0. so a 20 degree decrease -_-
Its also possible your cores cannot handle anything further. I hit walls like that on my old 7800gtx all the time. I could do 700 core + 50 delta @ 1.68v but couldnt get anything more even trying voltages up to 1.8v! I havent tested out my 7900gt, but the thing I would look into is pci-e frequency. Whats the highest you have been testing with because you would probably need to try around 120 - 125mhz.
I wouldnt recommend running PCi-freq. that high. Its been known to screw up Windows installs for quite a few people over 120ish. Its never happened to me but Im pretty sure the possibility is there. I never go over 110ish because it doesnt do anything after that for Me.Quote:
Originally Posted by coolingbug
What is the benefit (or not) of raising the PCI-E frequency? Achieve high clock but by how much usually? what are the risk? frying the card/damage mobo? Didn't AGP have a frequency but it was not recommended to touch that at all? :confused:
RE AGP OC-ing, it really depended on your board & AGP card in use. DOnt touch!?...Hmmm. I had an Abit board that allowed my hercules GF2 ULTRA to do 105Mhz AGP speed...and you could see it in the Benchies...then AGP went 4x & OC-ing AGP was no longer possible! Same card, different board & then we had 8x AGP, ....OC-ing only became possible on an ASUS P4C800E Deluxe for me, thro an ATi Radeon 9800 ULTRA, but it made the benchies run with less marks for some reason.Quote:
Originally Posted by iRoNeTiK
You might see about 10 to 15mhz over 100 on PCI-e but probably wont see any appreciable increases for it. If you go too far, you might scramble your registry/install, I know someone who did.... so be careful. :)
So what do you see for increasing below 110ish?Quote:
Originally Posted by smut
lol thanks for the info, could always do repair/checkdisk and my OS is on its own HDD so wouldnt lose anything, ill look into it more but i may or may not do it. Ill probably just go to the 1.55 vmod from 1.5 and call it good and enjoy my LONG hard work i put into this card heheQuote:
Originally Posted by Kin Hell
Yeah, just be sure to cool those Voltage Regs just to be sure! :DQuote:
Originally Posted by iRoNeTiK
Oh yah i got my vf1-plus last night and it came withheatsinks so i used those and took 2 heatsinks from my zalman vf700cu and put them on the voltage regulators. Those are the ones by the vmem measure point right? the ones where some of them been melting? If it is then those are the ones i got covered.
I should post pictures, maybe later tonight/tomorrow.
My clock speeds went up a bit higher raising the pci-freq. and a lil more stable as well. But once I hit 110 they did not go any higher so I stopped there.Quote:
Originally Posted by Kin Hell
I have a Forsa 7900GT that I have done the 1.4 penmod on and no other mods. I have though upgraded to a Zalman VF900 cooler and also put a copper heatsink on the 3 evil hot mosfets.
When I clock much above 600MHz core, I can audibly hear the GPU making squeeling sound. This reeks to me of simply taking the GPU clock too high, regardless of voltage or anything else. I have however settled on a clock of 625/800 with 50 delta, giving almost exactly 10000 in 3Dmark 05 (just an Athlon 3200+ CPU @ 10x240mem)
Does anyone else notice this sound? I hope it's not drowned out by fan noise! :confused:
The voltage regulator knows nothing of frequencies of the GPU.Quote:
Originally Posted by coolingbug
Since you hit the "wall" at 1.47V and at 1.68V it is unlikely that it is an Over Current Issue (although it would be relatively easy to replace this resistor)
The Over Voltage issue is positively influenced by increasing the frequency of the voltage regulator, but since you have tried that as well, this can almost be excluded.
Since both of your SLI boards are having this "wall" my guess would be that the frequency generator within the 7900GPU cannot produce or is inhibited to accept this frequency.
C-BUZz,
Haven't heard from you over the last week. What happened with your experiment with the external power supply?
The Squealing noise you are hearing is the Card complaining its NOT getting enough Voltage. My 7800GS card made the same noise cos my 550w True COntrol PSU was fugged on one of its rails. Check just what your PSU is pushing out on its 12v, 5v & 3.3v lines. They should all be above these values under whilst under load. Also to proove further, take off as many devices as you can, Case lights, CDROMS, Hard drives etc & then see if you can go further with the card. Anything over 3 hard drives is going to cause issues on most quality PSU's eventually. Hard drives seriously "tax" the PSU.Quote:
Originally Posted by xit
Your modded card is demanding aomewhere in the region of 110Watts alone.
I'll hope you dont have a nackered PSU, but don't hold your breath, check it out! :)
Is than an Opteron 180 you got there?Quote:
Originally Posted by coolingbug
I'd be VERY satisfied with 'only' 750 core on a core which runs 450 standard... ;)
So would i. :)
1.4Vmod netted me 650-655 on the core with the stock cooler but temps are becoming really frightning at this stage> 76C under load!! :eek:
Like others here i found that 1800 on the RAM was completly artefactfree where it gave slight artefacts prior to the mod.
Soon as my EK full cover waterblock arrives i'll be benching again.
Got 11K in 05, 23K in 03 and 37K in 01 so far.
Here's 11K for '05:
Hi t024484,Quote:
Originally Posted by t024484
I've given up & I dont want to keep hasseling you heh. Card is just dead. No signs of life whatsoever. I've tried every possible combination to power the card but no luck.
Thanks anyway for all your help on this matter.
C-BuZz
I had an accident while volt moding the D508 "B" point, and it now no longer bumps ups the voltage, is there any other points on the board (other then replacing resistors) that i can bump it up to 1.4 or 1.5 volts?
Ok i see now, just to make sure though,
in this pic
http://www.xtremesystems.org/forums/...7&d=1143538288
The lower set of points act the exact same way as the upper ones?
So i can connect the 2 yellow "B" points to the 3 "C" points to get 1.4?
and 1.5 if i include the purple points?
Yes. The lower points have the same effect as the upper points.