| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Processor Optimisation - Sandybridge vs Ivy Bridge |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 25
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I just bought myself a 3770k a few days ago, and put BOINC on it recently, and while I got the expected performance boost over my i7 M620, I noticed something funny:
Floating point benchmark was only about 10% faster, but integer performance was nearly 100% faster. From this, I wondered if some projects gain more than others. Here are some examples: Project/Old/New/Gain DSFL/512 minutes/239 minutes/114% Gain C4CW/270 minutes/139 minutes/94% Gain GFAM/544 minutes/156minutes/248% Gain FAAH/376 minutes/191 minutes/96% Gain HPFP2/602 minutes/277 minutes/117% Gain Is this a "known issue", and is it reliable enough to take advantage of? So for example, knowing that I get 248% gain on GFAM, I could potentially contribute more by doing nothing beside running GFAM on my 3770k, and not allowing GFAM to run on M620. Is this a fair comment? |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7849 Status: Offline Project Badges: 
|
Two things come immediately to mind:
----------------------------------------1. Use a bigger sample and see if the comparisons hold up. 2. If you want to use the new machine where it is most efficient, that is entirely your choice, but it makes sense to me, given the caveat that the project also meets with your desires. Crunch on. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Some projects have WUs of variable running time on the same machine, GFAM is one of them. So using just one WU on each of your machines is not representative enough. On the other hand, C4CW WUs running times have little variance, so maybe that 94% gain is more believable.
Keep in mind that there are other factors beside the CPU that may affect this comparison: OS, memory, other running processes, just to name a few. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Some projects have WUs of variable running time on the same machine, GFAM is one of them. So using just one WU on each of your machines is not representative enough. On the other hand, C4CW WUs running times have little variance, so maybe that 94% gain is more believable. Keep in mind that there are other factors beside the CPU that may affect this comparison: OS, memory, other running processes, just to name a few. Ah, okay... That makes more sense, and downloading a few more units, I can see that they're not all the same size. I guess in that case, it's going to be very hard to select the right WUs for the right processor, though I'm convinced by benchmarking alone, that some WUs should in theory be more efficient than others. (Those that are integer heavy should run better on the 3770k) If any advisors can share which programs that might be, it would be worth doing processor assignment optimisation, otherwise, I might just run everything on every computer. |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7849 Status: Offline Project Badges:
|
I believe HCC is integer based.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
It's a dream of the techs to be able to real time analyze what runs best on each device and get the volunteer's OK to assign whatever it's best at [either in throughput, or in sheer heaviness], so for instance long or high IO jobs go to the Bridge fitted devices and the HCC go to the centrinos. If there is a disproportion in different class devices, than the lighter stuff could also get assigned to the heavy duties or vice versa... with the consequential grump grump grump reactions doubtlessly to follow, so opt-in, or don't opt in and if you do, stay out of meddling [probably adds another 3 options demanded by users to override and mm]. If that were only true.
--//-- |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
!
----------------------------------------[Edit 1 times, last edit by skgiven at Jul 18, 2012 8:59:43 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Another thing I noticed... I get 13k Dhrystones per CPU on Linux, but only 10k on Windows 7 SP1 :(
I could dual boot to Linux when not using it, but I like being able to just use my computer whenever, and not wait for a reboot. Oh well. |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges: 
|
Anyone ever try running Linux in a virtual machine on win 7 64? Am set up dual boot but need to run win7 for other things. Just curious how the benchmarks would compare.
----------------------------------------
Crunching for humanity since 2007!
 |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7849 Status: Offline Project Badges:
|
Anyone ever try running Linux in a virtual machine on win 7 64? Am set up dual boot but need to run win7 for other things. Just curious how the benchmarks would compare. Why not run win 7 in a virtual machine under linux ? Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
|