Hi all

I was wondering if there is anyway for the gpu usage to be limited? I have tested against 9800GTX+ (killed it), 260GTX. 560GTX ti cards and all run flat out when running cuda/fermi tasks. As you can imagine things tend to get hot (very) and loud.
Any help would be good, cant keep killing cards :(

There's nothing built into BOINC that I'm aware of that controls how much of your GPU is used.

In my humble and professional opinion, the best cure for thermal issues is better cooling. Instead of spending all the money replacing graphics cards due to overheating, get some extra fans, or if possible a case with better air flow.

BOINC its CPU throttle (Use at most X% of CPU time) will throttle the GPU as well. However if you want more (temperature) control, use Tthrottle.

---

see now there be the problem, it doesnt appear to throttle the gpu just the cpu?
example: processor set to 20% max usage with 50% of processors allowed, 1/2 hr and the 260gtx was too hot to touch at the exhaust, but the processor is running cold.
anyway thanks for the links i'll give em a go.

and as to the suggestion of better cooling, well i have a cm stacker (orig ver)
with a cm V10 on the processor, 2 x 120 blowing in with a 120 on the back to extract. the PSU is a HX1000, 2x 560ti, quite in normal operation but do get loud when stressed.
the idea i had was turning it down to reduce the heat, not increasing the airflow to deal with the overheat :)

see now there be the problem, it doesnt appear to throttle the gpu just the cpu?

As long as all GPU applications run on the CPU, when you then throttle the CPU it'll throttle the GPU. You can check that, in Windows Task Manager you'll see that setiathome_6.09_windows_intelx86_cuda23.exe (or similar of name) will be running on one CPU (core).

The CPU takes care of translating the contents of the task to something that the GPU can understand plus transporting of this content to the videocard's memory, then after the GPU's done with the calculations, the CPU picks up the result from the videocard's memory, translates it back to something the humans can understand and writes it to disk.

If you then throttle the CPU, you will slow this process down, since throttling in BOINC-land means that you suspend and resume the task. 20% CPU time means that per 10 seconds the application is active for 2 seconds and sleeps for 8 seconds.

Of course, it depends on where you set the preferences. If on-line, but you're using the local preferences, then the on-line preferences won't be used as the local ones override the web-preferences. Even having looked once at the local preferences and exited the preferences window with the OK button means you are using the local preferences.

---

and as to the suggestion of better cooling, well i have a cm stacker (orig ver)
with a cm V10 on the processor, 2 x 120 blowing in with a 120 on the back to extract. the PSU is a HX1000, 2x 560ti, quite in normal operation but do get loud when stressed.
the idea i had was turning it down to reduce the heat, not increasing the airflow to deal with the overheat :)

Interesting. My friend is using a GTX 560 Ti (albeit just a single one) in a less air-efficient case and isn't experiencing nearly as much heat as you seem to be. The ambient air temperature probably makes a difference, but he's only using two 80mm case fans (intake and exhaust).

I have to agree with "Pond Grass" with the program THROTTLE. I also use that program on my main computer and does seem to keep things under control, but if I run the GPU full throttle then the THROTTLE program will compensate but doesnt do a very good job at it. So what I have done to try to help the GPU is go into the computer preference settings in boinc and unchecked the "use GPU while computer is in use" to keep the GPU from running all the time. It seems to help my situation but the hottest I have see my GPU get around is 150 degrees(f) which I think is about max temperature the GPU can handle. My Gpu will run with this setting but the computer use will not be interupted, as if it is, the GPU will stop. I also have noticed that the the GPU will run for a certain amount of time and then stop without me interupting the computer. This may be that something is running in the background at the moment and interupts the GPU which I dont have a problem with. I do use nortons 360 and is always doing constant checks which would interupt the computer but there may be other processes (which im not aware of) that may be stopping the GPU at that moment. So what I have done to compensate for this is change the amount of "computing allowed", "only after computer is idle for (x) minutes" to 1 minute. This way if anything interupts the computer while running the GPU, it will start back up again after 1 minute. Now you can set the amount of time to whatever you want and if your case you may want to increase the amount of time to try to give the GPU time to cool down before it runs again. Although I dont know what kind of ambient temperatures you have where your computer is at as it does play a role on cool air thru your computer. I have a HAF tower and with all the cooling fans going, I still need the ambient air surrounding the computer tower to be a cool temperature......needless to say the computer is in a room with an air conditioner at 68 to 71 degrees(f) which does help alot......mind you I run all the stock fans that came with the computer and have added a couple on the tower for extra measure and does seem to help. One other thing and im not saying that you dont do it but make sure you keep the GPU card and all other parts on the computer free of accumulating dust as this will increase unwanted high temperatures and cause your GPU to fail at some point. If I am wrong with anything I have said I would like some input as I am new to this and am learing but would except any great info that may come my way. Hope this was some help to ya.

