| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Discrepancy in then and now Dhrystone benchmarks |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 6
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Last August I was testing boinc under Win XP Pro and under Puppy Linux to see if Linux might be faster. I believe I was running ver. 5.8.16, but not certain. The machine is a 2.8GHz P4 w/Hyperthreading, 1GB 400MHz FSB DDR, 40GB HD, Intel 865GLC MB. Here are those benchmarks copied from the message log in boinc:
Under Windows XP Pro: Benchmark results: Number of CPUs: 2 8/20/2007 9:57:28 AM 1240 floating point MIPS (Whetstone) per CPU 8/20/2007 9:57:28 AM 1145 integer MIPS (Dhrystone) per CPU Under Puppy Linux: Benchmark results: Number of CPUs: 1 Mon 20 Aug 2007 10:19:05 AM PDT 1299 floating point MIPS (Whetstone) per CPU Mon 20 Aug 2007 10:19:05 AM PDT 2213 integer MIPS (Dhrystone) per CPU Since installing ver. 5.10.22 and now 5.10.30, I am getting very different Dhrystone results: 12/1/2007 6:18:17 AM Benchmark results: 12/1/2007 6:18:17 AM Number of CPUs: 2 12/1/2007 6:18:17 AM 1275 floating point MIPS (Whetstone) per CPU 12/1/2007 6:18:17 AM 2572 integer MIPS (Dhrystone) per CPU Here is the initial message string upon starting 5.10.30 this morning showing system info details: 12/1/2007 6:17:45 AM Starting BOINC client version 5.10.30 for windows_intelx86 12/1/2007 6:17:45 AM log flags: task, file_xfer, sched_ops 12/1/2007 6:17:45 AM Libraries: libcurl/7.17.1 OpenSSL/0.9.8e zlib/1.2.3 12/1/2007 6:17:45 AM Data directory: C:\Program Files\BOINC 12/1/2007 6:17:45 AM Processor: 2 GenuineIntel Intel(R) Pentium(R) 4 CPU 2.80GHz [x86 Family 15 Model 2 Stepping 9] 12/1/2007 6:17:45 AM Processor features: fpu tsc sse sse2 mmx 12/1/2007 6:17:45 AM OS: Microsoft Windows XP: Professional Edition, Service Pack 2, (05.01.2600.00) 12/1/2007 6:17:45 AM Memory: 1006.73 MB physical, 2.37 GB virtual 12/1/2007 6:17:45 AM Disk: 37.26 GB total, 26.13 GB free 12/1/2007 6:17:45 AM Local time is UTC -8 hours 12/1/2007 6:17:45 AM Version change (5.10.22 -> 5.10.30) 12/1/2007 6:17:45 AM World Community Grid URL: http://www.worldcommunitygrid.org/; Computer ID: 251466; location: (none); project prefs: default 12/1/2007 6:17:45 AM General prefs: from World Community Grid (last modified 06-Oct-2007 01:29:22) I am wondering if the inflated benchmark is causing my calculated claimed credits to be higher than they should be. Many, but not all, work units are validated with the granted credits some 33% to 40% lower than the claimed value. Additionally, it appears that I am experiencing a higher number of errors with zero credits granted under v.5.10; 3 in the last 1 1/2 weeks (2 for HCC, 1 for HPF) alone. I don't think I was getting even half that many fatal runs before.  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
bump. you could at least acknowledge my post... I'm not going to WASTE MY TIME AND ELECTRICITY on your bogus project if you ignore me
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I believe the benchmark code was improved in one of the recent major BOINC releases - or, to be precise, the compilation options were changed so that the benchmark is more consistent across platforms.
The WCG credit calculation now adjusts benchmarks anyway, so inaccurate benchmarks are no longer the problem they were. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello becca_cet,
You are running across a well known peculiarity of the Pentium 4. Hyperthreading is a good idea, but it can give really peculiar benchmark results. You only have 1 FPU, so your Whetstone results do not vary much. But you can run 2 threads through the integer pipeline. The total output is usually about 15% higher than 1 thread only, which is why hyperthreading is a good idea. But you have no guarantee that the Dhrystone integer benchmark thread will get the maximum use of the CPU chip. Some benchmark runs will only show a percentage of the true Pentium 4 Dhrystone capability. I personally recommend that members with Pentium 4 computers set the BOINC profile to only run 1 process per computer (by default it will run 2 processes) but leave hyperthreading on so that OS processes can run concurrently. Lawrence |
||
|
|
MD-Crusher
Advanced Cruncher Joined: Jul 23, 2007 Post Count: 52 Status: Offline |
becca_cet wrote:
bump. you could at least acknowledge my post... I'm not going to WASTE MY TIME AND ELECTRICITY on your bogus project if you ignore me Share with us what you know. The members really like to get to understand where this came from! |
||
|
|
KLiK
Master Cruncher Croatia Joined: Nov 13, 2006 Post Count: 3108 Status: Offline Project Badges: 
|
there is some notes that (beta) SP3 is 10% faster than the current SP2...so either wait for that or use more stable Linux!
---------------------------------------- |
||
|
|
|