| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Benchmarks in different versions |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 3
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I thought I'd try out the new version. As the one on this site doesn't work on my machines, I downloaded 5.10.45 command-line-only version from http://boinc.berkeley.edu/download_all.php.
The benchmarks are reporting about 65% of the floating point values and 80% of the integer values as compared to 5.8.15, which of course leads to massive under-claiming. Are there any known differences to account for this? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi Kremmen,
Sounds like you are running Linux. The benchmarks have not changed for decades. However, back in the 1990s the Windows compiler writers added in some optimizations that eliminated some of the loops in the benchmarks. This let them claim faster benchmark scores for their compilers (commercial competition - - that is to say: cheating). A few years back there was a big fuss about the difference in benchmark scores between Windows and Linux. I don't know the details about how it finally was resolved, but I am prepared to believe in different versions of BOINC for Linux giving big benchmark differences. This is probably a question best asked on the BOINC message boards. Lawrence |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Pass. You may want to ask at Berkeley.... maybe compiled with different libs?
----------------------------------------Do know that if i run the daemon at a higher priority that the test will produce better and quite consistent values for the Dhrystone part in particular.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|