could be a SLI issue with my 560ti cards, havent tried them as a single yet :)
1 does get hotter than the other, perhaps time for some directed airflow.
the 9800GTX+ in alt PC seems good but still a little hot.
i think i'll test some more, if i find anything intersting i'll report it back :)

i tried multiple settings for CPU and it doesnt appear to have any effect on GPU used, it still runs flat out when active. There needs to be a GPU limit parameter similar to CPU limit.

---

nvm, tthrottle has a GPU temp limit function

---

Hi all, thought i'd check in with an update, have now balanced my usage on the graphics cards (3 screens, main mon + 1mon on card1 and 1mon on card 0) and upgraded the coolers on both cards, that also required a MB upgrade so i could get them to fit :) Now have 2x Arctic PlusII(GPU) and a CM V10 HSF, 2x120 in and 1x120 & 1x80 out
This has helped to bring temps down dramatically, i still have the issue of the cuda/fermi units grabbing 80%+ of the GPU, although no overheat problems even after several hours. I would respectfully request from the powers that be, that a feature be added that will allow this level of control (pretty please) for the GPU.
and although i have looked at 3rd party controller S/W i'd prefer not to use it

What type of temperatures are you guys getting on your GPUs?

GTX 470
Idle 64c
Load 78c

GTX 295 (side panel off)
Idle 58c
Load 95c

I had to take the side panel off the case that the 295 is currently in due to load temps going as high as 105c! From what I've read that's a little warm.

---

My Computers:
â–ˆ Blue Offline
â–ˆ Green Offline
â–ˆ Red Offline

What type of temperatures are you guys getting on your GPUs?

GTX 470
Idle 64c
Load 78c

GTX 295 (side panel off)
Idle 58c
Load 95c

I had to take the side panel off the case that the 295 is currently in due to load temps going as high as 105c! From what I've read that's a little warm.

Have you set your fan speed manually. If not, why not?

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

How can I manually set the fan speed?

---

My Computers:
â–ˆ Blue Offline
â–ˆ Green Offline
â–ˆ Red Offline

Use MSI Afterburner or EVGA Precision
http://event.msi.com/vga/afterburner/download.htm
http://downloads.guru3d.com/Afterburner-2.0.0-Final-download-2562.html

http://www.evga.com/precision/
http://majorgeeks.com/EVGA_Precision_d6209.html

Screenshots:
http://setiathome.berkeley.edu/forum_thread.php?id=65894&nowrap=true#1165854
http://setiathome.berkeley.edu/forum_thread.php?id=65894&nowrap=true#1165516

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

I downloaded EVGA Precision and I'm now able to keep my 470's around 70c by turning the fan up to 75%. I'm going to be putting both my 470s in my Blue rig today so hopefully I'll still be able to keep temps down.

For the 470 & 295, what are good temps to shoot for?

---

My Computers:
â–ˆ Blue Offline
â–ˆ Green Offline
â–ˆ Red Offline

Also check that the heatsink/fan combo on you video card are working correctly: you may have to take off the heatsink/fan, scrape the thermal paste off and apply a new layer...

---

.

Hello, from Albany, CA!...

What type of temperatures are you guys getting on your GPUs?
GTX 295 (side panel off)
Idle 58c
Load 95c

Way way to hot.

My 2 cards are next to each other and read:

68, 57, 75 83 on full load at 24C room temp.
You need more fan speed.

---

TThrottle Control your temperatures. BoincTasks The best way to view BOINC. Anza Borrego Desert hiking.

That's what I thought. Seems like my Antec Sonata III isn't the best fit for for a 295. Going to try 3 of them in a HAF 912 next.

---

My Computers:
â–ˆ Blue Offline
â–ˆ Green Offline
â–ˆ Red Offline

HI,
Been lurking::-) Thanks for the info on afterburner, my gtx460 was running 77c and there isnt space between it and its compainion gtx560 to put another cooler in.. Afterburner allowed me to push the fan to a higher % it was running at 46% with the temp at 77c:-( Now its at 81% and the temp on the card is down to between 56c and 65c.. min 12c lower.

Regards,

---

Cliff,
Been there, Done that, Still no damm T shirt!