Advanced search

Message boards : Graphics cards (GPUs) : Anybody know the max GPU temp?

Author Message
Profile Deleted
Avatar
Send message
Joined: 7 Sep 08
Posts: 6
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 2174 - Posted: 7 Sep 2008 | 21:08:04 UTC
Last modified: 7 Sep 2008 | 21:09:24 UTC

I just got into this CUDA thing, and it raises the temp of my 88900GTS 512 to 51 deg C. I never let my CPU go over 45, and right now it's 12 (because of thermoelectric water cooling).

How hot is too hot for a GPU?

thanx!

____________
A cerebral mutant, damned by God and doomed by man to understand EVERYTHING and control NOTHING.

My blog

Profile Stefan Ledwina
Avatar
Send message
Joined: 16 Jul 07
Posts: 464
Credit: 135,911,881
RAC: 123
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 2177 - Posted: 7 Sep 2008 | 21:21:14 UTC

Hi and welcome to PS3GRID/GPUGRID!

I can't tell you exact numbers, but 51°C are for sure no problem for a GPU.
From what I've read in various magazines critical temps for a GPU are about 100°C...
____________

pixelicious.at - my little photoblog

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 2180 - Posted: 7 Sep 2008 | 21:57:01 UTC - in response to Message 2177.

I can't tell you exact numbers


There's no point in exact numbers for customers, they only exist because the manufacturer has to set definite limits. Let me answer his question first before I go on: 51°C is absolutely fine! Is that on water or with an Accelero / Thermalright?

Generally the limit for shutdown is around 100°C. However, that doesn't mean 90°C would be fine, even if same GPUs do get that hot with stock cooling. The rule of thumb is still that 10 degrees more halves the life time. GPUs normally run in the 70 - 90°C range (because the stock coolers are just too small) .. they are fine with that because they don't run this load 24/7. I'm not sure what kind of duty cycle the manufacturer expects for a gamer card. 8 h a day 365 days a year at 90°C? That would be quite a hardcore gamer, but possible. The majority of customers will taxe their GPUs less than that though, so if it survives the harsh conditions for 2 years (insert whatever warrenty time) that would be perfectly fine and yield few returns due to "burnout" or degradation.

Is it somehow possible to get my point? I'm actually a bit too tired right now ;) To summarize: it's not the maximum allowed temperature we should worry about, but rather the "time to failure" due to normal operation / crunching. And here we can efinitely say "50°C is fine" and "80°C should be too much". I don't like 70 due to personal experience (yes, quite a limited set for statistics) and am not sure about 60°C. I tend to give it the thumb up.

And please don't tell me "it doens't matter if the cards life span is reduced from 10 to 5 years". That's not what we're talking about here, because graphics cards are not designed for 24/7 load. When I ran folding@home1 on my 1950Pro the card started to give me errors after 4 months of 24/7 operation at 70 - 75°C. Its fan was programmed to let the card reach 90°C, I changed it using ATI Tray Tool. These 4 months are not the normal time to failure and the card was replaced without major problems, but this should give you an indication of how critical this business can be. Now imagine the same card at 90°C.. would have failed in 1 - 2 months.

Having said that, I'm again trying my luck with GPU crunching. But this time the GPU is cooled by an Accelero S1 and kept at 50 - 55°C.

MrS

(I know.. I should just go to bed :D )
____________
Scanning for our furry friends since Jan 2002

TomaszPawel
Send message
Joined: 18 Aug 08
Posts: 121
Credit: 59,836,411
RAC: 0
Level
Thr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 2196 - Posted: 8 Sep 2008 | 17:19:30 UTC - in response to Message 2180.

My 8800GTS 512 is about to 85-87C when load and automatic fan control is on.

So nvidia set it that, and if it fail they must give me back my money or new card.

so I am not woory :)

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 2200 - Posted: 8 Sep 2008 | 19:24:00 UTC

As long as you're within warrenty, sure.

