Frying 7900GT boards

[t024484]

When my first 7900GT board started to behaving strangly, I thought it had to do with all the experiments and changes I had made to the board.

I bought a second 7900GT board, and modified it to 1.5 Volt GPU and 2.14 Volt Mem and added a Zalman VF 700.
GPU temp was never higher than 62 C, and everything worked fine for two whole weeks until yesterday. As a first sign, the board started dispaying huge artefacts in "Deep Freeze", the last test of 3Dmark06. After that happened, I had to restart the system to get the display working again. 3Dmark05 still worked troublefree at that time.
At the end of the day, ever more problems ocurred, and now I cannot even run 3Dmark05 anymore.

It seems to me that the G70 GPU is not able to have 1.5 Volt on its pins permanently as well in the 2D as in the 3D mode.
With 1.5Volt it seems to be an accident waiting to happen.

With my experience based on two boards, I can only advice: Do not apply more than 1.4Volt on the GPU, or even better have a board that only has a higher GPU voltage when in 3D mode, like the EVGA SS serie.

The 7900 GPU is nowhere as robust as the 7800 GPU, since I never had the slightest problem with the last.

I am considering now to buy a 9750, mod nothing at all and return it under RMA if anything goes wrong. All I need is the right BIOS for my P5WD2-E.

[couppi]

Hmmmm... So even with the switching mod and everything, the board still fried?

[t024484]

Hmmmm... So even with the switching mod and everything, the board still fried?

Yes with all the mods. The board has been modded with the greatest care with SMD resistors only: 50Kohm for the 1.5Volt , 100Kohm for the switching mod and 10Kohm for the 2.14Volt Vmem mod.
From a distance it is hardly possible to see that the board has been modded. All possible negative side effects of having long wires and/or having mechanical instabilities have been completely avoided by using these SMD resistors, but nevertheless .....

[ManagHead]

So you think it's the core, not mem which has failed you?

[Krizby87]

did you sink all the hot ICs on the back of those boards ? those could be the cause for artifacts as I have read from EVGA forums.

[DTU_XaVier]

I've had trouble with that freezing thing as well, am able to run 3dmark05 no problem... Haven't tried 06 yet, but sounds like i should.. As far as I have been told, it's the MOSFETs and voltage-regulator on the back of the board which overheats... Try de-clocking your RAM
I'm running mod-free btw

Best Regards

[quijonsith]

All I've done is ground the two points at the top of the board that yield +.1v and +.2v respectively. I've had my card since the day they came out and have had the mod running just about as long and no problems here. I have the eVGA CO 500/1500 card.

[t024484]

So you think it's the core, not mem which has failed you?

The board still works in 2D with the high memory clock, only when 3D starts I am getting problems.
Going from 1.2 to 1.5 Volt is a 25% increase for Vgpu.
Going from 2.04 to 2.14 for Vmem is only 5% increase.
The memory is similar to the memory on the 7800Gt board.
According to Samsung, the max voltage on the memory is 2.5Volt.
So if I take all this into account, I bet it is the GPU that is making trouble.

did you sink all the hot ICs on the back of those boards ? those could be the cause for artifacts as I have read from EVGA forums.

I have made proper cooling with heatsinks to the Mosfet and the voltage regulator. Temp stays below 40C

I've had trouble with that freezing thing as well, am able to run 3dmark05 no problem... Haven't tried 06 yet, but sounds like i should.. As far as I have been told, it's the MOSFETs and voltage-regulator on the back of the board which overheats... Try de-clocking your RAM

Thanks for your suggestions. It is not "freezing" that is happening. The board starts wildly artefacting when in 3D, and when I interupt and go back in the 2D mode, the 7900GT is displaying partly rubbish on the screen and is no longer controleable. Only by restarting the system, everything is working properly again in 2D.
Declocking the GPU and/or MEM does not help me in any way !

[s e t h]

are you using coolbits or atitool for clock adjustment? or something else?
i found some cards really don't like coolbits
maybe try nibitor bios program and see if you can change the delta clocks and/or some other settings.
just a couple of ideas.
good luck

[Shark-357]

I lost 2x Asus EN7900GT in three weeks (first GPU@1.4V with VF900cu died in a week and second one died in two weeks it was watercooled GPU@1.55V ,ddr@2.2V it did nice 735/1030 artefact free ).
I had no problems with Club3d 7900gt which i ran @ 1.55V,2.2V for two months and is still running fine with the new owner.

[DTU_XaVier]

Thanks for your suggestions. It is not "freezing" that is happening. The board starts wildly artefacting when in 3D, and when I interupt and go back in the 2D mode, the 7900GT is displaying partly rubbish on the screen and is no longer controleable. Only by restarting the system, everything is working properly again in 2D.
Declocking the GPU and/or MEM does not help me in any way !

What you describe there is the same problem I have when oc'ing... At stock it runs fine, but it takes almost no oc before it starts acting like that... it runs benches (3dmark and AM3) with pretty high clocks, but no games are possible... It was possible the first 2-3 days, then it started... Dunno if that's the same thing you've expirienced??

