Linux vs Windows

meetme
meetme
Joined: 9 Dec 05
Posts: 1
Credit: 2,667
RAC: 0
Topic 190383

Hi,

I just compared one of my results with the result of another user (same workunit, same CPU type)

My machine..
CPU type GenuineIntel
Intel(R) Pentium(R) 4 CPU 1.80GHz
Number of CPUs 1
Operating System Linux
Memory 1011.32 MB
Cache 256 KB
Measured floating point speed 464.43 million ops/sec
Measured integer speed 1114.83 million ops/sec
CPU time 45029.153521
Claimed credit 41.1532953835418

The other machine...
CPU type GenuineIntel
Intel(R) Pentium(R) 4 CPU 1.80GHz
Number of CPUs 1
Operating System Microsoft Windows XP
Professional Edition
Memory 126.48 MB
Cache 976.56 KB
Measured floating point speed 933.55 million ops/sec
Measured integer speed 1795.02 million ops/sec
CPU time 46966.0625
Claimed credit 74.1610612229932

As I said...it's the same workunit, same CPU type...my computer was even faster than the other but I got much fewer claimed credit...?!?!

Hey...that's not fair...Claimed credit depends on the benchmark results...but this benchmark doesn't work correct on linux machines...I would suggest to fix this problem or change the formular for claimed credits...:(

Michael Roycraft
Michael Roycraft
Joined: 10 Mar 05
Posts: 846
Credit: 157,718
RAC: 0

Linux vs Windows

meetme,

First, Welcome to Einstein, and thank you for contributing to a worthy science research!

Quote:
Hey...that's not fair...Claimed credit depends on the benchmark results...but this benchmark doesn't work correct on linux machines.

Not fair? You are absolutely right. This is NOT an Einstein problem. The Einstein app only does the crunching, and as you have noted, the app crunches quicker than does the Windows version on a similar machine.

The reason for your Linux' lower credit claims lies with the BOINC client, which does the benchmarking (note the discrepancy there) and crediting. Claimed credit is directly proportional to crunchtime times the sum of the benchmarks.

Quote:
..I would suggest to fix this problem or change the formula for claimed credits...:(

Third option, and one that is directly in your hands - look for, or compile and install an optimized BOINC client, to bring your benchmarks in line with the "norm", or better! That alone will claim more credit, nothing else.

Regards,

Michael

microcraft
"The arc of history is long, but it bends toward justice" - MLK

Stick
Stick
Joined: 24 Feb 05
Posts: 790
Credit: 31,225,324
RAC: 659

RE: Third option, and one

Message 21811 in response to message 21810

Quote:
Third option, and one that is directly in your hands - look for, or compile and install an optimized BOINC client, to bring your benchmarks in line with the "norm", or better! That alone will claim more credit, nothing else.

You can find a variety of optimized BOINC software here: Other sources of BOINC client software.

FalconFly
FalconFly
Joined: 16 Feb 05
Posts: 191
Credit: 15,650,710
RAC: 0

I'm using the Linux BOINC

Message 21812 in response to message 21811

I'm using the Linux BOINC V5.2.13 from here : here

Works fine, Claimed Credits are actually a tad higher than the Default Win32 BOINC, but still way more accurate than the ~50% too low ones from the default Linux BOINC.

I still don't understand why BOINC got never fixed on that issue (they claimed the Win32 compilates actually claimed way too high but never done anything effectively against the difference Linux vs. Win32), therefor it's still required to use optimized (=normal scoring) Linux BOINC builds and hope others do as well.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.