Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Beta Testing Forum: Beta Test Support Forum Thread: HCC GPU Testing v6.45 (Started June 25, 2012) |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 62
|
Author |
|
katoda
Senior Cruncher Poland Joined: Apr 28, 2007 Post Count: 170 Status: Offline Project Badges: |
Interesting, but again no results returned by nVida 2xx cards, including my GTX280 begging for data to crunch on 32bit Vista. I'm even more surprised seeing one workunit returned by nVidia 8600GT card, which is far less powerful than nVidia 2xx cards. Why such gap?
---------------------------------------- |
||
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges: |
Scribe, does collatz use part of the CPU along side the GPU? Since the beta GPU wu's here uses part of the CPU, it's going to share resources if your crunching 100% CPU. Since I reduced my crunchers by 1 thread, have never had issues crunching with the GPU. When I crunch Milky way, don't have to reduce a thread as it doesn't (AFAIK) use CPU resources.
----------------------------------------
Crunching for humanity since 2007!
|
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Boinc Manager shows it using 0.01 CPUs + 1 NVIDIA GPU
|
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
For Milkyway it shows 0.05 CPUs and 1 NVIDIA GPU
|
||
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: |
@ retsof- You can't run all of you CPU cores at 100% and run GPU tasks at the same time. That's what is causing your computer to lag. Set your preferences to use 83% of your CPUs and that will leave 1 to feed the GPU. Yes you can, I do it all the time running WCG at 100% on all 4cores and Collatz on GPU, no problems whatsoever. Collatz and WCG are not anywhere near the same animal. WCG GPU tasks use 20%+ (0.20) of a CPU core while running on ATI cards and even more on Nvidia. Collatz uses <1%. If you don't have a core free to help the HCC GPU tasks then I can guarantee that you will have issues.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
|
||
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: |
@ retsof- You can't run all of you CPU cores at 100% and run GPU tasks at the same time. That's what is causing your computer to lag. Set your preferences to use 83% of your CPUs and that will leave 1 to feed the GPU. I wondered about that, but where would the gain come from, a GPU faster than a core? GPU processing is at least 10-20x faster than CPU processing. I am finished with the HCC project pre GPU, and only got a few GPU betas to try to see what issues may arise. This is one issue, it seems. I don't want to leave a core open all of the time when other projects are running. I could go back to graphic processing and then knock it down to 83% if more HCC betas come along. I suppose. I would have a few days in the buffer queue to notice them. You could try some other GPU projects while you wait for the HCC GPU to go live. That will give you some practice and feedback on what to expect.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
|
||
|
ryan222h
Senior Cruncher Joined: Sep 4, 2006 Post Count: 425 Status: Offline |
I have a question. If you set your prefs to allow one thread free, won't that thread be free all the time, whether or not you are crunching with gpu or cpu? I'm thinking the GPU will not benefit with a thread free since the GPU is also limited to using that 83% cpu as well, is it not?
----------------------------------------Put another way, the GPU is still not allowed to use the extra 17% of cpu that is available, is it? Because the boinc prefs are set to 83% total usage. I do see how freeing up a thread can help with total system responsiveness when running applications alongside boinc. [Edit 5 times, last edit by ryan222h at Jun 27, 2012 6:21:55 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The 83% refers to the number of cores *for CPU only tasks*, This means, that CPU tasks on an octo can use 7 cores. This is not the limit *per CPU core*, that's still 100%. If need be the GPU task can use up to 100% of the free core. The effect is, no GPU crunching, no use of the 8th core, but if GPU crunching, the 8th core is used.
----------------------------------------edit: One important note... some GPU tasks use very little CPU resource, so in that case you need not *reserve* the core, but when that GPU task does use a substantial portion of the CPU core, you have contention for the same resource, which is when the user experience becomes less dandy. Could turn out so bad that the overall performance of BOINC on that system drops more than just reserving the one core and have it idle. --//-- [Edit 1 times, last edit by Former Member at Jun 27, 2012 6:27:37 PM] |
||
|
ErstwhileIII
Cruncher Joined: Nov 16, 2004 Post Count: 6 Status: Offline Project Badges: |
Received 4 Beta tasks a few days ago --- none since then.
Has something changed?! |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: |
The 83% refers to the number of cores *for CPU only tasks*, This means, that CPU tasks on an octo can use 7 cores. This is not the limit *per CPU core*, that's still 100%. If need be the GPU task can use up to 100% of the free core. The effect is, no GPU crunching, no use of the 8th core, but if GPU crunching, the 8th core is used. That is why this will be a difficult project to support, until they have a steady supply of work units, since you will need to leave a core idle most of the time. But once the work units start pouring in (from HCC or others down the road), then it will tie up a graphics card to the point of un-useability. So if I do this project, it will be only when I can devote a spare card to it (not a display card), and can give it a core all its own. It will be great then, but this beta thing is going to be painful. |
||
|
|