7600GT failed torture test....

S

shadowwyvern

Gawd
Joined
Sep 22, 2005
Messages
604
I just got a new 7600GT from evga. It came clocked at 580Mhz, but when I ran 3dmark05 for several hours (I went to work, so i don't know how liong it was before it crashed), but i came back and my screen was black and white and all messed up (lots of short flickering lines), with a message about some nvidia driver crashing. I rebooted, and now have the core clocked at 570Mhz. What should be my next step? Tonight I am going to test at these settings, but an not happy about the card not being stable at the clock it came at.
 
This is just me but...I would never loop a synthetic benchmark for hours simply to test stability. My stability test consists of loading up my favorite games and playing them for my usual amount of time and if there aren't any issues it's all good. Having said that, I rma'd my 7900GT KO not only because of artifacting in 3DMark 06 but also BF2 and a sudden reboot in NFSMW. If it was just 06 I wouldn't have bothered..
 
Regardless, if the cards are not supposed to ever crash, no matter how hard you run them (which I think is the case provided you have enough cooling), I do nto want mine crashing after only 5-6 hours of heavy use. I am currently running at 580Mhz to see if temperature is an issue here.
 
I agree with you shadow, this is not acceptable. I suggest you rma the card as it should run stable at it's out of box frequencies. This is reason enough to rma. Time to put that EVGA customer service to the test.
 
I agree with jomamma; if it doesn't give you any problems while you're actually **using** the card, why worry about some synthetic benchmark crashing?

I think nitpicking things like that will cause companies to end any factory overclock and maybe tighten up their RMA standards since it could be seen as abuse from some obscured angle.
 
rfwinters77 said:
I agree with jomamma; if it doesn't give you any problems while you're actually **using** the card, why worry about some synthetic benchmark crashing?

I think nitpicking things like that will cause companies to end any factory overclock and maybe tighten up their RMA standards since it could be seen as abuse from some obscured angle.
Click to expand...
what difference does it make if its synthetic or not? it shouldnt screw up after a few hours.
 
Overnight I ran the video stress test avaible here and the HDR demo with the flying balls. It did not crash, or develop any huge artifact (I did not even llok closely), over 538 min. Temp as good too, around 60C. Is it possible that the issue I saw with 3Dmark was just a crash due to overheating? It was during the day, which was around 80F here, and my machine is crunched in with 2 others. Also, as I found out last night, I did not properly llop 3dmark. I told it to repeat each test 99 times (the max) thinking it would go around its series 99 times, but it actually runs reach test 99 tiems bfore moving on to the next one. And when it does this with Battle of Proxycon (first test), the framerate for the first cycle starts at around 18, but a few cycles later, that same section of the test is rendered at only 9. Other parts of the test apeared to suffer from this as well.
 
UT2004 crashed a few times on me, giving me a nVidia driver error much like yours, but I simply rolled back my drivers and it has yet to happen again.
 
I am going to repeat the test this evening with a big fan blowing over my case. I very well may RMA if it keeps crashing. Does anyone know evga's policy regarding shipping I do not have a spare card). Also, how do you make 3Dmark loop the series of tests, instead of just one test? The option seems to be there, but it is always unavaliable.
 
cyks said:
UT2004 crashed a few times on me, giving me a nVidia driver error much like yours, but I simply rolled back my drivers and it has yet to happen again.
Click to expand...


yep could easily be a driver issue or a bug within 3Dmark, I definitely would not blame
the card just yet. ;)
 
joemama said:
This is just me but...I would never loop a synthetic benchmark for hours simply to test stability. My stability test consists of loading up my favorite games and playing them for my usual amount of time and if there aren't any issues it's all good. Having said that, I rma'd my 7900GT KO not only because of artifacting in 3DMark 06 but also BF2 and a sudden reboot in NFSMW. If it was just 06 I wouldn't have bothered..
Click to expand...

Shhhh....3DMark is a game... :)
 
This is very common of the 7900's to fail on benchmarking that they can't run their oc'ed speeds... which is why from now on I'm just going to buy the refrence cards... I've had enough with these promised high locked in speeds that can't deliver.
 
shadowwyvern said:
Regardless, if the cards are not supposed to ever crash, no matter how hard you run them (which I think is the case provided you have enough cooling), I do nto want mine crashing after only 5-6 hours of heavy use. I am currently running at 580Mhz to see if temperature is an issue here.
Click to expand...


EXACTLY.

'Maybe it's just me', but I would never consider my normal games to be enough of a 'torture test' to root out problems with a card I've just purchased. I would find the most graphically intensive thing I could and run it for hours or even days especially if I had a suspect card like a 7900GT. That happens to be 3DMark 06. The whole synthetic benchmark thing aside (don't even get that whole debate started) it's a damn good GPU stress test, especially the SM3.0 tests. Only thing I've found that stresses the card equally sofar is Oblivion.

Sorry if I sound like an ass, but I'm so sick of people saying 'screw running those synthetic benchmarks, play your games, benchmarks suck, blah blah blah blah!'
 
merlin704 said:
Shhhh....3DMark is a game... :)
Click to expand...
It really is to some people...I suspect some of the highest scorers rarely ever play the "other" types of games. :p
 
Why play a game when you can get the same thrill from real life activities, like overclocking? Regardless, I am going to clock the card down to stock speeds so I can absolutly sure of stablility. Does anyone know how to make the underclock permanent (ie make it the new "stock"), as I occasionally use other OSs, like sometimes vista, and maybe linux, and want the thing clocked at stock speeds there too, when I cannot be sure drivers will do it as nicely as in XP.
 
shadowwyvern said:
Why play a game when you can get the same thrill from real life activities, like overclocking? Regardless, I am going to clock the card down to stock speeds so I can absolutly sure of stablility. Does anyone know how to make the underclock permanent (ie make it the new "stock"), as I occasionally use other OSs, like sometimes vista, and maybe linux, and want the thing clocked at stock speeds there too, when I cannot be sure drivers will do it as nicely as in XP.
Click to expand...

I respect you decision to do that but keep in mind you paid a premium to buy a factory overclocked card. So basically you've been had (no offense) so instead of buying a reference card you bought a factory OCd card for maybe $20-30 more. Only to have to downclock it for stability. The fact is that I agree with your initial assessment. No videocard should crash anything synthetic or not when running at it's advertised speeds. It's not like you were overclocking it further.

Furthermore benchmarks may be more demanding than real world gameplay and even give a rough idea of things to come. How comfortable would you feel playing unreal tournament 2007 and quake wars knowing 3dmark 06 craps out.

I feel evga should either get you a stable factory overclocked card or pay you the difference this card costed you over a stock card. Or give you some insentive to keep it "underclocked" at factory settings.
 
You must log in or register to reply here.
Back
Top