Advanced search

Message boards : Graphics cards (GPUs) : OMG it's Summer!

Author Message
Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 16633 - Posted: 29 Apr 2010 | 18:29:45 UTC

Is it only me, or is Summer causing a drop in RAC in other people's PC's? Winter was great! I never saved as much money on heating as I did this winter. Now it's Spring & that's not Summer. My open window shut while I was away & it was like a furnace at home, almost all WU's failed due to the heat.

Would there be a smaller chance of WU's failing if they were shorter & if so, would it be a good idea to have short WU's in Winter & long ones in Summer. Or maybe being able to choose between short & long WU's so that people in the South won't get miffed with the people of the North? Also, if there was a choice, smaller GPU owners (or those with bad cooling) can choose the short WU & monster water-cooled GPU owners can choose the long ones.
____________

fractal
Send message
Joined: 16 Aug 08
Posts: 87
Credit: 1,248,879,715
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16649 - Posted: 29 Apr 2010 | 22:53:55 UTC

I keep a thermostatically controlled exhaust fan in the window of the room with "the grove" year round. The only time it gets taken out of the window and the window closed is when it is below freezing outside over night, something that happens here less than a week a year. I suppose that is less viable for people in more varied climates.

Either way, the grove DOES get trimmed during the summer to keep the room tolerable.

cristipurdel
Send message
Joined: 31 Mar 10
Posts: 45
Credit: 103,429,292
RAC: 0
Level
Cys
Scientific publications
watwatwatwat
Message 16662 - Posted: 30 Apr 2010 | 7:37:18 UTC - in response to Message 16649.

My laptop is also starting to heat up. Too bad that TTrottle isn't recognizing the temp sensor. I need a temp limit option in BOINC...fast.

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 16665 - Posted: 30 Apr 2010 | 7:56:59 UTC

There are a couple of things you can do:

- increase the GPU fan speed
- lower GPU core and Shader clocks a bit if you get errors
- clean the heatsink / fan (be careful not to damage your card)
- replacing the thermal compound (be more careful not to damage your card)
- add / increase case air flow
- lower the GPU voltage (and clocks accordingly) (mostly only possible via BIOS editing - not for the unexperienced!)

I need a temp limit option in BOINC...fast.


Don't count on that. There are many chips and different ways to read their temperatures out. That's a lot of work, so the only practical way would be to interface with some existing 3rd party tool. Not exactly what BOINC traditionally does ;)

MrS
____________
Scanning for our furry friends since Jan 2002

Snow Crash
Send message
Joined: 4 Apr 09
Posts: 450
Credit: 539,316,349
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16673 - Posted: 30 Apr 2010 | 13:08:03 UTC

Would there be a smaller chance of WU's failing if they were shorter

If your failures are due to heat, then no, shorter WUs would not help.
____________
Thanks - Steve

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16677 - Posted: 30 Apr 2010 | 14:38:08 UTC - in response to Message 16665.
Last modified: 30 Apr 2010 | 14:38:49 UTC

I need a temp limit option in BOINC...fast.

Don't count on that. There are many chips and different ways to read their temperatures out. That's a lot of work, so the only practical way would be to interface with some existing 3rd party tool. Not exactly what BOINC traditionally does ;)

Try eFMer TTthrotle to limit temps. Also works in conjunction with BoincTasks:

BoincTasks BOINC Client Monitor & Control

BOINC TThrottle Client Temp Control & Monitor

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16680 - Posted: 30 Apr 2010 | 15:01:24 UTC - in response to Message 16677.

A couple of additional suggestions.

Configure your CPU (that you also crunch with) to run slower.
That will reduce the system heat and in turn the temps of your GPU!

- Start, Control Panel, Hardware and Sound, Power Options,

Alternatively,
Use Less CPU cores :)

EVGA Precision is a simple (and free) tool to reduce/increase your core speeds (GPU core, RAM, Shaders). It also lets you control the fan speed on some cards.

PS. Laptops are best sat on a flat surface, soft chairs and beds are bad!
There are also USB trays/ laptop cooler pads that you can get to sit your laptop on.

