My GTX 295 is having problems

So, I bought a GTX 295 a while back thinking I'd use it to get some higher scores when running 3D benchmarks.

But, the card has never been stable at stock speeds (576MHz core/1008MHz memory clock), even when I first bought it. Always crashes within 30 seconds to 2 minutes of starting a 3D benchmark or beginning a game. There are no visual artifacts, tearing or otherwise, it will just crash to a black screen. Then, I have to fully power down the system and power it back up to get the system to run again, as a soft-reset does nothing. I've run this card in probably five different systems now with the same behavior in each.

After doing some research on this, it seems it is a COMMON problem with nearly all GTX 295 cards (the single PCB version at least, which is what I have) as I am finding a lot of cases of people reporting these same problems.

https://forums.geforce.com/default/...-200-100-series/gtx-295-black-screen-crashes/

I had to underclock my core clock down to 547MHz just to keep it from crashing. Then, once I do that it is completely normal, though the performance is no doubt lower than it would be at stock.

I did a test and ran 3DMark 03 for comparison sake. At 547MHz core and the stock 1008MHz memory clock the GTX 295 scored 20,500-20,600 3DMarks. One of my GTX 260 192-CUDA core cards scores consistently at 25,000 3DMarks or greater, while at the same clocks as the GTX 295 (I lowered them intentionally for comparison sake). Both cards were benchmarked in the same system using the same drivers.

How can it perform so badly when it is based on the same technology as the GTX 260, and is a higher end card?

Did you enable SLI in the nvidia control panel? If not, that's the issue. I don't think it defaults to enabled even with dual GPU cards. Not positive on this.
If you did, it's something more complicated.

The crashing sounds like OCP is being tripped, time for a voltmod
No-warning crash to black or grey that requires a power-off to get rid of is classic OCP/OVP/UVP type stuff.


And yes, it is terrible. It always has been, just like the GTX480 that replaced it. Much like the 480, it was very nice, if power hungry and delicate, in its time.

Bobnova said:

Did you enable SLI in the nvidia control panel? If not, that's the issue. I don't think it defaults to enabled even with dual GPU cards. Not positive on this.
If you did, it's something more complicated.

The crashing sounds like OCP is being tripped, time for a voltmod
No-warning crash to black or grey that requires a power-off to get rid of is classic OCP/OVP/UVP type stuff.


And yes, it is terrible. It always has been, just like the GTX480 that replaced it. Much like the 480, it was very nice, if power hungry and delicate, in its time.

Click to expand...

Yes, I tried it both with SLI enabled and with SLI disabled. My particular card defaults to enabling SLI as soon as the driver is installed.

The result is the same either way.

Also ran it with probably four different power supplies. Corsair HX750 750W, Corsair HX520 520W, Corsair TX750 750W, and PC Power & Cooling Silencer MK III 850W. Same result in each case.

Does GPUz report SLI being enabled?

I would try HWBot's Heaven benchmark also, the DX9 version. It scales very nicely with SLI and isn't as CPU bound as 3D0anything. 3D03 is an odd one at times. Not as odd as some, but odd.

you check where temps are prior to crashing ?

Is it a sandwitch or a singel ?

Bobnova said:

Does GPUz report SLI being enabled?

I would try HWBot's Heaven benchmark also, the DX9 version. It scales very nicely with SLI and isn't as CPU bound as 3D0anything. 3D03 is an odd one at times. Not as odd as some, but odd.

Click to expand...

Yes, it shows SLI is enabled.

Witchdoctor said:

you check where temps are prior to crashing ?

Is it a sandwich or a single ?

Click to expand...

Not a clue. I have no way to monitor temperatures while running benchmarks on it (I only have it hooked up to one monitor). But, when running lower clocks and not crashing, I don't get past 58°C according to Hardware Monitor and EVGA Precision IIRC (need to check that though).

Single PCB.

This is my card, the EVGA GTX 295 CO-OP Edition.

Idle temperatures:



Load temperatures:

Temps seem ok, shame these dual cards have a way of burning themselves out.

My 295 is finiky, but can get it to work most of the time

Just so you have tried everything physically I would take it all aprat, clean it up spic and span throw on some new TIM and give it a whirl, if going out on OCP at idol there is something seriously wrong with it, before flashing it you may try some different drivers to take a driver issue out off the table

last you could try to bake it

Witchdoctor said:

Temps seem ok, shame these dual cards have a way of burning themselves out.

