| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: Work Unit benchmarks for CPUs |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 5
|
|
| Author |
|
|
daichi
Cruncher Joined: May 24, 2006 Post Count: 9 Status: Offline Project Badges: 
|
Hello,
Let me start by saying how refreshing it is to find a knowledgeable group of members who are dedicated to humanitarian grid computing. I have a question on benchmarks for WCG work units between different processes, speeds and operating systems. I see processor benchmarks from Intel and AMD, but have wondered how they apply to WU processing. Does such information exist? This information would help me determine the optimal server configuration I might dedicate to processing work units at the best throughput rate. These would be stock vendor configurations. Perhaps in the format like: OS = Windows FAAH Time to Completion HDC Time to Completion Opteron 256 Dual Core Opteron 270 Dual Core Opteron 275 Dual Core Xeon 5050 Dual Core Xeon 5110 Dual Core Intel Core 2 [blah] P4 etc. And the same for OS=Linux. This would all be using the BOINC client. I've seen similar posts in this forum about P4 processors or specific user configurations, but not a consolidated view … especially with the new dual-core processors. Any information would be appreciated. Regards, --] Eric |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Good idea.
Unfortunately, it is very difficult to benchmark in the way you describe. Each project is different, and each work unit is different. However, you can get a good approximation by using the Whetstone benchmark. This is the benchmark BOINC uses for awarding points, and since most of the projects use floating point operations heavily, it should be enough for your purpose. You should be able to find published tables of Whetstone benchmarks for the systems you describe. In theory, there will be no difference between Windows and Linux. This hasn't stopped people from claiming that one or the other is better.... |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello Ekodak.
100% agree with Didactylos. I might add that I have seen, in house, benchmarking for similar FP intensive work and the Core 2 Duo is a 'lay down misere' There simply isn't anything close, save the new Macs. You said the OS was to be Windows or Linux so it has to be C2D. Cheers. ozylynx  |
||
|
|
daichi
Cruncher Joined: May 24, 2006 Post Count: 9 Status: Offline Project Badges:
|
Thanks guys. I guess I'll just have to get some evaluation boxes and run the BOINC CPU benchmark on them and see what they turn up.
I was able to find some data for AMD Opteron processors here: http://www.amd.com/us-en/Processors/ProductIn...8_8796_8800~72730,00.html Not sure if that's the same as the Whetstone benchmarks. Can't find anything similar for the Xeon 51XX series, but have heard rumors that those processors have superior calc performance over the Opterons. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi Again.
The best I can do to help is here This link is to a privately run site keeping track of actual scores/day of various computers on the Foldin@home project. All of the information is submitted by the folders so is not varifiable and much of it wont apply to you. Having said that it can be browsed by cpu type and on average will be quite representative of the performance expected of each make represented. The numbers are meaningless in the outside world but they do have proportion and higher is better. As you are likely aware this project is 'human proteome folding exclusively and I suspect, without knowing, places similar computational requirements on the cpu to our own FAAH and folding experiments. Good luck. Cheers. ozylynx |
||
|
|
|