MrS
____________
Scanning for our furry friends since Jan 2002

Thamir Ghaslan
Send message
Joined: 26 Aug 08
Posts: 55
Credit: 1,475,857
RAC: 0
Level
Ala
Scientific publications
watwatwat
Message 2202 - Posted: 8 Sep 2008 | 19:26:46 UTC - in response to Message 2174.

I just got into this CUDA thing, and it raises the temp of my 88900GTS 512 to 51 deg C. I never let my CPU go over 45, and right now it's 12 (because of thermoelectric water cooling).

How hot is too hot for a GPU?

thanx!


My GTX 280 runs at 79C with 43% stock fan speed, this percentage is automatically set which makes me believe Nvidia dont want the card go above 80C.

I just set it to 100% fan speed and the hair dryer style noise is unbearable, however, it went down to 64C within 2 minutes.

Bottom line, I'll trust the control panel to decide which threshold is acceptable for which model. So I'm back to automatic non noisy fan speed.

Profile Krunchin-Keith [USA]
Avatar
Send message
Joined: 17 May 07
Posts: 512
Credit: 111,288,061
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 2228 - Posted: 9 Sep 2008 | 22:19:26 UTC

My two XFX NVIDIA 600MHz both run at about 70% fan speed and stay around 65C. They run whipser quiet. I did try to crank one up to 100% fan, I coul hear it a little bit.

At home XFX NVIDIA 640MHz (Factory Overclocked version) of same board, runs the fan at 39% and maintains a temp of 63C.

All are on auto control.

I like the fact the fan speed is so low on the factory overclocked model, it gives it room to work with.

DO NOT TRY THIS AT HOME.

Turn off you air conditioner while showering so cold air does not blow on you, and forget to turn it back on.

I came home yesterday and it was about 90F in the house, thermostat does not register above that. Computer #2 was wailing like a stuck pig. Its only a P4-HT and running only 1 thread for boinc. I've done this before at full power, it can handle it but the fan screeches loudly. I do not have any temp monitoring on that computer, so do not know what CPU temp was. It was above 99F in my room. Computer #1 with the NVIDIA board was quieter. It has good cooling too, but the case is open. The NVIDIA was now making some noise, no wailing like the other, but I could hear it, normally its whisper quiet. The fan speed was up to 42% with temp of 79C. It ran that way about 8 hours. Not bad I'd say. Not to worry either as the XFX NVIDIA boards have a double lifetime warranty.

That means (pick one)
A) When I die, They resurrect me for another go round.
B) When I die, I get to take it with me to the afterlife.
C) The warranty transfers to the next owner, if I sell it or it is passed along in my estate.

Christian Buchner
Send message
Joined: 27 Aug 08
Posts: 4
Credit: 27,844
RAC: 0
Level

Scientific publications
wat
Message 2232 - Posted: 10 Sep 2008 | 17:01:13 UTC - in response to Message 2228.
Last modified: 10 Sep 2008 | 17:04:01 UTC

Hmm, four nVidia G92 GPUs ran at up to 90 degrees on a single MSI K9A2 Platinum V2 mainboard. Then I added an extra case fan blowing directly onto the boards and now it's betwen 65 and 70 degrees under full load (Folding@Home actually).

I also operate one passively cooled nVidia 8500 GT which is in the bottom most slot of one PC (poor airflow) - and it reaches 102 degrees celsius. Thermal throttling would kick in at 105 degrees. The GPU computes fine, but I suspect it would die soon at this temperature, so I am considering an extra case fan here as well.

Thermal throttling usually kicks in at 105 degrees. This is about the limit where a GPU does not take immediate damage, but would cease to function when heated even further. So it should not be operated in that range for extended periods of time.

Thamir Ghaslan
Send message
Joined: 26 Aug 08
Posts: 55
Credit: 1,475,857
RAC: 0
Level
Ala
Scientific publications
watwatwat
Message 2234 - Posted: 10 Sep 2008 | 18:47:01 UTC - in response to Message 2232.