My 295 is finiky, but can get it to work most of the time

Just so you have tried everything physically I would take it all aprat, clean it up spic and span throw on some new TIM and give it a whirl, if going out on OCP at idol there is something seriously wrong with it, before flashing it you may try some different drivers to take a driver issue out off the table

last you could try to bake it

Click to expand...

I've taken it completely apart and replaced the paste twice already. It hasn't helped with the crashing issue, but my temps did get lower.

Tried at least five different driver versions.

Works great at idle, just crashes under heavy 3D loads.

I'm really hoping to avoid having to reflow it, though I no doubt have the knowledge to know how to do so (I've reflowed several video cards and motherboards).

WD to the rescue.
Hello my old friend.

Hey Scott,

Good to see a face I know around here, hope all is well on this georgouse day in the Northeast, stuck tiling my backsplash today.


Try 185.85

It will rock 05 on that card

Witchdoctor said:

Try 185.85

It will rock 05 on that card

Click to expand...

Well, that sort of helped.

It made it through a pass on 3DMark03 (scored 23k), but it quickly failed in 3DMark06.

The oldest I'd tried previously was 280.26.

Tech Tweaker said:

Well, that sort of helped.

It made it through a pass on 3DMark03 (scored 23k), but it quickly failed in 3DMark06.

The oldest I'd tried previously was 280.26.

Click to expand...

Old card needs old drivers.
197.45 for 06.

Like the guys above said, older is better for older stuff. Usually.

If I recall correctly, the GTX2xx cards want drivers somewhere in the >170 to <200 range, GTX4xx and 5xx need 2xx drivers or newer and tend to prefer 2xx, and GTX6xx/7xx need use 3xx drivers.

The benching for top scores reality is somewhat more complicated, but the above will at least get you into the ballpark.

Using drivers that are "too new" can cause issues. Instability for 2xx cards on new drivers for instance. There was also a 3xx driver edition that tended to kill GTX580s, that was charming.


My brain hurts now.

Bobnova said:

Like the guys above said, older is better for older stuff. Usually.

If I recall correctly, the GTX2xx cards want drivers somewhere in the >170 to <200 range, GTX4xx and 5xx need 2xx drivers or newer and tend to prefer 2xx, and GTX6xx/7xx need use 3xx drivers.

The benching for top scores reality is somewhat more complicated, but the above will at least get you into the ballpark.

Using drivers that are "too new" can cause issues. Instability for 2xx cards on new drivers for instance. There was also a 3xx driver edition that tended to kill GTX580s, that was charming.


My brain hurts now.

Click to expand...

You recall correctly.
I have a notebook so my brain doesn't hurt as much as it used to.

Bobnova said:

Using drivers that are "too new" can cause issues. Instability for 2xx cards on new drivers for instance. There was also a 3xx driver edition that tended to kill GTX580s, that was charming.

Click to expand...

I vaguely remember reading something about this.

I think it was this thread. http://www.overclockers.com/forums/showthread.php?t=733131

I never tried the 320.18 driver myself, but I had some serious issues with the 320.49 driver. The 320.49 caused my system to crash, constantly, I was probably crashing and blue screening 4-6 times a day, and I wasn't even gaming or running any benchmarks. It would just decide to crash while sitting at the desktop, browsing the web, or checking email.

I wonder, if this is a driver issue then why has it only effected my GTX 295, and not my GTX 260. They are based on the same architecture, the only difference being the 295 is a 55nm process, and the 260 is 65nm, and the 295 having two cores.

The GTX 260 has been stable with every driver I've run from 280.26 to the current 335.23.

Also, I finally managed to install 197.45, took three attempts. Didn't want to install for some reason, seemed like Windows was trying to update to a newer driver at the same time I was trying to install 197.45. After that I was able to run 3DMark06 without crashing (for one run at the very least).

It's like watching a slideshow on 3DM06, 8-15FPS most of the time. Don't remember my GTX 260 performing this badly.

To be fair, 3d06 likes CPU. A lot. But only uses one core outside the CPU test.

Try AM with that driver

Could be simply the card is failing

Witchdoctor said:

Try AM with that driver

Could be simply the card is failing

Click to expand...

It crashes in Aquamark3 around the time it gets to the third or fourth test.