Is my GTS 250 dying?

[zoran555]

I just got a second hand GTS250 1gb with Zalman VF1050.

It's working fine in Win7 x64 on the desktop and in a few games I tried (Assassin's Creed 2, Dead space).
Found some time to test it out with GPUtool and OCCT.
What happens is I get errors reported on default clocks as soon as temps go over 70-74C.
I get visible artifacts splotches in the test window, which correspond to the errors reported.
OCCT:gpu with error checking starts reporting errors as soon as it hits 60C on the GPU.

Furmark showed and reported no errors up to 88C (the temps stabilized there).

I also run OCCT gpu:memtest and it never showed any errors no matter how hot the gpu was.

All these temps are well in the normal working range for the G92 chip (IIRC the nVidia reference card runs at 105C.

In other HW/OC forums I got told
a) that the memory is dying
b) that all is well and I should stop running needless tests.
I got reports from other GTS250 users that they also get errors in OCCTgpu/GpuTool while their card is running normally in games and such.

Can someone with a GTS 250 please run some tests to confirm, please?

Tested the card in another system (p45/e6300), same results.
Can any one pinch in their thoughts and preferably some hard data as I'm stumped with this all.


My specs:
Win7 x64 ultimate
e6550
Gigabyte p35-ds3r
4x2gb ddr 1066
Gigabyte gv n250oc 1gb
Antec p182 (4x12cm)
Enermax PSU 12v@33A single rail

 

[trickson]

Sounds like it is . When the vid card starts to get them artifacts and starts to over heat that is a tell tail sign things are going to go bad for that card . And if you tested the card in a different setup and it is doing this then YES that card is on it's way out . JMO . I could be wrong .

 

[erocker]

Two different systems, yes. The card is on its way out.

 

[zoran555]

Sounds like it is . When the vid card starts to get them artifacts and starts to over heat that is a tell tail sign things are going to go bad for that card . And if you tested the card in a different setup and it is doing this then YES that card is on it's way out . JMO . I could be wrong .

Well it's not overheating, if you've read my post. If anything it's cooler than normal thanks to good cooling.

The other thing is that outside OCCT:gpu and GPUTool there are no instabilities or artifacts of any kind and I have another user run the tests with he same results.
Runs games, Furmark, 3Dmark vantage and 3dmark 06 without a hitch.

I have an inkling that the tests might be bugged and would very much like to have it confirmed by other users with the same card, it only takes a minute or two.

 

[erocker]

Well it's not overheating, if you've read my post. If anything it's cooler than normal thanks to good cooling.

The other thing is that outside OCCT:gpu and GPUTool there are no instabilities or artifacts of any kind and I have another user run the tests with he same results.
Runs games, Furmark, 3Dmark vantage and 3dmark 06 without a hitch.

I have an inkling that the tests might be bugged and would very much like to have it confirmed by other users with the same card, it only takes a minute or two.

Ah, only with OCCT and GPUTool. Well it's a sign that it may start happening with your games and other benchmarks at some point. The tests aren't bugged as thousands of end-users use these tests without problems. I am not one of those users as I'm against torturing my hardware with OCCT. Anyways, use your card and don't bother with those tests anymore. Once the problems start happening with the applications you use then you may want to consider replacement. That could be months or years down the road.

 

[zoran555]

Ah, only with OCCT and GPUTool. Well it's a sign that it may start happening with your games and other benchmarks at some point. The tests aren't bugged as thousands of end-users use these tests without problems. I am not one of those users as I'm against torturing my hardware with OCCT. Anyways, use your card and don't bother with those tests anymore. Once the problems start happening with the applications you use then you may want to consider replacement. That could be months or years down the road.

Well, your reply makes sense, but consider this:
a) there were bugs like this in OCCT:gpu, I've read up their forums.
b) all the people experiencing these kind of errors had instabilities already (bsods, graphic drivers crashes and such)
c) the errors point to the gfx memory failure, yet OCCT:gpu memtest never found an error

 

[erocker]

Well, your reply makes sense, but consider this:
a) there were bugs like this in OCCT:gpu, I've read up their forums.
b) all the people experiencing these kind of errors had instabilities already (bsods, graphic drivers crashes and such)
c) the errors point to the gfx memory failure, yet OCCT:gpu memtest never found an error

I understand, which is why I said not to run them. If I was in your shoes, I'd try to get the card RMA'd or I would just use it and wait until it can no longer work with the applications I use. But that's just me. I've owned a couple GTS250's.. well a GTS250 and a 9800GTX+ and neither had problems while runing any sort of stability test. I suppose the things you can try is:

Reinstall the drivers
Contact the makers of OCCT
Replace the card or..
Just don't run these stressful programs on it.

