Brand new GTX 460, crashes at 82C. Is it the card?

[OCNewbie]

I just received my EVGA GTX 460 SSC today - Link.

Installed it without issue, played a bit of WoW without issue. I played with overclocking a bit, got the core/memory up to 900/1000 (default of 850/975) and ran some stress tests. I first ran OCCT's GPU stress test. Things seemed fine until the card got up to around 80-82C. No artifacts or anything, but the test kind of froze, but the computer didn't lock up. I was able to hit Esc, and close the program.

Anyway, I've run OCCT probably 5 or 6 more times now, and at default clock speeds, every time the card gets to right about 82C, and it appears the fan is about to spool up a bit more, my monitor goes into standby, and I have to turn off the computer, and turn it back on to get it to boot; pushing the reset button won't wake the card up.

After 4 different phone calls with EVGA tech support, disabling my CPU overclock, running HW Monitor to monitor my PSU's 12v line, disabling the High Definition Audio Controller, rolling back to an older (supposedly more stable driver), nothing changed the issue.

Does this sound like a video card problem? I ran with the same specs for at least a year, year and a half, and had no issues.

I've had this PSU - BFG-ES800 PSU since 12/2008. I'm assuming it's still in sufficient condition to power this card. I've had zero issues with my GTX 260 216/55 up to this point.

 

[3DVagabond]

It's probably not the GPU getting too hot. It's likely the PWM. Apparently, from the link below, nVidia doesn't require cards to be stress tested as hard as you are doing to be certified stable.

http://www.behardware.com/articles/809-5/roundup-14-geforce-gtx-460-1-gb-cards.html

<snip> EVGA did tell us that they had also observed this issue but said that it hadn’t come up in the games and benches recommended by NVIDIA, which is considered to be sufficient to validate the card at these clocks.

 

[OCNewbie]

This very well could be the issue. Thanks very much for the reply.

 

[timma]

it's looks like PSU problem,and check your FAN still work normal.

 

[BFG10K]

This isn’t a PSU issue. I’d underclock the GPU to nVidia reference clocks and see if it works.

Also that quote about using recommended games and benches is absolutely ridiculous. If that’s how factory overclocked cards are validated then it’s no wonder people are reporting issues that are fixed when running at nVidia reference clocks.

 

[MustangSVT]

although I don't own that card, could it be issue with heat sink in bad contact?

But if you remove the heatsink , it "voids the warranty". right?

 

[OCNewbie]

I’d underclock the GPU to nVidia reference clocks and see if it works.

I think it's likely an overheating problem, and probably not of the core. Maybe the PWM, or mosfets, or? (I really don't even know what all the parts that heat up are called).

I lowered the clocks from default SSC 850/975, to 750/900. I let this run for a little over 5 minutes with OCCT, but at those clocks, it didn't seem to be able to get hot enough. It seemed to settle right around 79/80. I then raised the core to 800, which is still 50 MHz under default, and left memory at 900. After about 6-7 minutes, the card was around 81, and just flirting with 82. It crashed, same thing as before... monitor to standby, etc. Turned it off, back on, and came here.

I've already requested an RMA, but it sure sucks to have to do it on day 1. I'm just concerned that if I get another SSC that I could very well have the same issue, as I'm definitely going to stress that one in the same manner to see if it passes, which I believe it should. Would be cool if they gave me a GTX 560, or even a 560ti. I believe the 560 uses less power, and runs cooler too possibly. I know the cooler on the 560ti is very good, but of course, that would be an upgrade, and something I'm not really entitled to.

 

[VirtualLarry]

This isn’t a PSU issue. I’d underclock the GPU to nVidia reference clocks and see if it works.

Also that quote about using recommended games and benches is absolutely ridiculous. If that’s how factory overclocked cards are validated then it’s no wonder people are reporting issues that are fixed when running at nVidia reference clocks.

remind me never to purchase a factory-overclocked reference-design NV card for F@H.

