| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: The Clean Energy Project - Phase 2 Forum
Thread: Poor crunching on this project |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 156
|
|
| Author |
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
New policy. Plz see this post http://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,30182 that was put as a sticky to top of the CEP2 forum. Under ''custom'' in the device profile you can override the "1 per device in progress". It's near the bottom.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges: 
|
I7 with 6GB = 95%-97% This suggests the i7 loses less time during I/O, but why?Q6600 with 2GB = 90-92% Yes, more RAM, but is this a Triple Channel vs Dual Channel RAM comparison, are they both x64 and what are their timings? Your i7 could be faster because it uses triple channel RAM, it has more RAM, the RAM is faster, or more likely a combination of 2 or 3 of these. The RAM could be clocked back, and 2GB / 4GB could be pulled out to test dual vs dual and amounts of RAM. |
||
|
|
Rarusu
Advanced Cruncher Germany Joined: Feb 7, 2006 Post Count: 64 Status: Offline Project Badges: 
|
I think you should also think of some other parts that may have an effect on the performance. There only needs to be one tiny little service program running in the background which is reading from or writing to the disc from time to time and the whole performance of CEP2 with its enormous I/O rates is falling back. That's my experience so far and I don't think that both of the systems are totally 100% similar except for the hardware
----------------------------------------
Cheers,
----------------------------------------Rarusu  [Edit 1 times, last edit by Rarusu at Oct 29, 2010 8:47:09 PM] |
||
|
|
X-Files 27
Senior Cruncher Canada Joined: May 21, 2007 Post Count: 391 Status: Offline Project Badges:
|
BTW, anyone tried running CEP2 on RAID0 config?
----------------------------------------  |
||
|
|
mikaok
Senior Cruncher Finland Joined: Aug 8, 2006 Post Count: 489 Status: Offline Project Badges: 
|
Under ''custom'' in the device profile you can override the "1 per device in progress". It's near the bottom. I knew that there would be an easy answer, but somehow didn't spot that. But thanks and everything running ok for now.
to infinity and beyond
 |
||
|
|
X-Files 27
Senior Cruncher Canada Joined: May 21, 2007 Post Count: 391 Status: Offline Project Badges:
|
After upgrading to 2.6.36 kernel and previous config(original ubuntu server kernel) carried over plus changing timer frequency to 1000 from 100 and processor type to Core 2/newer Xeon from generic X86-64, here's the results:
----------------------------------------q6600 with 2gb previous: 90% nice to: CPU %user %nice %system %iowait %steal %idle Average: all 0.01 94.61 4.99 0.11 0 0.28 i7 920 with 6gb previous: 95% nice to: CPU %user %nice %system %iowait %steal %idle Average: all 0 96.79 2.87 0.06 0 0.28 not much in i7 but with q6600 its a big improvement On Virtual Machine with 2GB and set to use 4 core/thread host is: i7 920 with 6gb CPU %user %nice %system %iowait %steal %idle Average: all 0.01 96.09 3.56 0.07 0 0.27 not much of a difference from a full blown linux. But what I like most on the VM is that I can suspend the session and resume without losing any progress at all. How I wish this kind of process can be implemented to the science app or to boinc itself. [Edit 3 times, last edit by X-Files 27 at Oct 31, 2010 8:21:50 PM] |
||
|
|
codes
Advanced Cruncher Joined: Oct 20, 2009 Post Count: 142 Status: Offline |
How about swap/paging statistics? sar -W
----------------------------------------How about and disk I/O activity? sar -dp How about memory utilization? sar -r I've reduced my RAM from 2 GB to 1.5 GB to see what difference that makes. So far, my sysstat data points are a little higher. Will try 1 GB in a few days, that should result in a higher disk I/O activity and lower CPU processing time I suspect. Must be a sweet spot to get optimum processing/crunching. [Edit 1 times, last edit by codes at Nov 1, 2010 1:32:34 AM] |
||
|
|
X-Files 27
Senior Cruncher Canada Joined: May 21, 2007 Post Count: 391 Status: Offline Project Badges:
|
q6600
----------------------------------------pswpin/s pswpout/s Average: 0.00 0.00 DEV tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util Average: sda 8.95 55.93 5406.72 610.45 2.30 257.13 5.31 4.75 kbmemfree kbmemused %memused kbbuffers kbcached kbcommit %commit Average: 360251 1695857 82.48 13965 1122023 1083567 13.41 i7 920 pswpin/s pswpout/s Average: 0.00 0.00 DEV tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util Average: sda 14.14 29.66 8977.03 637.04 3.06 216.43 4.31 6.09 kbmemfree kbmemused %memused kbbuffers kbcached kbcommit %commit Average: 1420580 4696900 76.78 44689 3509750 2080860 12.96 VM pswpin/s pswpout/s Average: 0.00 0.00 DEV tps rd_sec/s wr_sec/s avgrq-sz avgqu-sz await svctm %util Average: sda 6.92 14.95 3825.12 554.80 0.66 95.66 2.52 1.74 kbmemfree kbmemused %memused kbbuffers kbcached kbcommit %commit Average: 87516 1969100 95.74 37084 1328592 1163012 34.24 AWAIT on VM is 95 - this is great. [ot]btw, how to align tables?[/ot] my next test is to compile with cflags, any thoughts? [Edit 1 times, last edit by X-Files 27 at Nov 1, 2010 2:51:11 AM] |
||
|
|
codes
Advanced Cruncher Joined: Oct 20, 2009 Post Count: 142 Status: Offline |
btw, how to align tables? Highlight your text and click the "Code" button.my next test is to compile with cflags, any thoughts? Always a good idea to optimize software if you can. I'm running Slackware with the kernel Patrick said he uses -mcpu=i686 for optimization. I have not modified or recompiled any software on the device I'm running. It's pretty much stock like when it came of the install DVD + security patches.Your performance data points look very good, doesn't look like they're under a heavy I/O load. For best performance what I'm discovering on my device: no swap used, low %idle, low %iowait, low await, low %util, lower %memused than BOINC maximum % memory use when computer in use setting, and as already pointed out by other crunchers - more RAM is better. I'm experimenting with RAM to see what my sweet spot is. At what amount of RAM does the CEP2 CPU time performance go down. So far 1.5 GB and greater is good for my setup. |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
Nice work X-Files 27
 |
||
|
|
|