| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: It is time to move to CUDA/BOINC !!!! |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 87
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hellloo again, hahahaha
----------------------------------------almost more than 9 month from the start of my post here ... i had time to have a baby girl ... still no cuda upgrade .... what a shame .... oupsss, time feed the monster ... [Edit 1 times, last edit by Former Member at Sep 26, 2009 11:33:36 PM] |
||
|
|
twilyth
Master Cruncher US Joined: Mar 30, 2007 Post Count: 2130 Status: Offline Project Badges: 
|
It's not really a problem since most of the people into DC just put their 2.66 Tflops rigs on Folding@Home.
----------------------------------------  [Edit 1 times, last edit by twilyth at Sep 27, 2009 12:51:24 AM] |
||
|
|
realracer2
Cruncher Joined: Mar 16, 2007 Post Count: 5 Status: Offline |
I would be happy if World Community Grid projects started using Cuda. I am currently running other DC projects in Cuda, but would drop them for a World Community Grid Cuda project.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
just for the record.. on an AMD phenom9950 quad2600 + nvidia9500gt 512mb ram + 4GB 800mhz RAM
----------------------------------------on linux ubuntu 9.04 64bit project HelpCureMuscularDystrophyphase2 took 3hours 12minutes to end.. on win7 64bit same project,same machine it took 6 hours to complete.. on win Vista 64 bit Ultimate it took 4hours 59minutes.. great difference .... I vote linux for crunching.. [Edit 2 times, last edit by Former Member at Sep 30, 2009 5:53:27 PM] |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Tasks vary in size. Gather some average statistics, then we'll talk.
|
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
You cannot take HCMD2 as base for comparisons of run time because they are highly variable from an hour and less to 12 hours and more. The project that has a highly consistent run time is HCC and you'd have to take probably a dozen per platform to get an approximation.
----------------------------------------But, I've missed the link of different platforms crunching on CPU to CUDA/BOINC. Got some insight for me/us?
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
H_A
Cruncher Joined: Sep 14, 2009 Post Count: 6 Status: Offline |
ok some average statistics by me...
----------------------------------------same machine , no virtual box, clean installs... Ο.S | Project | cputime WinXp SP3 | HCMDphase2 | 6:23:12 >> | Nutricious Rice 6.17 | 7:23:58 Win 7 64bit | HCMDphase2 | 6:09:26 WinVistaSP2 32bit | HCMDphase2 | 6:27:14 >> | HumanProteomephase26.03 | 9:27:33 >> | FightAIDS 6.07 | 9:56:04 Ununtu 9.04 64bit | HCMDphase2 | 4:15:40 >> | FightAIDS 6.07 | 6:49:22 the "weird" is that in benchmarks win7 64 bit gives .. 2144 floating point MIPS(Whetstone) per cpu 6542 integer MIPS (Dhrystone) per cpu in ubuntu 9.04 64bit gives.. 1959 floating point mips 4547 integer mips in winXP.. 2574 floating point mips 5909 integer mips in benchmarks win7 have the highest numbers but in "action".. poor results i am not trying to start any kind of flame war operating systems.. :) p.s Preferences and options in BOINC was the same.. [Edit 1 times, last edit by H_A at Sep 30, 2009 7:07:09 PM] |
||
|
|
roundup
Veteran Cruncher Switzerland Joined: Jul 25, 2006 Post Count: 838 Status: Offline Project Badges: 
|
ok some average statistics by me... same machine , no virtual box, clean installs... Very interesting - especially the relation between benchmarks and runtimes. How many results per OS and project work units are the basis of your averages? |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
in benchmarks win7 have the highest numbers but in "action".. poor results Most WCG-wu's has too large variation in cpu-times between wu's in the same sub-project to give any meaningful comparison of performance between OS. An example of how large the variations can be is my own results from a single computer running a single OS. In flu, the highest is... 28016 seconds The lowest is... 15776 seconds... Difference, 78%. CMD2 is... highest, 19382 seconds The lowest is... 192 seconds... Difference, 101x OK, there's been some adjustments on target run-time for CMD2, and the shortest is most likely a beta, so let's look only on the two last CMD2-results: Last result: 7708 seconds. Previous result: 19382 seconds Difference... 151%... With differences on 78% and 151% within same project, any comparisons would be difficult to do. Even if you're looking on averages, it's doubtful you've run enough results to get a better comparison of the different OS. A much better comparison would be to download some wu's, disable network-connection, stop BOINC and make a backup-copy, and run the exact same group of wu's across XP, Vista and Win7. Also doing linux would be more difficult, since can't use the same backup-copy...  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
H_A
Cruncher Joined: Sep 14, 2009 Post Count: 6 Status: Offline |
ok some average statistics by me... same machine , no virtual box, clean installs... Very interesting - especially the relation between benchmarks and runtimes. How many results per OS and project work units are the basis of your averages? generally i use linux 64bit, but i make this "tests" on my own to find the "best crunching os" for my hardware. i kepp the records and after 5-6 projects each (HCMD2, FAAH, HFCC) the results show me that in linux, i was "faster".. |
||
|
|
|