 

[yottabit]

I've had stock clock cards artifact

Some of it depends on your case airflow too. I know my 5770 was artifacting because my case has poor airflow and although the core temps were fine the memory on the backside of the card was getting very hot. You may need a better temp monitoring utility- you want to see Core temp as well as mem temps for the front and back side. It could be the PWM's as someone pointed out, but in my expereince an 80+ deg C core temp usually corresponds with a 95-100 deg C plus mem temperature, which can lead to issues.

If you find the hot spot then you can just buy some little zalman thermal glue heatsinks and that should take care of the problem honestly

EDIT: Hrm, didn't see your card had a backplate.

 

[OCNewbie]

Click on the link in the 2nd post. It should detail why there is an issue with this particular card. Basically, the card is using a reference design, but because of the hefty overclock, the card is consuming up to 44&#37; more power than a reference clocked card. The components on the reference board are just not cut out to handle those kind of huge overages, so they overheat when stressed. I may consider trying to position a fan that is aiming at the back of the card somehow, though I don't believe that should be necessary, and wouldn't be necessary if the cooler and board were designed to handle the heat produced by the massive OC.

BTW, I'm not having ANY artifacting, just a sudden shut down of the card. The airflow in my case, while probably not optimal, is not horrible either. I had been running a power hungry GTX 260 for the better part of 2 years in the same case, with the same fans, without issue prior to this.

 

[edplayer]

is it crashing under OCCT at stock clocks? When you talked to EVGA support, did they say it was OK to return the card under warranty?

I have the same card clocked to 900MHz and have no problems in games. Never tried OCCT though

 

[Diogenes2]

82 sounds high to me ..

My MSI 460 768gb only hits about 60c at 850/1025 99&#37; load ..

 

[OCNewbie]

is it crashing under OCCT at stock clocks? When you talked to EVGA support, did they say it was OK to return the card under warranty?

Yes, stock clocks for the SSC+ (850/975). It also crashes underclocked at 800/900. Yes, I have my RMA confirmation, etc. Just have to pack it up and ship it out, which I will do tomorrow.

 

[edplayer]

Hope you get a good card back. What are your ambient temps and what case (and cooling) are you using?

It's probably not the GPU getting too hot. It's likely the PWM. Apparently, from the link below, nVidia doesn't require cards to be stress tested as hard as you are doing to be certified stable.

http://www.behardware.com/articles/809-5/roundup-14-geforce-gtx-460-1-gb-cards.html

Thanks. Never seen that. I backed my card down to stock clocks just in case.

 

[Cerb]

My initial was is the same as what 3DVagabond posted.

82C seems fairly hot. If you can expect the RMA to be fully processed before your return window ends, go for it, and keep the return option as a fallback. Otherwise, I'd go ahead and return for replacement, getting a non-reference MSI or GB, known for the power circuitry running at moderate temps.

 

[OCNewbie]

Finally got my RMA replacement card today.... Guess what?! CRASSSSHHHHH!!! Just like before...

And before anybody says anything about my system being the issue: I tested with a PNY GTX 460 and ran Furmark without any issues (other than the card getting toasty =P)

 

[96Firebird]

And before anybody says anything about my system being the issue: I tested with a PNY GTX 460 and ran Furmark without any issues (other than the card getting toasty =P)

What are the clocks on the PNY? Overclocked cards will use more power, which means it could very well be your PSU.

 

[OCNewbie]

What are the clocks on the PNY? Overclocked cards will use more power, which means it could very well be your PSU.

Default was 675/900 but I had it running OC'd at 800/900 (or 975, not sure tbh). The first EVGA GTX 460 SSC that I bought retail from Newegg (haven't tested this RMA unit yet) crashed at the same clocks (underclocked compared to factory clocks).

Here are a couple reviews for my PSU - Hardware Canucks HardOCP.

Supposedly jonnyGURU helped design this PSU.

I'd like to think this 800w PSU, even though it's a little over 2 1/2 years old (bought in 12/08), is capable of powering a single GTX 460 + typical extra components (1 HDD, 1 SSD, 1 DVD drive, 2 sticks of RAM, Intel Q9550).

