| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 28
|
|
| Author |
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
Jacob:
----------------------------------------Here's a simple answer to solve your problem: BUY a ATI card. Use the one for the project your involved in which works best. I have a EVGA 460GTX downstairs on my 2600K machine and from what I see trying to run more than one GPU WU at a time is pretty pitifull.. Here's 20 cpu threads and 12 gpu threads using a cheap $200.00 Asus 7850 card.. This machine did app 140,000 WCG PPD on just cpu and npw with that cheap card is doing app 620,000 PPD Yes, TWO 8 core cpu's.. http://img255.imageshack.us/img255/8365/12gpuwu.jpg Oh, one last point and no sarcasm meant, all in the spirit of friendship: Cut nanoprobe a little slack. He really is a decent guy and I can tell you from the heart that he helped many of us at XS understand the in's and out's of GPU based work here when it first started.. I think you misread his first comments and took them the wrong way. It happens and to me also but God's truth we are all here for the same thing and thats to get the work done. Better as friends yes? I wish they had a smiley for a hand extended in friendship because that is where all our heads should be at.. ![]() |
||
|
|
Jacob Klein
Cruncher Joined: May 31, 2007 Post Count: 28 Status: Offline Project Badges:
|
Hello Movieman,
----------------------------------------:) Thanks for the kind words. I think you guys also mis-understand me too! My "problem" is not "I want to crunch for this project a lot more than I can currently". If that was my problem, then you're right that that the solution would be to buy an ATI card. Instead, my "problem" is: given the tools I have, I believe performance can still be improved by an application change. Regarding running 2-at-once on NVIDIA: You mention having a GTX 460, which is the same video card I'm using for my testing. My results, using that card, are that I can do a single WCG HCC GPU unit in about 12 minutes, but I can do 2 WCG HCC GPU units (at the same time on the same GPU) in about 19-22 (average 21) minutes. So... That's about a (((12 * 2) - 21) / (12 * 2)) = 12.5% improvement. I don't consider that improvement pitiful at all! Are you able to replicate those results? Note: The reason it varies so much (19-22 minutes) is because, at least on NVIDIA, these tasks have 2 lengthy segments where they only do CPU work and leave the GPU idle. When running 2-at-a-time, sometimes both tasks are doing those segments at the same time (leaving the GPU idle; 22 minutes). But sometimes the segments are offset (and the GPU is never left idle; 19 minutes). And, of course, the reason it's quicker even when the segments happen at the same time, is that: GPU Load for 1-at-a-time is 94-96%, but GPU Load for 2-at-a-time is 98%. As you can see, I'm presenting my findings through actual data, and actual results. And I'm always looking to improve, using the tools that I have (NVIDIA cards). Knowing all of this... I believe performance can be even further improved, if the GPU tasks ran with a process priority of 6, and a cpu-intensive-thread priority of 6 or 7. Kind regards, extending that hand of friendship (good to meet you!), Jacob [Edit 7 times, last edit by Jacob Klein at Apr 10, 2013 9:28:11 AM] |
||
|
|
rousku
Cruncher Joined: Mar 21, 2013 Post Count: 9 Status: Offline |
My Radeon HD 6950 works at about 97-98% load (according to GPU-Z) with HCC projects, causing massive stuttering if i try to use my browser or file explorer etc. Strangely, Milkyway@Home GPU tasks do not cause the same issue, even though GPU-Z says the load is 100%. I use this computer a lot and as such, setting an idle timer for GPU work would equal pretty much no GPU work done whatsoever (i have to shut down the PC overnight anyway).
----------------------------------------I hope future BOINC GPU supported projects don't have this issue, and can be run while using the computer for normal activities simultaneously. Obviously i have my gaming executables set as exclusive applications, so that's not an issue here. Not that i have a lot of time for gaming these days anyway...[Edit 1 times, last edit by rousku at Apr 10, 2013 12:31:38 PM] |
||
|
|
branjo
Master Cruncher Slovakia Joined: Jun 29, 2012 Post Count: 1892 Status: Offline Project Badges:
|
AFAIK, Asteroids@Home doesn't support GPGPU crunching
----------------------------------------![]() ![]() Crunching@Home since January 13 2000. Shrubbing@Home since January 5 2006 ![]() |
||
|
|
rousku
Cruncher Joined: Mar 21, 2013 Post Count: 9 Status: Offline |
You are correct - i meant to say Milkyway@Home. Doh!
|
||
|
|
Movieman
Veteran Cruncher Joined: Sep 9, 2006 Post Count: 1042 Status: Offline |
Hello Jacob. I won't argue your logic that a 10-12% increase is meaningfull. It is but when you can buy a AMD 7850 like I did for under $200.00 and run not 2 but 12 GPU WU at a time and in my case add 480,000 WCG PPD it at least to me loses it's importance especially when the program is ending shortly.
----------------------------------------Now I'm not wealthy or I would have bought this card long ago but it seemed like a very good investment for what it does and so far is as smoothe as a baby's bottom! ![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
... I found my own answer!
----------------------------------------[Edit 1 times, last edit by Former Member at May 2, 2013 8:55:48 PM] |
||
|
|
Jacob Klein
Cruncher Joined: May 31, 2007 Post Count: 28 Status: Offline Project Badges:
|
Gandalf,
I'm not sure why you replied to this thread titled "HCC GPU - Thread Priority should be increased (to 6) to not starve the GPU". You should consider creating a new thread for your question. Thanks, Jacob Klein |
||
|
|
|