 

[trickson]

All I know is when my GTX260 started this is was already too late . you should be looking into a new card and just scraping that one .

 

[FreedomEclipse]

try putting some new thermal grease on the card & see if that helps - it might not seem like the most likely solution one would usually suggest. but Ive had 2 cards in that past that started artifacting n shit when it got hot & not because of of the overlock i put on them. had a look at the TIM under the cooler n it was drier then oprahs vagina under there. after i put a small application of MX-2 both the cards worked flawlessly & are still working without any issues

 

[Benetanegia]

I just tried my 9800 GTX+ on the latest builds and got no errors. Only 2 mins each, but I think it's more than enough to test the programs.

I wish you luck with your card, it might not happen anything before you upgrade, if you have no problems on games. One thing that I test when a new game is released is how much % loaded are my cards in games and the higher I have seen on the 9800 GTX+ is 80% (on Crysis, Metro2033). That's almost 20% difference to the load that is achieved in burn tests, so your card could work forever at 80% load. There's not going to exist a game (or game seting) that puts more load on that card AND offer playable framerates at the same time, so like I said, I wish you are lucky.

 

[zoran555]

try putting some new thermal grease on the card & see if that helps

Seems reasonable I was planning to do that, but heck, I'm lazy.

I just tried my 9800 GTX+ on the latest builds and got no errors. Only 2 mins each, but I think it's more than enough to test the programs.

Hmm, did you enable error checking in OCCT:gpu?

All I know is when my GTX260 started this is was already too late . you should be looking into a new card and just scraping that one .

Ah, it's still under warranty for a few months, but the store is out of country. Though I wonder how they'll react: Look the card runs everything you throw at it perfectly, except 2 freeware stress apps. Gimme a new one ASAP!

 

[trickson]

Then just stop using them harsh stress tests the card can not handle it . Do you have any games like crysis or just cause 2 ? try running a game every thing maxed out and see if it can still take the heat .

 

[zoran555]

Then just stop using them harsh stress tests the card can not handle it . Do you have any games like crysis or just cause 2 ? try running a game every thing maxed out and see if it can still take the heat .

Just played half an hour of Crysis:Warhead. Everything seemed fine, if a tad smoother than my last card (9800gt 512mb). Put most of the things on the highest level and the temps on the GPU kept around 65C.

 

[trickson]

Just played half an hour of Crysis:Warhead. Everything seemed fine, if a tad smoother than my last card (9800gt 512mb). Put most of the things on the highest level and the temps on the GPU kept around 65C.

Then I would use it that way only . stop torching and killing your parts with them DUMB ASS programs !

 

[FreedomEclipse]

Just played half an hour of Crysis:Warhead. Everything seemed fine, if a tad smoother than my last card (9800gt 512mb). Put most of the things on the highest level and the temps on the GPU kept around 65C.

trust me - just re-apply some new thermal paste. is that fan on the Zalman cooler plugged directly into the card?? cuz you might need to turn it up a little (if it isnt) - Zalmans coolers usually come with the sort of plugs that can go directly into a mobo (the standard 3pin fan plugs) & they come with this fan controller unless whoever you recieved the card from didnt bother passing that on with the card.

 

[zoran555]

trust me - just re-apply some new thermal paste.

Well, I did reapply the paste, the original one was some dry sh*t, it took a while to remove.

Didn't change much, maybe a tad lower temps, that's all.

But I did find out the cause: clocks....I lowered the clocks until I got rid of errors.
What's funny, the clocks which work perfectly are: 741/1782/1100. They kinda remind me of the DEFAULT GTS250 clocks.
This is supposed to be a factory OC-ed card (765/1912/1100), which is borderline stable at it's factory frequencies

The other guy who ran the test and got errors has the same identical card at the same frequencies.

Hmmm... Gigabyte pushing things a tad too much?


At least I got my faith back in the tools.