 

[Termie]

Default was 675/900 but I had it running OC'd at 800/900 (or 975, not sure tbh). The first EVGA GTX 460 SSC that I bought retail from Newegg (haven't tested this RMA unit yet) crashed at the same clocks (underclocked compared to factory clocks).

Here are a couple reviews for my PSU - Hardware Canucks HardOCP.

Supposedly jonnyGURU helped design this PSU.

I'd like to think this 800w PSU, even though it's a little over 2 1/2 years old (bought in 12/08), is capable of powering a single GTX 460 + typical extra components (1 HDD, 1 SSD, 1 DVD drive, 2 sticks of RAM, Intel Q9550).

It's not an issue of PSU quality - it's an issue of whether it's functioning correctly. It should be able to support that hardware. That's doesn't mean it can.

But honestly, this looks like a temp issue. Those EVGA cards are clocked to the max, and the fact that the FTW (850/1000) was replaced with the SSC+ (850/975) suggests that EVGA knew it wasn't a sustainable design (even though it was heavily marketed and touted right here on Anandtech).

82C is very, very hot for a GTX460. What kind of case cooling do you have? And how hot does it get in gaming? I'm willing to bet it won't touch 82C in games, and if that's true, you'll be all set...I certainly wouldn't bother RMAing it again. And if it still overheats, just take the OC and voltage down a notch using EVGA Precision. A lot of the GTX460s weren't able to hit 850, so although this one is pre-overclocked to that point, that doesn't mean it was an ideal speed for a GTX460.

 

[BD231]

I'd definitely sit happy at 850, that's a lot of power and about all you can get out of the 460 safely IMO. 82c is to high though, I can't stay stable at 850 past 7Xc range for some reason.

I'm using aftermarket cooling though so it's not an issue, however, stock cooling just didn't cut it. If you can get the 850mhz card replaced I definitely give another one a shot, very good card if you can get it working.

 

[happy medium]

I think you fan profile is not working right. I see no mention of fan speed in this thread.
Do you have the fan speed set at 100&#37; when running this OCCT program?
Have you tried to run any games with it? Does you fan spin up in line with temps?
DId you try a different fan profile in MSI afterburner?
What case are you running it in, and what kind of fans/airflow does it have?

 

[OCNewbie]

EVGA made this card to run at 850/975. I will not underclock it to get it to run. If a company designs and sells a retail card at a specific core/clock frequency, there should not be a limit of what programs you can run at those speeds. You should be able to stress that card to 100&#37; of it's ability, at it's default (set by the manufacturer, in this case EVGA) speeds, without crashing. 80-82C is significantly below the temp threshold for these cards. When idling, this card is around 34-36C.

 

[OCNewbie]

I think you fan profile is not working right. I see no mention of fan speed in this thread.
Do you have the fan speed set at 100% when running this OCCT program?
Have you tried to run any games with it? Does you fan spin up in line with temps?
DId you try a different fan profile in MSI afterburner?
What case are you running it in, and what kind of fans/airflow does it have?

No, I haven't tried a different fan profile. I don't understand why I should need to though? Don't these cards know how to cool themselves? Again, this card is running with the default specs designed by EVGA. My old EVGA GTX 260, in this same case, same PSU, same fans, etc., got up to about 85C in Furmark, and settled there. No crashes, nothing like that.

BTW, I haven't really tested it with this card yet, but the card I got from Newegg ran about 60C max running Borderlands.

 

[happy medium]

When idling, this card is around 34-36C

Yea , you should see post #21, you fan profile is not aggressive enough.
See my card in my signiture................

When running with a 20&#37; factory overclock and OCCT, I suggest you make a better fan profile. DO you have Crysis or any other stressfull game to try?

 

[happy medium]

No, I haven't tried a different fan profile. I don't understand why I should need to though? Don't these cards know how to cool themselves? Again, this card is running with the default specs designed by EVGA. My old EVGA GTX 260, in this same case, same PSU, same fans, etc., got up to about 85C in Furmark, and settled there. No crashes, nothing like that.

BTW, I haven't really tested it with this card yet, but the card I got from Newegg ran about 60C max running Borderlands.

You missed quite a few important questions in post #21.
Fan speed ramp up, case, airflow, games?