Best Regards

[humeyboy]

This problem is well know on some brands of 7900's esp evga and xfx which have problems, some remove deltas in bios as thay can be as high as +40 which gets higher as you overclock and if your card only overclocks for a period of time then wont run anything but stock, you need reflash bios and it will again overclock, someone pointed towards not using coolbits to overclock till this is fixed.

There is many threads about 7900's issues esp them 2 brands but not just them.

[t024484]

are you using coolbits or atitool for clock adjustment? or something else?
i found some cards really don't like coolbits
maybe try nibitor bios program and see if you can change the delta clocks and/or some other settings.
just a couple of ideas.
good luck

I am using Coolbits, and also Ati Tool and Riva Tuner.
I have noticed no difference between the three when I change the Clock.
The point is however that everything worked fine for two weeks, and now the damned thing starts to desintegrate.
I have tried delta clocks with Nibitor of 0, 20 and 40 but with no effect.
Something starts to misbehave as soon as 3D starts, no matter what the clocksetting is.

I lost 2x Asus EN7900GT in three weeks (first GPU@1.4V with VF900cu died in a week and second one died in two weeks it was watercooled GPU@1.55V ,ddr@2.2V it did nice 735/1030 artefact free ).
I had no problems with Club3d 7900gt which i ran @ 1.55V,2.2V for two months and is still running fine with the new owner

This is just another prove that the 7900GT is still very vulnerable for whatever, and that Nvidia should spend some time in getting things better under control.

This problem is well know on some brands of 7900's esp evga and xfx which have problems, some remove deltas in bios as thay can be as high as +40 which gets higher as you overclock and if your card only overclocks for a period of time then wont run anything but stock, you need reflash bios and it will again overclock, someone pointed towards not using coolbits to overclock till this is fixed.

As far as I am corrrectly informed, all boards come from the same manufacturer in China, so if this is true, there are no differences, apart from having more components on the board like the EVGA SS.
Reflashing the Bios did not bring anything, nor did changing the delta.
I have also reinstalled the Forceware driver and Direct X9C, but all with no effect.
I will see what happens if I do not install Coolbits.

[Highland3r]

The issue with these cards is the memory NOT the core...
The mem mod you did would probably have accelerated any damage being done to the memory regulation, bet you didnt add any extra cooling to the rear of the card either?
Some of the cards seem "worse" affected than others, if they make it passed the first 2-3 days of their lives without artifacting/having issues with Deep Freeze in 06 then chances are it'll be fine.
Just had an XFX card die on me, ran fine for the first few days then the deep freeze issue presented itself. From then on things got worse and worse.
Not getting XFX again, probably plump for another EVGA CO as this ones been pretty decent so far...

[Lerun]

The issue with these cards is the memory NOT the core...

I tend to agree with you. But I think it's a combination memory / volt controll components.

Maby it's a crapy batch that does not hold a steady voltage regulation as the tempratures rise. After the inital break (pushing the components to the first breakdown), the behavior presist or are easier to provoke.

My eVga also started making some high pitched sounds, like a MOSFET or IC on the last leg. Pointing to problems with freqs switching.

I just don't know what's up with some of these cards.

[BigDov]

I've had my XFX at the speeds in my sig since day one of ownership, no mods, no problems to report either.

[Lerun]

I've had my XFX at the speeds in my sig since day one of ownership, no mods, no problems to report either.

I get the problem every time I loop DeepFreeze in 3Dmark06.
First run went fine, every run after initial gets massive tearing.

Tried all different drivers...and so on.

[t024484]

The issue with these cards is the memory NOT the core...
The mem mod you did would probably have accelerated any damage being done to the memory regulation, bet you didnt add any extra cooling to the rear of the card either?
Some of the cards seem "worse" affected than others, if they make it passed the first 2-3 days of their lives without artifacting/having issues with Deep Freeze in 06 then chances are it'll be fine.
Just had an XFX card die on me, ran fine for the first few days then the deep freeze issue presented itself. From then on things got worse and worse.
Not getting XFX again, probably plump for another EVGA CO as this ones been pretty decent so far...

I already mentioned that I have cooled the Mosfet and the Voltage regulator, and that they do not get hotter than ca 40 C.
Why are you so sure that the Memory is causing the trouble, it will be interesting to hear if you have more info to share.
My Vmem seems as stable as a rock as far as I have seen so far, but I will investigate this further with my osciloscope to see if anything is behaving strangely around the voltage regulator.

[Highland3r]

I already mentioned that I have cooled the Mosfet and the Voltage regulator, and that they do not get hotter than ca 40 C.
Why are you so sure that the Memory is causing the trouble, it will be interesting to hear if you have more info to share.
My Vmem seems as stable as a rock as far as I have seen so far, but I will investigate this further with my osciloscope to see if anything is behaving strangely around the voltage regulator.

It seems (from reading around and personal experience) that if the cards going to die it will. I'd suggest its not the mods which are killing the cards, (although they may well accelerate death) just the fact that the cards going to die anyway.

