7950 ... is my card about to die ( long sry )

[Nomercy1]

Hey guys, making a fresh post here because i dont want to continue to blow up the 7059 owners thread with my problems. Looking for advice/ expirience with a issue with my gpu.

http://www.newegg.com/Product/Produc...82E16814202006 is the card im having issues with.

Back when i got this card i overclocked it to 1125mhz core/1300mhz memory and ran a lot of benchmarks for a week or so to verify that it was stable. It passed a lot of tests and i stopped messing with it as i figured i had dialed it in. A few days ago ( 5 months later now ) i decided to see how its doing and noticed the card was now throttling after 5 min of use in either a benching program or a game.

The temps on the card are not exceeding 62c vram temps 63c under full load for multiple hours. Power is set to +20% as it always has been.


Here's where it gets interesting. After much more benching i came to the result that 1075mhz core was running stable without throttling, anything over would throttle. benched for about 4 - 6 hours and then began to play battlefield 3 ect at these speeds without issue for several hours. 97 - 100% gpu usage no throttling. The memory i ended up going back to stock with 1250mhz as it showed a improvment in benches over 1300mhz, anything over would crash and 1200mhz mem would crash.

Now TODAY those same speeds are throttling and going a step further, crashing BF3, 3dmark and Heaven 4.0/Valley benches. At this point i put the card back to stock and dropped the +/- power to 0 as is normal and the tried more testing. Ther card would throttle beetween 733mhz ( weird) 850mhz ( normal clock ) 925mhz ( boost clock ) Constantly. put the power to +20% solved this issue and the card would run at 925 boost clock constant.

I thought Perhaps it was the power supply so i grabbed working 100% 700w PSU and hooked it directly up to the video card ONLY and the results were the same, when i clocked higher than 1000mhz core it would throttle after 5 min in any program/game. same results at stock clocks with the power slider at 0 ect.


So its definetly not the power, and the temps are fine. The performance of the card has significantly gone downhill in a matter of a few days. Now at any speed even stock speeds the GPU wont get used up to 100% ever. It is constantly going beetween 65 % - 99% where as it usually is 97 - 99% always. This is definetly indicating somethings wrong. Its not exactly throttling at stock speed ( going from 66 - 99% usage every second and dropping the core speed) its just flat out not using the card to its fullest. No hardware or bios or driver changes have been made to this card. It is absolutley running at stock voltages. I cant help but think tommorrow the card will be dead. I have never had a issue like this come up and any advice / theorys would be awesome. If you need screenshots of anything just let me know.

 

[MeldarthX]

Drivers?

temps on the chip itself might be fine; but what about the vram? voltage reg?

You might need to redo the tims on it; and possibly add ram sinks for the vram if its not being directly cooled.

Thing is - its running while its inspec; but throttles when you try to OC - check with Sapphire - if they allow you to take off the heat sink - reapply tims and check vram cooling...

If not I'd talk with them about RMA -

Last which drivers are you running? they sometimes can have effect; but never found an effect like this before

 

[zaydq]

Try pluggin in another 2x6 or 1x6,1x8 into the card and see how it works. You could also try flashing the bios to see if your problems get solved. I wouldn't be surprised if the boost bios is conflicting with your overclocking.

Also, check your Vrm temperatures, the memory could be running hot. On the 7950, you'll see majority of performance increases from increasing the core clock, since the 7 series is not short of memory bandwidth by any means.

 

[Silverforce11]

check VRM temps, if its one of twin fan cards with NO heatsinks on VRM it could have overheated over a long period of time (>115C) and is starting to degrade. as it degrades or overheat, the current conversion becomes less efficient causing instability and throttling.

 

[Nomercy1]

check VRM temps, if its one of twin fan cards with NO heatsinks on VRM it could have overheated over a long period of time (>115C) and is starting to degrade. as it degrades or overheat, the current conversion becomes less efficient causing instability and throttling.

VRM temps according to both HWinfo64 and GPU-Z ahve never exceeded 63c in any of my benching or gaming.

 

[Nomercy1]

Try pluggin in another 2x6 or 1x6,1x8 into the card and see how it works. You could also try flashing the bios to see if your problems get solved. I wouldn't be surprised if the boost bios is conflicting with your overclocking.

Also, check your Vrm temperatures, the memory could be running hot. On the 7950, you'll see majority of performance increases from increasing the core clock, since the 7 series is not short of memory bandwidth by any means.

Thanks for the input, i did take a different power supply and hook it up with different cords ect. and got the same results. VRM temps in the low 60's


Also it may be worth noting now that the card when overclocked just 25mhz to 950 core will not throttle, however the GPU usage will stagger significantly, usually starting at 70% usage and wondering its way up to 90, but spending the majority of its time in the mid ti low 80:s, in games such as battlefield this obviously is a FPS loss, and no im not using VSYNC or anything that would limit FPS

I have tried both BIOS on the card and they both are reacting the same way.

 

[Piotrsama]

Were you overvolting when Ocing?
That can kill your card and/or reduce it's life span.

 

[Tweak155]

Were you overvolting when Ocing?
That can kill your card and/or reduce it's life span.

I was thinking this. It is just a risk you take.

 

[Nomercy1]

Were you overvolting when Ocing?
That can kill your card and/or reduce it's life span.

The card has never been overvolted, i belive to change it on this card anyway i would have needed the modded trixx to do that, in either case no.

I took the card and put it in another PC to make sure its the card, got the same results. pic of the card struggling at stock and going below stock speeds and not boosting with normal power setting

Here it is STOCK power everything trying to play BF3

 

[UNhooked]

A bios re-flash may fix your problem.

 

[Silverforce11]

VRM temps according to both HWinfo64 and GPU-Z ahve never exceeded 63c in any of my benching or gaming.

No way. VRMs do not run that cool under max load.

vram and VRM are different.

 

[Nomercy1]

No way. VRMs do not run that cool under max load.

vram and VRM are different.

i can take pic of hwinfo if u want want but they both read exactly the same, i just ran valley benchmark to get this, but i assure you if i run it for many hours it will max at 63c

 

[Nomercy1]

A bios re-flash may fix your problem.

Thinking this as well, probably going to be my last resort, at this point something clearly wrong sent in a ticket to sapphire, waiting to hear back from them.

 

[BallaTheFeared]

I would imagine simply moving your slider + a few percent on powertune would stop all this, does it not?

 

[Silverforce11]

i can take pic of hwinfo if u want want but they both read exactly the same, i just ran valley benchmark to get this, but i assure you if i run it for many hours it will max at 63c

This doesnt look right, in every card ive owned, even ones with heatsinks on the VRM they run hot.

Someone else thinks this is normal? Such low mosfet temps is so unnusual..

 

[BallaTheFeared]

my gpuz never read a vrm temp.

This is with my TF3, pretty sure the OPs card has better memory and vrm cooling.

 

[UNhooked]

Thinking this as well, probably going to be my last resort, at this point something clearly wrong sent in a ticket to sapphire, waiting to hear back from them.

I would just do it, since it's a painless process. Just back up your old bios and flash using atiflash in dos mode.

This is how I fixed my 4870X2 couple of times

 

[Nomercy1]

Talking with sapphire now, told them my VRM temps and they say thats completly normal for this card silver if you were still wondering.\

Long story short sapphire is going to RMA the card on the grounds that it cant sustain stock clocks w/o the need for upping the power to +20%. Thanks everyone for your replies, if they let me know why this card is doing this ill certainly provide a update.


Thanks !