| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: AMD A8-3850 & AMD FX-8120 processors |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 35
|
|
| Author |
|
|
kateiacy
Veteran Cruncher USA Joined: Jan 23, 2010 Post Count: 1027 Status: Offline Project Badges: 
|
@ PPL: When you look at the numbers for the FX remember that when BOINC does it's benchmark I believe that it "suspends" all cores and then runs the benchmark process on one core. I don't think this is the case under Linux. There is a thread somewhere (I'll try to find it later today) in which benchmarks for machines with hyperthreading were compared under Windows and Linux. Here's the link to the other thread that discusses Windows vs. Linux benchmarks and hyperthreading. It includes comparative numbers for an Intel 2600K. It goes on for several messages. http://www.worldcommunitygrid.org/forums/wcg/...ad,29825_offset,70#314532  [Edit 1 times, last edit by kateiacy at Nov 24, 2011 3:16:27 PM] |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
@ PPL: When you look at the numbers for the FX remember that when BOIC does it's benchmark I believe that it "suspends" all cores and then runs the benchmark process on one core. Why is this important? The FX architecture is based on two cores per module. Each core within a module has its own integer unit but shares one 256 bit floating point unit (which I understand can be divided into two logical 128 bit units) The fact that the BOINC CPU baseline is run on only one core goes a long way towards explaining why FP performance on the test appeared to be better than expected when compared to integer performance. Then BOINC benchmarks, it runs 1 instance of the integer-benchmark but for the floating-point-benchmark it runs N instances there N equals number of usabe cores. Old BOINC-clients also ran N instances of the integer-benchmark, but this gave results "all over the place", so it was changed to only running one instance.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Minor correction: The BOINC benchmark runs one instance of the integer benchmark on Windows, but N instances on Linux and OS X.
----------------------------------------Source: The BOINC source code, where the relevant part is #ifdef _WIN32So: * whetstone / float: Runs on N cores, where N is the number of cores available to BOINC. Same for all OS. * dhrystone / int: Runs on one core on Windows, N cores on Linux and OS X. The final score is averaged over all cores that were used in the benchmark, so the scores should be comparable between different OSes when all cores have roughly the same performance. But when the cores have very different performance characteristics, e.g. because virtual cores are included, the integer scores from Windows are no longer comparable to integer scores from Linux. [Edit 1 times, last edit by Former Member at Nov 24, 2011 10:45:25 PM] |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges:
|
I don't think it runs benchmark test on one core. I can see all cores I assign to crunching peg for both benchmarks on my xeons.
----------------------------------------
Crunching for humanity since 2007!
 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
@Bearcat Are your Xeons running Windows or Linux?
On my Windows 7 box (i7, 4 cores/8 threads), I get one 100% CPU spike (all cores) for the floating point benchmark, and a 12% CPU (one core) spike for the integer benchmark. |
||
|
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges:
|
Just did a benchmark on my win7 64 (dual hex) and ubuntu 11.4 64 (hex). I was going off my system monitor icon I have on top of the screen to monitor (ubuntu). I opened it up to view all cores on the graph. Both tests showed all assigned cores I crunch with (11) active, but not pegged. On my windows PC, all assigned cores (22) are active on floating point, but on integer only one thread.
----------------------------------------
Crunching for humanity since 2007!
----------------------------------------[Edit 1 times, last edit by Bearcat at Nov 25, 2011 4:11:00 AM] |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges:
|
Minor correction: The BOINC benchmark runs one instance of the integer benchmark on Windows, but N instances on Linux and OS X. Ah yes, since I'm only running windows didn't remember the change to integer-benchmark was only for windows-computers. In any case, the BOINC-benchmark isn't a good indicator of performance, using actual science-applications is a much better way to measure this. Of course, not all science-applications gives a useful measure, example Rice while running was a terrible benchmark since all tasks took N hours regardless of computer (there N was changed a couple times during the projects lifetime). Now, the best benchmark is always to re-run the exact same tasks on different computers, but as long as there's small variations between tasks it's possible to average over many tasks instead. "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges:
|
Hi all. I found a FX 8150 running on P.O.E.M. looks to be stock clocked i believe that's 3.6Ghz, & the O.S. is linux 3.2.0-rc2+ so that takes windows out of the equation. ( mine is Ubuntu 10.04lts x64 ) Ran some POEM over night here's the average per task. poempp_barrelmutations__4,770 sec - my 1055T @ 2.8Ghz poempp_barrelmutations__4,920 sec - FX 8150 @ 3.6Ghz - guessing here. FX 8150 POEM. Closer then the FX 8120 to my 1055T running on DOCKING but is still slower! very disappointing. Hmm, going by these numbers, the 1055T crunches 108.7 Poems/day while the FX-8150 crunches 140.5 Poems/day meaning the FX-8150 is producing 29.3% more Poem-tasks per day than the 1055T. If takes the clock-speed into account, the FX-8150 is clocked 28.6% higher so based on these numbers MHz for MHz there's no difference in speed between 1055T and FX-8150 running Poem. BTW, one interesting thing to look at is how many physical FPU's the FX-8150 really has. If takes this into account, the 1055T with 6 FPU's crunches 18.1 Poems/day/FPU while FX-8150 crunches 35.1 Poems/day/FPU meaning 93.9% more per FPU. Even if takes into account the higher clock-speed this still means the FX-8150's FPU is 50% faster per MHz running Poem than the 1055T-FPU is. "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
be warned most of the bulldozer 8core chips run very warm...
only thing about it is 8 cores the 1100t oced can blow it away.. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hey Chris.
----------------------------------------If your running any H.C.C. tasks i'd be interested to see the run times from the 8120, i've got my x6 running a few to extra to try and get to 365days before it's finished. Stock of coarse,  not fair otherwise.Edit/ Here's a quick 15 just finished. X0000129421093201108311019_ 1-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 06:56:31 1.57 39.5 / 0.0 X0000129421354201108311015_ 0-- Black-AMD-X6 Valid 11/27/11 00:34:17 11/27/11 02:43:01 1.62 41.2 / 36.5 X0000129421356201108311015_ 1-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 02:43:01 1.49 37.7 / 0.0 X0000129421353201108311015_ 1-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 04:03:09 1.28 32.4 / 0.0 X0000129421339201108311016_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 04:03:09 1.47 37.3 / 0.0 X0000129421337201108311016_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 04:03:09 1.32 33.5 / 0.0 X0000129421325201108311016_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 04:03:09 1.55 39.2 / 0.0 X0000129421322201108311016_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 04:03:09 1.53 38.7 / 0.0 X0000129421318201108311016_ 1-- Black-AMD-X6 Valid 11/27/11 00:34:17 11/27/11 05:17:47 1.50 38.0 / 21.8 X0000129421309201108311015_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 05:17:47 1.59 40.3 / 0.0 X0000129421273201108311016_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 05:42:25 1.51 38.1 / 0.0 X0000129421256201108311016_ 1-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 05:42:25 1.42 36.0 / 0.0 X0000129421206201108311017_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 06:52:29 1.69 42.7 / 0.0 X0000129421164201108311018_ 0-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 06:52:29 1.51 38.2 / 0.0 X0000129421357201108311015_ 1-- Black-AMD-X6 Pending Validation 11/27/11 00:34:17 11/27/11 02:43:01 1.56 39.5 / 0.0 Average is 1.5073 per task. [Edit 5 times, last edit by Former Member at Nov 27, 2011 7:18:04 AM] |
||
|
|
|