The reason I say its the memory which is the issue is thus: Overclock a 7900's memory too high games/benchies "appear" to run clean, but on going back to desktop it seems to flash/corrupt. Downclocking the memory fixes the issue.
Just lost an XFX card, the regs were sinked and a 120mm fan blowing over them. 3 days after getting the card (and running SLI) arti's/hangs started to occur. The usual symptoms of a dying card.
On running 03 and other 3d a day or 2 later the card produces the same flashing screen etc as the EVGA did when the memory was clocked too high.
To me that suggests the memory is the cause of the issue in this (and probably other) cases.

[t024484]

It seems (from reading around and personal experience) that if the cards going to die it will. I'd suggest its not the mods which are killing the cards, (although they may well accelerate death) just the fact that the cards going to die anyway.

The reason I say its the memory which is the issue is thus: Overclock a 7900's memory too high games/benchies "appear" to run clean, but on going back to desktop it seems to flash/corrupt. Downclocking the memory fixes the issue.
Just lost an XFX card, the regs were sinked and a 120mm fan blowing over them. 3 days after getting the card (and running SLI) arti's/hangs started to occur. The usual symptoms of a dying card.
On running 03 and other 3d a day or 2 later the card produces the same flashing screen etc as the EVGA did when the memory was clocked too high.
To me that suggests the memory is the cause of the issue in this (and probably other) cases.

Thanks for your reply. I will try to dig further into the problems, because there are things that are illogical to me.

1) I run my memory at 1750 Mhz as well in 2D as in 3D mode. No signs of trouble are there as long as I stay in 2D with this high memory clock.
The memory is unaware wether you are using the system in 2D or 3D, so how could the memory cause trouble only in 3D ?
2) Samsung gddr3 memory is used all over the world, in all kind of graphic boards, and I have not heard of similair problems in other high end boards.
Also the 7900GTX with the same memory chip in a selected 1.2 spec does not seem to cause any problems with frequencies up to 1800Mhz.

My suspicion is much more that it has to do with electro erosion within the GPU.
With the smaller 90 nm details, it is more critical then ever to keep signals carrying different voltages apart far enough, more critical then it was with the 7800 GPU with 110 nm details.
When you have 1 volt difference on traces only 100nm apart, this translates into 10.000Volt/mm. This is enough to fire a spark plug.
Things are made even worse when you increase the Vgpu.

If signal traces are too close, you will see an avalanche (or electro erosion) taking place after some time, which in effect causes the same destruction as a short circuit. Just by small individual differences in the production process, on one chip the distance can be several nm more then on another chip, which leads to one chip dying quite soon and the other staying alive for some time.

So it is much better if you only allow the higher Voltage when in 3D mode like the 7900GTX and the 7900GT SS are doing, that's probably the reason why, because most of the time a PC is working in 2D.

If my suspicion is true, it will force NVIDIA to come with a redesign quite soon.

[Lerun]

In 3D there must also be other parts of the core that get activated. Parts that lay dormant in 2D.

Strange it's tearing in whole lines as I've heard is a indication of memory problems.

What about different power IC beeing used in 2D and 3D?

Just tossing some ideas out there.

I've been reading that alot of ppl get issues in DeepFreeze (3Dmar06), including myself. Downclocking the core does nothing.
Taking the mem to nVida stock speeds is another ting, DeepFreeze passes without insident; if one does not take into consideration the crapy score

[t024484]

In 3D there must also be other parts of the core that get activated. Parts that lay dormant in 2D.

Strange it's tearing in whole lines as I've heard is a indication of memory problems.

What about different power IC beeing used in 2D and 3D?

Just tossing some ideas out there.

I've been reading that alot of ppl get issues in DeepFreeze (3Dmar06), including myself. Downclocking the core does nothing.
Taking the mem to nVida stock speeds is another ting, DeepFreeze passes without insident; if one does not take into consideration the crapy score

It is not tearing whole lines, firstly things are happening randomly and then the screen starts flashing, which only stops after a complete restart. The same power IC´s used for 2D are also used for 3D

[Highland3r]

You can also fix it by reloading the gfx drivers via rivatuner

[SF3D]

My 7900GT is dying. I have used 1.5V all the time from day one, but randomly screen flashes, goes black or tearing. Only help is that Max Vgpu is 1.5V or under and no mem mod. I think my problems are mem related.

Core can still do 700 but mem do only 1600. Anything over 1600 and screen starts to act weird or flash after 3D mode.

Every temperature are ok.

[Krizby87]

if you have a samsung ram on the card, I suggest flashing it with EVGA beta bios as this Bios is supposed to fix issues with the memory -> flashing particles, screen goes black. See if it help.

http://www.yousendit.com/transfer.ph...5C91575BA7C88E

I also have the screen flashing thing when I overclock memory too high, GPU core would downclock to bios clock, FEAR benchmark would run fine (with lower GPU clock) until I quit the game and the screen went nuts, seems like the refresh rate goes down to 2 or something, if I move the mouse the screen will flash, every movement on the screen will make it flash.