| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: CPU Benchmarks (what does it mean/where do I rank) |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 71
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Here is my Compaq Presario with AMD Sempron 3100+ (1.81 Ghz) running BOINC 5.4.11 on Windows XP:
1/11/2007 7:06:45 PM|| 1708 floating point MIPS (Whetstone) per CPU 1/11/2007 7:06:45 PM|| 3140 integer MIPS (Dhrystone) per CPU |
||
|
|
cio_redulla
Advanced Cruncher Philippines Joined: Apr 24, 2006 Post Count: 130 Status: Offline Project Badges: 
|
Hello Lawrence,
----------------------------------------Do the values of the Whetstone and Dhrystone benchmarks affect the time needed to crunch a work unit? (Assuming that the BOINC Agent being used is not one of those 'tweaked' agents)  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello cio_redulla,
Yes. The higher the value, the faster the processor. The Whetstone benchmark measures the speed of the floating point instructions. The Dhrystone benchmark measures the speed of the integer instructions. BOINC points are awarded by adding the 2 benchmarks together, then dividing by ?400?, then multiplying the CPU time it takes to process a work unit. The problem is that the Windows compiler writers, competing with each other, put some optimizations into their code to ignore some instructions that do not produce any result outside of the program. This enabled them to point to high benchmark scores and claim that their compilers were highly efficient but has reduced the usefulness of standard benchmarks. THe Linux compiler leaves all the instructions in, giving a true (and lower) benchmark score for a processor. Lawrence |
||
|
|
cio_redulla
Advanced Cruncher Philippines Joined: Apr 24, 2006 Post Count: 130 Status: Offline Project Badges:
|
My CPU usually claims a lower credit than what is really granted. Take the following (actual) workunit as an example:
----------------------------------------faah1170_ d245n120_ x1MEU_ 01-- Valid 01/07/2007 02:07:26 01/07/2007 12:16:18 5.49 73 / 79 faah1170_ d245n120_ x1MEU_ 01-- Valid 01/07/2007 02:02:48 01/07/2007 18:32:39 11.13 85 / 79 faah1170_ d245n120_ x1MEU_ 01-- Valid 01/07/2007 02:01:40 01/08/2007 00:06:51 7.59 57 / 79 My CPU crunched the third unit in the list. Does it mean that the first CPU in this list is faster? I think so because it claimed a higher credit even though it took the CPU less time to process a unit. Also, it is obvious that the CPU took a shorter time to finish the unit compared to the other CPUs. That makes the CPU in the middle the slowest of the three, :-). |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
G'Day cio_redulla
Yes your machine was in the middle of the time taken for this work unit. The first machine in the quorum is significantly faster than yours and the second machine is a slow one. It is interesting that the first machine claimed as many points as it did, due to the fact that very fast modern processors usually outperform their benchmarks. i.e. they complete tasks quicker than expected and therefore tend to have low claims. The example is unusual. Slow, struggling machines, such as the second in the list, often under perform, for many reasons including marginal ram or cpu power and often take longer than expected to complete a task thus claiming high points. If your machine is claiming low, as the norm, and is not a speed demon, it would suggest that something has changed since your last benchmark. Perhaps it was unusually busy or running a program which might have chronic memory leakage when it was benchmarked but doesn't run that software in normal use. Try shutting down all unnecessary programs and force a new benchmark to see how you go with it. Cheers. ozylynx  |
||
|
|
cio_redulla
Advanced Cruncher Philippines Joined: Apr 24, 2006 Post Count: 130 Status: Offline Project Badges:
|
G'Day cio_redulla Yes your machine was in the middle of the time taken for this work unit. The first machine in the quorum is significantly faster than yours and the second machine is a slow one. It is interesting that the first machine claimed as many points as it did, due to the fact that very fast modern processors usually outperform their benchmarks. i.e. they complete tasks quicker than expected and therefore tend to have low claims. The example is unusual. Slow, struggling machines, such as the second in the list, often under perform, for many reasons including marginal ram or cpu power and often take longer than expected to complete a task thus claiming high points. If your machine is claiming low, as the norm, and is not a speed demon, it would suggest