Hmm, four nVidia G92 GPUs ran at up to 90 degrees on a single MSI K9A2 Platinum V2 mainboard. Then I added an extra case fan blowing directly onto the boards and now it's betwen 65 and 70 degrees under full load (Folding@Home actually).

I also operate one passively cooled nVidia 8500 GT which is in the bottom most slot of one PC (poor airflow) - and it reaches 102 degrees celsius. Thermal throttling would kick in at 105 degrees. The GPU computes fine, but I suspect it would die soon at this temperature, so I am considering an extra case fan here as well.

Thermal throttling usually kicks in at 105 degrees. This is about the limit where a GPU does not take immediate damage, but would cease to function when heated even further. So it should not be operated in that range for extended periods of time.


I've seen posts claiming all over the web with screen shots to prove it that the max is 120 C.

Passive cooling is a bad idea, I once had passively cooled 8500 which ran fine at 80C, however, my q6600 was not happy with the ambient case temperature and would cause a reboot every 3 to 6 hours, despite providing it with excellent in case air flows, front, side, and rear fans, and even tried open cased.

The quad dropped 20C after I upgraded to a GTX 280.

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 2236 - Posted: 10 Sep 2008 | 19:48:34 UTC

Relief the poor 8500 and attach a big and slow fan to it, by wire or cable tie. It's easy and will do wonders for the temp!

Concerning 105 or 120°C max: guys, it absolutely doesn't matter! Running your card at 90°C 24/7 is unhealthy enough, anything above is totally out of the question.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Deleted
Avatar
Send message
Joined: 7 Sep 08
Posts: 6
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 2245 - Posted: 11 Sep 2008 | 2:22:48 UTC
Last modified: 11 Sep 2008 | 2:28:14 UTC

> four nVidia G92 GPUs ran at up to 90 degrees on a single MSI K9A2 Platinum V2 mainboard.

Why do you run 4 G92s on one MB? Quad SLI? Woh, you must get some 'mendous frame rates! What rate do you get with Crysis at max quality?


> critical temps for a GPU are about 100°C

!!!
BOILING temperature?? No WAY!! I'm using peltier to cool my CPU waterblock to almost freezing. You can like totally *forget* 100C anywhere near my computer!

BTW, It turns out 51 degrees was my idle temp. This CUDA distriputed computing place has the... peculiar... policy of only letting you do 2 WUs a day, and when I checked, it wasn't running. I can't check now because I already did my quota of 2.

But I can tell you for sure that if I'm anywhere over 70, that I'm not running CUDA anymore!

Hmmmm, or I could water cool my GPU. But I never game, so it would be just for CUDA. I wonder how cold that'll make it. Time to go to Tom's hardware!

--flk

PS
Does anyone else run into their daily quota and have an idle GPU for most of the day? Is that okay with y'all?

____________
A cerebral mutant, damned by God and doomed by man to understand EVERYTHING and control NOTHING.

My blog

Profile Stefan Ledwina
Avatar
Send message
Joined: 16 Jul 07
Posts: 464
Credit: 135,911,881
RAC: 123
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 2249 - Posted: 11 Sep 2008 | 4:45:43 UTC - in response to Message 2245.



...policy of only letting you do 2 WUs a day, and when I checked, it wasn't running. I can't check now because I already did my quota of 2.


PS
Does anyone else run into their daily quota and have an idle GPU for most of the day? Is that okay with y'all?


Well... Please see My answer in the other thread. Normally the daily quota would be higher, but because your card only produces errors (because it's not compatible), the quota gets reduced.

Runtimes on my GTX 260, which is one of the faster cards, are about 9 hours and a few minutes. The normal quota is at 4 WU/day I think. 9x4=36, so there's no way to hava a GPU sitting idle if everything works as it should, even with the faster GTX 280...
____________

pixelicious.at - my little photoblog

Post to thread

Message boards : Graphics cards (GPUs) : Anybody know the max GPU temp?

//