| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Bogus Bulldozer BOINC Benchmarks - blooking blor Badvice  |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 38
|
|
| Author |
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges: 
|
FX8150 + Buckets of DDR3-1600 RAM OC'ed to 4Ghz (Windows Server 2008 R2)
----------------------------------------1620 FP MIPS and 6702 Integer MIPS 1055T + DDR3-1066 RAM OC'ed to 3Ghz (Windows 7 64bit) 2492 FP MIPS and 7795 Integer MIPS Both on 6.10.58  What am I doing wrong? The Bulldozer spits out Work Units commensurate with it's clock rate and core count when compared to the PhenomII and always picks up the higher points from the Wingman's Work Unit so despite the bad press the FX8150 rocks I suspect the benchmark is causing the Bulldozer to underclock but try as I might I can't get those Benchmark scores any higher and they are clearly wrong. Anyone any advice? n.b. I don't need a lecture on how the FX sucks and how I should have waited for Ivybridge, that the 8 cores is really 4 etc etc etc - all of this will fall on deaf ears as for a sub £200 processor it's rocketing through Work Units  Dave  |
|||
|
||||
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7677 Status: Offline Project Badges: 
|
Just a thought, but it might be the OS.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
|||
|
||||
|
kateiacy
Veteran Cruncher USA Joined: Jan 23, 2010 Post Count: 1027 Status: Offline Project Badges:
|
Very interesting, Dave. Can you give us some comparative run times on the two machines for sciences that have very consistent times -- say C4CW?
----------------------------------------I know BOINC calculates benchmarks differently under Linux versus Windows for machines running more than 1 thread; I don't have any idea about different versions of Windows.  |
|||
|
||||
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
Sgt Joe, I know what you are saying but they are both built around the same Kernel (W7) and I don't yet have it busy doing anything else
----------------------------------------kateiacy I'll put it back on Xubuntu, after the C4CW test below, and see if there's any difference I can honestly say that it burns up WU's at a rate you would expect for a 8 core 4Ghz processor compared to the PhenomII Personally even with it's 125W TDP I think it is more efficient - but the benchmarks are not showing this Right I will have a go for a day on C4CW and let you know |
|||
|
||||
|
mikey
Veteran Cruncher Joined: May 10, 2009 Post Count: 822 Status: Offline Project Badges:
|
FX8150 + Buckets of DDR3-1600 RAM OC'ed to 4Ghz (Windows Server 2008 R2) 1620 FP MIPS and 6702 Integer MIPS 1055T + DDR3-1066 RAM OC'ed to 3Ghz (Windows 7 64bit) 2492 FP MIPS and 7795 Integer MIPS Both on 6.10.58 What am I doing wrong? The Bulldozer spits out Work Units commensurate with it's clock rate and core count when compared to the PhenomII and always picks up the higher points from the Wingman's Work Unit so despite the bad press the FX8150 rocks I suspect the benchmark is causing the Bulldozer to underclock but try as I might I can't get those Benchmark scores any higher and they are clearly wrong. Anyone any advice? n.b. I don't need a lecture on how the FX sucks and how I should have waited for Ivybridge, that the 8 cores is really 4 etc etc etc - all of this will fall on deaf ears as for a sub £200 processor it's rocketing through Work Units Dave I too have a similar setup, except I have an AMD 1090T and an FX-6100: Here are some of my 1090T valid results: cfsw_ 0529_ 00529959_ 0-- Boinc14 Valid 4/27/12 23:46:05 4/29/12 12:06:53 3.07 102.2 / 102.2 cfsw_ 0527_ 00527545_ 0-- Boinc14 Valid 4/27/12 22:55:08 4/29/12 12:05:17 3.09 103.8 / 103.8 cfsw_ 0527_ 00527023_ 0-- Boinc14 Valid 4/27/12 22:51:50 4/29/12 12:02:45 3.08 103.1 / 103.1 cfsw_ 0527_ 00527243_ 0-- Boinc14 Valid 4/27/12 22:51:33 4/29/12 11:23:42 3.10 101.1 / 101.1 Here are some of the FX-6100 results: cfsw_ 0564_ 00564451_ 0-- Backup2 Valid 4/28/12 13:15:02 4/29/12 11:57:13 4.33 102.0 / 102.0 cfsw_ 0560_ 00560892_ 0-- Backup2 Valid 4/28/12 11:50:11 4/29/12 11:40:37 4.36 101.0 / 101.0 cfsw_ 0558_ 00558730_ 0-- Backup2 Valid 4/28/12 10:49:20 4/29/12 09:19:19 4.27 97.8 / 97.8 cfsw_ 0558_ 00558996_ 0-- Backup2 Valid 4/28/12 10:49:20 4/29/12 11:25:20 4.35 99.5 / 99.5 I am running Win7 64 bit Ultimate in both and have 16GB of DDR-1333 memory nothing overclocked in each. I also have one gpu in the FX-6100 crunching, an AMD 5870 and two AMD 5770 gpu's in the 1090T crunching, they are similar in performance between the machines. Both are basically Boinc only machines. The FX is SLOWER doing the same tasks! I am running Boinc version 7.0.25 in FX and 6.12.34 in the 1090T.   |
|||
|
||||
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
first results for C4CW
----------------------------------------104 mins 1055T at 3Ghz 87 Mins FX8150 at 4Ghz same target 05 and batch 2626.... Struggling with the maths - I'm having a blind spot It's 33% faster (by clock rate) but should knock 25% off the time ..that's right isn't it?  clock for clock the fx core is 0.9 PhenomII ...so the benchmark result for a 4Ghz clocked FX core should be 2990 FP MIPS not 1620 6 core PhenomII at 95W TDP 14952FP MIPS 8 Core Bulldozer FX at 125W TDP should be 23920 MIPS 21.5% more efficient per watt than the very green version of the PhenomII I'll run C4CW a while longer [Edit 5 times, last edit by David Autumns at Apr 29, 2012 4:07:12 PM] |
|||
|
||||
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
There you go
----------------------------------------The 905E2 is the 1055T at 3Ghz (the PC started it's crunching life as a 905E)  twice as many complete WU's out of the 8150 with 2 extra cores and another Ghz of head room per core and that's with the 1055T having a head start That Benchmark is hardly representative still not sure why the 8150FX is asking for more points with the slower Benchmark result The plot thickens |
|||
|
||||
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
FX8150 + Buckets of DDR3-1600 RAM OC'ed to 4Ghz (Windows Server 2008 R2)
----------------------------------------1620 FP MIPS and 6702 Integer MIPS exactly the same hardware on Xubuntu 2542 FP MIPS and 11236 Integer MIPS How bizarre ! I wonder what I would get if I paid for another Windows 7  We'll never know..... [Edit 1 times, last edit by David Autumns at Apr 29, 2012 5:52:09 PM] |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
WOW that's a HUGE difference. Interesting.
|
|||
|
||||
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
5pot it is !
----------------------------------------But looking at how many points the FX8150 asked for for each result I don't think the points returned will be suffering as a result OK OK I will go and run some C4CW on Xubuntu and see what happens....... No rest for the wicked  |
|||
|
||||
|
|