| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: Amount of work to still run on HCC |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 86
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I think that gb009761 is correct. As far as I know, the HCC work units are (currently) made up of 2 images. I call each image a 'job'. Each work unit can be processed by a CPU algorithm or by an OpenCL GPU algorithm. If we ever create a work unit with a lot of jobs (images) in it, we will have to run them on GPUs only, because the run time on a CPU would be impractically long.
Lawrence |
||
|
|
KieX
Cruncher Spain Joined: Dec 19, 2009 Post Count: 8 Status: Offline Project Badges: 
|
as far as I know, no. Obviously I could well be wrong here, but from what I've understood, it's just a 'big pot' - so, whether it's a CPU or a GPU asking for work, if there's some WU's there, they'll grab them. I think that gb009761 is correct. As far as I know, the HCC work units are (currently) made up of 2 images. I call each image a 'job'. Each work unit can be processed by a CPU algorithm or by an OpenCL GPU algorithm. If we ever create a work unit with a lot of jobs (images) in it, we will have to run them on GPUs only, because the run time on a CPU would be impractically long. Lawrence Cool. Thanks for the explanation  |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges: 
|
Great news.
----------------------------------------
Crunching for humanity since 2007!
 |
||
|
|
ThreadRipper
Veteran Cruncher Sweden Joined: Apr 26, 2007 Post Count: 1321 Status: Offline Project Badges:
|
Great news, thanks! :)
---------------------------------------- Join The International Team: https://www.worldcommunitygrid.org/team/viewTeamInfo.do?teamId=CK9RP1BKX1 AMD TR2990WX @ PBO, 64GB Quad 3200MHz 14-17-17-17-1T, RX6900XT @ Stock AMD 3800X @ PBO AMD 2700X @ 4GHz |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks for the update, nice to hear that this technology is implemented to analize further data.
|
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges:
|
Excellent news Viktors. My ROI for the ATI boards improve by the day.
----------------------------------------  [Edit 1 times, last edit by Hypernova at Dec 11, 2012 11:56:55 PM] |
||
|
|
JEklund2
Advanced Cruncher Finland Joined: Aug 10, 2006 Post Count: 119 Status: Offline Project Badges: 
|
Great
---------------------------------------- Now heading for 10 years for this project   |
||
|
|
coolstream
Senior Cruncher SCOTLAND Joined: Nov 8, 2005 Post Count: 475 Status: Offline Project Badges: 
|
I'd still like to know why the run time is allowed to go so high ( up to 10 hours before a WU times out).
---------------------------------------- Crunching in memory of my Mum PEGGY, cousin ROPPA and Aunt AUDREY. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
coolstream, the topic of this thread is how much work is left for the project, not a technical problem. My suspicion though is, that as HCC tasks come from 1 pool, not pre-assigned to GPU or CPU when generated, and the run time for CPU is set to time out at X times Y Flops. For a CPU task it would be about the regular time. Don't know what time-out factor is used... 5 times or 10 times or something in between [can be equated from data in the client_state.xml]
|
||
|
|
mmstick
Senior Cruncher Joined: Aug 19, 2010 Post Count: 151 Status: Offline Project Badges: 
|
Now, if only I could find a 7950 or 7970 at a good price!  The HD 8000 series will be out sometime in the first part of next year. You could wait for it, or at least the prices on the 7000 series will drop. But to get the most out of those cards, you have to devote several CPU cores to it, so if you are working on other projects, it may not be worth it. More or less, coming from an owner of a highly overclocked 7950, my 4GHz 8 core processor is just barely able to properly keep it fed. If you are to buy one, you'd have to dedicate all CPU cores to helping the GPU stay fed with data. Therefore, you'd want to be running the 7950 with a really high end CPU and overclock it as far as you can so that the GPU isn't spending half the time at idle. Don't even think about buying two, one 7950 per machine (Unless you magically have a 5GHz six core Intel CPU); 100K PPD per 7950 machine (700K WCG PPD). [Edit 3 times, last edit by mmstick at Dec 14, 2012 10:38:13 AM] |
||
|
|
|