| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: HCC with GPU |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 486
|
|
| Author |
|
|
KWSN - A Shrubbery
Master Cruncher Joined: Jan 8, 2006 Post Count: 1585 Status: Offline |
It is also unfair to compare credit systems across projects. Credit inflation is a known issue and WCG is consistently the lowest granting credit project out there. Milkyway is one of the highest, outdone only by one math project in particular.
----------------------------------------At the end of the day, credits mean nothing cross-project. I will agree that GPUs can way outperform a CPU, but as you said yourself "on the correct project".  Distributed computing volunteer since September 27, 2000 |
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges: 
|
It is also unfair to compare credit systems across projects. Credit inflation is a known issue and WCG is consistently the lowest granting credit project out there. Milkyway is one of the highest, outdone only by one math project in particular. At the end of the day, credits mean nothing cross-project. I will agree that GPUs can way outperform a CPU, but as you said yourself "on the correct project". You can't compare MW to WCG unless you are talking about the MW CPU client. BTW, I think you forgot about Moo! and DNETC. |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges: 
|
You can compare a SB at MW to a GPU. You can then look at the relative credits of the CPU at MW to another CPU project and extrapolate the ratio to a GPU, to find out what sort of credits could be had, if a suitable GPU app could be used at that project.
You could do the same at other projects that use both a CPU and a GPU. |
||
|
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges: 
|
Well, I just picked up a new Radeon HD 4350 (25 max watt load) for $10 after rebate. I know it isn't a heavy hitter, but will fit nice in an empty PCI Express slot in one of my farm PC's. I guess it will crunch one of those Math projects until WCG decides they want it.
----------------------------------------    [Edit 2 times, last edit by Coleslaw at Sep 10, 2011 11:37:17 PM] |
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges:
|
Well, I just picked up a new Radeon HD 4350 (25 max watt load) for $10 after rebate. I know it isn't a heavy hitter, but will fit nice in an empty PCI Express slot in one of my farm PC's. I guess it will crunch one of those Math projects until WCG decides they want it. Not a heavy hitter, but it's 7x faster than the GPU that came in my other PC running CC.  |
||
|
|
TimAndHedy
Senior Cruncher Joined: Jan 27, 2009 Post Count: 267 Status: Offline Project Badges:
|
Heard nothing on a 1:1 comparison... reading posts here [on the public side] that an OpenCL performance is barely matching a CPU, but then, it's still an extra "core" so to speak and WCG has tackled the NVIDIA/ATI differences without a need of a 6.12 client, smelling like platform optimization. How is the stuff org'd. Who actually does the work on these things. The WCG staff? |
||
|
|
rilian
Veteran Cruncher Ukraine - we rule! Joined: Jun 17, 2007 Post Count: 1443 Status: Offline Project Badges:
|
---------------------------------------- |
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges:
|
---------------------------------------- |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 442 Status: Offline Project Badges: 
|
it has to do a lot with whether single or double precision maths are used. double precision currently is verry slow on gpus iirc Depends on what GPU. Another BOINC project prefers high-end AMD/ATI GPUs because they have hardware support for double precision, and their project runs the fastest on those GPUs. The Fermi series of Nvidia GPUs also has hardware support for double precision, but of a type that can run at maximum speed only for the rather expensive Tesla series of graphics cards. The GPUs that don't have hardware support for double precision are rather slow in their software versions of double precision, though. |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 442 Status: Offline Project Badges:
|
Yeah, only Tesla capable of doing double precision point at the moment (AFAIK). I think the Fermi boards series 4XX and new 5XX are double precision capable. They cost much less than the Tesla. I am not so sure regarding the 2XX series. They work very well on GPUGrid but I do not know if double precision is needed. Anyway if HCC comes in GPU format I will be ready  There are a few boards at the low end of the 4xx series for which Nvidia has not made it clear if they use Fermi series chips or not, so it's not clear if those have hardware double precision. Apparantly, all the Fermi series GPU chips for desktops have hardware double precision, but rather few of those for laptops and other mobile computers. GPUGRID makes little if any use of double precision; if any at all, it's little enough they can put it in the CPU portion of those workunits. However, there's rather little support for OpenCL GPU workunits in BOINC yet; Milkyway@Home appears to be the only BOINC project that has developed a workaround for that yet. It looks like it will be past the 6.13.* series of BOINC versions before that is changed. |
||
|
|
|