that something has changed since your last benchmark. Perhaps it was unusually busy or running a program which might have chronic memory leakage when it was benchmarked but doesn't run that software in normal use. Try shutting down all unnecessary programs and force a new benchmark to see how you go with it. Cheers. ozylynx Hello ozylynx, It terms of clock speed, my machine is really fast (Pentium D, 3.4GHz). But I hate my machine when it comes to crunching. It is easily outrun by AMD CPUs with lower clock speeds. My benchmarks are relatively low (1700 Whetstone and 2750 Dhrystone per cpu). It hovers between these values everytime I benchmark my machine so I guess that's how fast my machine crunches. The first machine in the previous example may have very large benchmark values that's why it claims a larger credit than mine even though it finished the unit in such a short time. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Not sure if it goes for all projects of WCG, but certainly one of them (think its HDC) on interruption and restart would skip back to the last checkpoint, but not loose the CPU time. It can result in inflated hours thus 'high' claim, as the benchmark is still the same.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
cio_redulla
Advanced Cruncher Philippines Joined: Apr 24, 2006 Post Count: 130 Status: Offline Project Badges:
|
Here's my AMD X2 4600+ @ 2.74Ghz results: 2612 floating point MIPS (whetstone) per CPU 4874 integer MIPS (Dhrystone) per CPU Cant understand why they are so different tho... My AMD 64 3200+ (standard clock) 06/01/2007 16:58:22||Benchmark results: 06/01/2007 16:58:22|| Number of CPUs: 1 06/01/2007 16:58:22|| 1853 floating point MIPS (Whetstone) per CPU 06/01/2007 16:58:22|| 3450 integer MIPS (Dhrystone) per CPU 06/01/2007 16:58:22||Finished CPU benchmarks My Intel Conroe E6600 @ 3.24Ghz Number of CPUs: 2 2991 floating point MIPS (Whetstone) per CPU 6247 integer MIPS (Dhrystone) per CPU Managed to get the Intel to post and boot into windows (and its fairly stable at that the speed too) 3.61Ghz which I'm very pleased at. I just needed better cooling to run it at full load (each core) as it was rather warm at that speed with the retail heatsink... Results to follow when I get that sorted  Hello Phill23, I couldn't overclock my machine because I'm using an Intel board. Is there any other way to make it go faster? I tried using softwares to change the clock speed but it won't run on my system because the PLL of the motherboard is not yet supported. I guess I have to try those other boards, :-). |
||
|
|
cio_redulla
Advanced Cruncher Philippines Joined: Apr 24, 2006 Post Count: 130 Status: Offline Project Badges:
|
G'Day cio_redulla Yes your machine was in the middle of the time taken for this work unit. The first machine in the quorum is significantly faster than yours and the second machine is a slow one. It is interesting that the first machine claimed as many points as it did, due to the fact that very fast modern processors usually outperform their benchmarks. i.e. they complete tasks quicker than expected and therefore tend to have low claims. The example is unusual. Slow, struggling machines, such as the second in the list, often under perform, for many reasons including marginal ram or cpu power and often take longer than expected to complete a task thus claiming high points. If your machine is claiming low, as the norm, and is not a speed demon, it would suggest that something has changed since your last benchmark. Perhaps it was unusually busy or running a program which might have chronic memory leakage when it was benchmarked but doesn't run that software in normal use. Try shutting down all unnecessary programs and force a new benchmark to see how you go with it. Cheers. ozylynx I was able to obtain the following benchmark after I 'overclocked' my machine: BOINC client version 5.4.11 for windows_intelx86 (no tweaks, whatsoever) 1/18/2007 8:38:30 AM||Suspending network activity - running CPU benchmarks 1/18/2007 8:38:32 AM||Running CPU benchmarks 1/18/2007 8:39:31 AM||Benchmark results: 1/18/2007 8:39:31 AM|| Number of CPUs: 2 1/18/2007 8:39:31 AM|| 2052 floating point MIPS (Whetstone) per CPU 1/18/2007 8:39:31 AM|| 3044 integer MIPS (Dhrystone) per CPU 1/18/2007 8:39:31 AM||Finished CPU benchmarks 1/18/2007 8:39:32 AM||Resuming computation A nice improvement from my previous benchmark, :-). Keep crunching! |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
G'Day
cio_redulla, that's interesting. Have you any idea what caused the difference? Was it a 'nasty' program in the background and if so which one? Cheers. ozylynx |
||
|
|
|