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 16682 - Posted: 30 Apr 2010 | 15:50:07 UTC

I moved a few things around. I've got a casing where the 2 front fans can be mounted differently. I put them both at the bottom, the HDD in the middle, cleaned up the cable mess & made sure my window didn't shut unexpectedly. But even though I got better temps, the problem was the RAM sticks that used a higher frequency, that I remembered to increase in voltage, but I forgot to increase the NB voltage due to the higher frequency of the RAM. Now it looks good! ;-)
____________

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 16714 - Posted: 1 May 2010 | 10:47:06 UTC
Last modified: 2 May 2010 | 11:26:41 UTC

I've got my HD4870 running Milkyway at full throttle since a year now. I mounted an Accelero S1 with 2 x 120 mm fans and initially it ran in the mid 50°C. After clients optimization temperatures increased into the mid 60°C range. During winter I noticed that I had reached mid 70°C, even without major changes to the PC. By now I was at 80°C and had to throttle back 10 MHz - not pleasant. Today I replaced the stock thermal pad (it had become rather solid) with my age old Artic Silver 2. Now I'm back to 73 - 74°C at 23°C ambient, so will likely reach 80°C again during real summer. But an improvement of 6°C is nothing to sneeze at!
EDIT: today I'm down again into the mid-60°C range, at similar room temperature. Very nice! Maybe the AS2 needed some time to reach its full potential.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16715 - Posted: 1 May 2010 | 11:12:13 UTC - in response to Message 16714.

I've got my HD4870 running Milkyway at full throttle since a year now. I mounted an Accelero S1 with 2 x 120 mm fans and initially it ran in the mid 50°C. After clients optimization temperatures increased into the mid 60°C range. During winter I noticed that I had reached mid 70°C, even without major changes to the PC. By now I was at 80°C and had to throttle back 10 MHz - not pleasant. Today I replaced the stock thermal pad (it had become rather solid) with my age old Artic Silver 2. Now I'm back to 73 - 74°C at 23°C ambient, so will likely reach 80°C again during real summer. But an improvement of 6°C is nothing to sneeze at!

Eight HD 4770 cards currently running MilkyWay, all stock cooling, 6 are 2 cards per machine. Temps are 45C - 61C depending on the case and card placement. All are OCed to 830MHz. GPU utilization is 99%. Unlike my NVidia cards, they just run. Not a problem ever. No crashes, no failed WUs, no problems at all. Same with Collatz except you can subtract about 3-4 degrees temp/card.

Profile liveonc
Avatar
Send message
Joined: 1 Jan 10
Posts: 292
Credit: 41,567,650
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwat
Message 16718 - Posted: 1 May 2010 | 14:16:05 UTC

I'm starting to wonder if there have been some GTX260(216) unfriendly WU's. I got errors on 2 different PC's today. One running Windows 7, the other running Mint Linux 8. It's not been extra hot today & my windows didn't shut so ambient temps haven't been bad.
____________

Snow Crash
Send message
Joined: 4 Apr 09
Posts: 450
Credit: 539,316,349
RAC: 0
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16720 - Posted: 1 May 2010 | 14:40:53 UTC - in response to Message 16718.

It is not you. The *GA* tasks have been having difficulties. There is a post in the GA thread that we should abort all GA8* and GA9* but they have not been pulled from rotation so they still show up from time to time. I assume the project has decided it is easier to just let them hit the "too many errors" status instead of pulling them manually. I personally have had quite a few failures with GA10* and hope we can get them straightened out or get some official work on if we should abort them or not.
____________
Thanks - Steve

Profile K1atOdessa
Send message
Joined: 25 Feb 08
Posts: 249
Credit: 392,702,681
RAC: 1,417,376
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 16730 - Posted: 1 May 2010 | 19:41:23 UTC - in response to Message 16720.

It is not you. The *GA* tasks have been having difficulties. There is a post in the GA thread that we should abort all GA8* and GA9*... I personally have had quite a few failures with GA10* and hope we can get them straightened out or get some official work on if we should abort them or not.


I've had all failures of GA10 the past week or so (about 3-4 WU's, costing about 15 hours of crunching). No / minimal issues with other WU's, so I'm manually killing GA10's if I see them.

Post to thread

Message boards : Graphics cards (GPUs) : OMG it's Summer!

//