| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Mapping Cancer Markers Forum
Thread: Haswell is faster |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 26
|
|
| Author |
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges: 
|
This reminds me of something I wish WCG would add. I wish they'd have a store of well coded WU's from each project that people could run on their systems to benchmark. That would be nice, but I use WCGDAWS, which gives me some nice averages across machines and by project if I want to do it that way. Cheers Right, but different people/rigs doing the same tasks would allow for a more accurate comparison and would not depend on batch variations in WU's for a particular project. You are absolutely right, but I do not know of any place that would be available, so I tried to suggest the next best thing. Perhaps some ambitious BOINC programmer could make a 50 or 100 unit benchmark set of WU. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
ryan222h
Senior Cruncher Joined: Sep 4, 2006 Post Count: 425 Status: Offline |
I would think one "sample" work unit for each project would be a great way to compare system performance.
----------------------------------------Since it would be an exact duplicate every time it should give very accurate results. Not sure if its possible to implement such a thing.  |
||
|
|
yojimbo197
Advanced Cruncher Joined: Jun 30, 2012 Post Count: 83 Status: Offline Project Badges: 
|
I would think one "sample" work unit for each project would be a great way to compare system performance. Since it would be an exact duplicate every time it should give very accurate results. Not sure if its possible to implement such a thing. That's what I had in mind, though I was thinking about enough WU's from a particular project to generate accurate statistics. So you'd need more than just one WU.  |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges:
|
I decide to give a sample of the results of my systems running MCM1 and let the reader draw their own conclusions on what it may mean. Just as an example of how to rate systems.
----------------------------------------Specs Processors(Cores) OS Avg. Points/Hr Total # of WU Avg. CPU Time Intel(R) Xeon(R) CPU E5410 @ 2.33 GHz 2(8) Linux 64 bit 21.4348 787 5.397 Intel(R) Core(tm)2 Quad CPU Q6600 @ 2.40 GHz 1(4) Windows Vista 32 bit 19.3605 477 5.585 AMD Phenom(tm) 9150e Quad-Core Processor 1(4) Windows 7 Ultimate 64 bit 12.0287 238 8.782 Intel(R) Xeon(R) CPU E5320 @ 1.86GHz 2(8) Linux 64 bit 16.5428 1156 6.460 Intel(R) Core(tm)2 CPU 6600 @ 2.40GHz 1(2) Linux 32 bit 21.3474 383 5.015 Intel(R) Core(tm)2 CPU 6600 @ 2.40GHz 1(2) Linux 32 bit 20.6421 376 5.232 Intel(R) Core(tm)2 CPU 6600 @ 2.40GHz 1(2) Linux 64 bit 20.9611 357 5.476 I can not get it to line up properly but I think you can follow the top legend. The bottom three systems are all identical except for the OS on the the last one. So even identical systems with identical OS's can vary a bit. Cheers
Sgt. Joe
----------------------------------------*Minnesota Crunchers* [Edit 6 times, last edit by Sgt.Joe at Jan 20, 2014 8:35:36 PM] |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
Jim thanks for shearing your findings. Out of interest have you found any increase in one times. As of 8 January overall run-time was around 5.59 hours per task. At the moment, they seem to be decreasing for me, but they go up and down. An easy way to get some idea is by using the "Time Left" column in Boinc Tasks. That averages over some time period to get its estimate, though I don't know how long. Of course the anticipated difficulty of the work unit enters into it also in some manner, but the estimates at present are in the range of from 3 hours 30 minutes to 3 hours 47 minutes. (And thanks; the days can add up fast now that I can place four Haswell cores on it 24/7, since I have to leave this machine on for other purposes anyway.) [Edit 2 times, last edit by Jim1348 at Jan 13, 2014 12:09:58 AM] |
||
|
|
Jack007
Master Cruncher CANADA Joined: Feb 25, 2005 Post Count: 1604 Status: Offline Project Badges: 
|
I'm really surprised that Haswell has done that well.
----------------------------------------Intel has kinda been worried more about GPU. I have a sandy bridge and 2 ivy bridge, for the price not going to upgrade, however, broadwell is coming  more power with less power!Or the 8 core extreme cpu... that's kinda pricey, but if i call it my new gaming rig I can justify it...ish  |
||
|
|
l_mckeon
Senior Cruncher Joined: Oct 20, 2007 Post Count: 439 Status: Offline Project Badges: 
|
For MCM the percentage completed column gives the best idea of time for a task.
|
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1326 Status: Offline Project Badges:
|
Hi Jim, I am interested to hear/read what your average turnaround time is for current tasks on your Haswell machine? I am currently running batch 5877 on a first gen Intel I 7 at 3.43 GHz stock is 3.33 and the turn around time is 14 to 15 minutes
---------------------------------------- |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
Hi Jim, I am interested to hear/read what your average turnaround time is for current tasks on your Haswell machine? I am currently running batch 5877 on a first gen Intel I 7 at 3.43 GHz stock is 3.33 and the turn around time is 14 to 15 minutes I haven't gotten to the 5877 yet, but there are a few in the buffer. However for the recent 58xx series from 5832 onward, it is about 2 1/2 hours. On the last three 5865's, it has been 2 3/4 hours. I last saw short work units on the 581x series on 11 July, which ran about 45 minutes. EDIT: I have now done 10 of the 587x work units, and they are all about the same time, around 7 minutes 40 seconds. But note that that is with only 4 virtual cores running (out of 8), so they are lightly loaded otherwise, and probably act more like 4 physical cores, as noted above. [Edit 3 times, last edit by Jim1348 at Jul 14, 2014 5:07:54 PM] |
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1326 Status: Offline Project Badges:
|
Hi Jim, I am interested to hear/read what your average turnaround time is for current tasks on your Haswell machine? I am currently running batch 5877 on a first gen Intel I 7 at 3.43 GHz stock is 3.33 and the turn around time is 14 to 15 minutes I have now done 10 of the 587x work units, and they are all about the same time, around 7 minutes 40 seconds. But note that that is with only 4 virtual cores running (out of 8), so they are lightly loaded otherwise, and probably act more like 4 physical cores, as noted above. Thank you for the information Jim. I know we cannot categorically say that Haswell CPUs are twice the speed of the first gen I 7's, it gives me a good indication at how much quicker the Haswell chip is in terms of data crunching speed. Thanks again for the information. |
||
|
|
|