| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: FightAIDS@Home
Thread: Autodock & CUDA |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 40
|
|
| Author |
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Please read the comments over at SIMAP and what their programmers observed between their 32 bit and recently released 64 bit version. Barely any difference and indeed, on my 32 bit their jobs used to run 45 minutes and now after upgrading to 64 bit OS they do 43 minutes... 4%.
----------------------------------------FAIK the Floating point portions of the calculations are still performed on 32 bit hardware. The integers do move much quicker so even without a recompile you'll see for some WCG sciences considerable throughput increase... 25-30% as has been measured for HCC and to a lesser extend also for all other WCG projects. And on FAAH CUDA... see my first post in this thread.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Sekerob> Could you point me to the reference numbers? I only found the following info:
There might be other posts on how to use boinc on 64 bit using 32 bit apps, but there weren't any when I wrote this. This is also only for the simap project, which from what I've read is the only project that actually uses the extra computational power of the 64 bit processor (i.e. their 64 bit app runs a lot faster than their 32 bit, so I'm contributing more than the average joe to their project). http://ubuntuforums.org/archive/index.php/t-259828.html |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 444 Status: Offline Project Badges: 
|
There might be other posts on how to use boinc on 64 bit using 32 bit apps, but there weren't any when I wrote this. This is also only for the simap project, which from what I've read is the only project that actually uses the extra computational power of the 64 bit processor (i.e. their 64 bit app runs a lot faster than their 32 bit, so I'm contributing more than the average joe to their project). http://ubuntuforums.org/archive/index.php/t-259828.html Check the date on that post - to me it looks like it says simap ran much faster on 64-bit processors back in 2006 (when it still used workunits that ran in 32-bit mode), but says nothing about how much faster it is now. Also, it doesn't appear to take into account the fact that 64-bit processors generally have a higher clock frequency than processors which can only do 32-bit, which should make them faster even when running workunits in 32-bit mode. I've been using 64-bit Windows Vista to run both WCG projects and simap for months, and have seen some signs that there are a significant number of crunchers using 64-bit Windows XP. [Edit 1 times, last edit by robertmiles at Dec 1, 2009 1:49:55 AM] |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 444 Status: Offline Project Badges:
|
Thanks a lot. Nvidia and ATI will soon be leaders in supercomputers ... or I hope so. Judging by what I see, I'd say that Nvidia will soon be a leader in supercomputers. ATI might eventually be a leader also, if they ever catch up in providing compilers for computer languages such as C++ and Fortran. |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: 
|
Also, it doesn't appear to take into account the fact that 64-bit processors generally have a higher clock frequency than processors which can only do 32-bit, which should make them faster even when running workunits in 32-bit mode. It's already difficult enough to run comparison tests without adding different clock speeds in the picture! All the tests I have done and the numbers I have mentioned in various threads have been done with the same machine and the same processor running at the same speed. I was just dual booting either an XP SP3 for 32-bit mode tests and an Ubuntu 64-bit for 64-bit mode tests. I was also making the tests in the two modes in the same period of time for trying to eliminate the variations that can affect the duration of WUs of a given project over time. And last I have chosen projects with consistent WU durations like HCC (the most consistent one), HFCC, FAAH and IADS. Cheers. Jean. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
One question. There is a paper published by NVIDIA http://www.lunarc.lu.se/Documents/nvidia-work...tion/90_Life_Sciences.pdf that shows that software Autodock 4 modified to use CUDA can give around 10x speedup, and they compare it to base dual Intel Xeon. Autodock is the base for the FAAH project, so it is hard to believe the project would really have little benefit of CUDA optimization. The optimized version is available as commercial product from Silicon Informatics. http://www.siliconinformatics.com/products.html
While of course NVIDIA source is prone to bias, there might be quite some truth in that. Anybody used their optimized version? |
||
|
|
mwgiii
Advanced Cruncher United States Joined: Aug 17, 2006 Post Count: 131 Status: Offline Project Badges: 
|
Sekerob> Could you point me to the reference numbers? I believe this is the thread he was referring to: http://boinc.bio.wzw.tum.de/boincsimap/forum/viewtopic.php?f=17&t=981   [Edit 1 times, last edit by mwgiii at Dec 11, 2009 6:21:28 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
That's the one :D
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 444 Status: Offline Project Badges:
|
Also, it doesn't appear to take into account the fact that 64-bit processors generally have a higher clock frequency than processors which can only do 32-bit, which should make them faster even when running workunits in 32-bit mode. It's already difficult enough to run comparison tests without adding different clock speeds in the picture! All the tests I have done and the numbers I have mentioned in various threads have been done with the same machine and the same processor running at the same speed. I was just dual booting either an XP SP3 for 32-bit mode tests and an Ubuntu 64-bit for 64-bit mode tests. I was also making the tests in the two modes in the same period of time for trying to eliminate the variations that can affect the duration of WUs of a given project over time. And last I have chosen projects with consistent WU durations like HCC (the most consistent one), HFCC, FAAH and IADS. Cheers. Jean. So you haven't tried to account for any speed differences between XP and Ubuntu, independent of the bit size? |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges:
|
So you haven't tried to account for any speed differences between XP and Ubuntu, independent of the bit size? Robert, I am not sure I understand your question correctly. Are you asking for comparisons 32-bit XP vs 32-bit Ubuntu, and 64-bit XP vs 64-bit Ubuntu? If so, no I have not done that, and I don't remember having read some. For 32-bit XP vs 32-bit Ubuntu maybe, some day, I will convert my P4 HT from XP to a dual-boot setup XP/Ubuntu. But it won't be before weeks, if ever. For 64-bit XP vs 64-bit Ubuntu, clearly, it's "no". Because, right now, I have no intention to give money to MS for a 64-bit OS that I have not.  But anybody else is welcome for such a test. Cheers. Jean. |
||
|
|
|