I have access to 3 different CPUs and here are the first results:
1.) Pentium 4 2GHz
E@H 4.79: average 43178 (24 values from 42972 to 43740)
E@H 0.03: average 42060 (04 values from 41529 to 42558)
that is 2.5% faster
2.) Pentium III Xeon 1.4 GHz (dual CPU Server System)
E@H 4.79: average 40856 (24 values from 40746 to 40964)
E@H 0.03: average 44869 (02 values from 44822 to 44916)
that is 10% slower
3.) AMD Athlon XP1700+
E@H 4.79: average 32471 (03 values from 32357 to 32567)
E@H 0.03: no WU finished yet, but estimated usage: 35400
that is 9% slower
Only running one machine on the 0.03 test application -- Celeron 1.0gig with 524meg memory, WindowsME for OS and an nVidia TNT M64 video card ... using the "run always" option in BOINC Manager.
Completed first WU in 84,945.
This is about 20% slower than previous WUs, but since this is only the first result, that figure may change (based on additional returns). Next WU is in process, but estimated time to complete is of similar duration.
ScreenSaver is visible and stable when called (no dropouts), but I don't normally use it in day to day operation here. I'll post again when more data is available.
I have access to 3 different CPUs and here are the first results:
...
3.) AMD Athlon XP1700+
E@H 4.79: average 32471 (03 values from 32357 to 32567)
E@H 0.03: no WU finished yet, but estimated usage: 35400
that is 9% slower
System is an Athlon XP 2200+ (1800MHz), 512MB; video card is ATI Radeon 7500 series, 64MB RAM. OS is WinXP Pro.
Computed and returned three results, average time 28,660 seconds. Previous results were in 26,100 range.
Can [Show graphics] at will. No graphics seen under previous version.
Can select BOINC screensaver and [Preview] at will.
Screensaver activates and deactivates normally.
It would be cool if you could eventually compile an SSE, SSE2, SSE3 version for Windows.
I once compiled the code with SSE optimization and found it to be about 1% faster than without, so there's not much in the normal code that the compiler found to be suitable for SSE.
Over at SETI@home, there has been some significant success increasing the performance of their application by using a modern, up-to-date compiler (Intel C++ v9.0). They were able to furhter increase performance by utilizing FFTW / IPP library v4.1. I'm not aware of the data processing methods the Einstein@home application is executing, but perhaps there's some value in exploring the same approach?
Only running one machine on the 0.03 test application -- Celeron 1.0gig with 524meg memory, WindowsME for OS and an nVidia TNT M64 video card ... using the "run always" option in BOINC Manager.
Completed second WU in 84,213.
Looks like as long as I keep crunchin' 24/7, I'll be the high side limit for claimed credit. Elapsed time remains at 20% above previously reported values, so to get close to everybody else I can only hope that someone puts together an 'optimized' version for us users with old PIII machines ....
Okay, I just ended crunching my one unit with the normal client. Will now crunch one with the 0.03 client and check results afterwards. It's nice to see that the graphics are working again, but I do miss the interactive keys we had in Pirates. :(
I have access to 3 different
)
I have access to 3 different CPUs and here are the first results:
1.) Pentium 4 2GHz
E@H 4.79: average 43178 (24 values from 42972 to 43740)
E@H 0.03: average 42060 (04 values from 41529 to 42558)
that is 2.5% faster
2.) Pentium III Xeon 1.4 GHz (dual CPU Server System)
E@H 4.79: average 40856 (24 values from 40746 to 40964)
E@H 0.03: average 44869 (02 values from 44822 to 44916)
that is 10% slower
3.) AMD Athlon XP1700+
E@H 4.79: average 32471 (03 values from 32357 to 32567)
E@H 0.03: no WU finished yet, but estimated usage: 35400
that is 9% slower
Udo
Only running one machine on
)
Only running one machine on the 0.03 test application -- Celeron 1.0gig with 524meg memory, WindowsME for OS and an nVidia TNT M64 video card ... using the "run always" option in BOINC Manager.
Completed first WU in 84,945.
This is about 20% slower than previous WUs, but since this is only the first result, that figure may change (based on additional returns). Next WU is in process, but estimated time to complete is of similar duration.
ScreenSaver is visible and stable when called (no dropouts), but I don't normally use it in day to day operation here. I'll post again when more data is available.
If I've lived this long - I gotta be that old!
Completed 3 WU's. All were
)
Completed 3 WU's. All were between 35 and 40 minutes slower. My average before was 7 hours. CPU is AMD Athlon XP 2500+.
Gregg
RE: I have access to 3
)
The result is indeed 35488 (9.3% slower).
Udo
First results of 0.03
)
First results of 0.03 test:
System is an Athlon XP 2200+ (1800MHz), 512MB; video card is ATI Radeon 7500 series, 64MB RAM. OS is WinXP Pro.
Computed and returned three results, average time 28,660 seconds. Previous results were in 26,100 range.
Can [Show graphics] at will. No graphics seen under previous version.
Can select BOINC screensaver and [Preview] at will.
Screensaver activates and deactivates normally.
No problems noted. --Don
RE: RE: It would be cool
)
Over at SETI@home, there has been some significant success increasing the performance of their application by using a modern, up-to-date compiler (Intel C++ v9.0). They were able to furhter increase performance by utilizing FFTW / IPP library v4.1. I'm not aware of the data processing methods the Einstein@home application is executing, but perhaps there's some value in exploring the same approach?
My first WU done with v. 0.03
)
My first WU done with v. 0.03 validated ok and finished a couple of hundred seconds faster than stock client.
Happy crunching,
Only running one machine on
)
Only running one machine on the 0.03 test application -- Celeron 1.0gig with 524meg memory, WindowsME for OS and an nVidia TNT M64 video card ... using the "run always" option in BOINC Manager.
Completed second WU in 84,213.
Looks like as long as I keep crunchin' 24/7, I'll be the high side limit for claimed credit. Elapsed time remains at 20% above previously reported values, so to get close to everybody else I can only hope that someone puts together an 'optimized' version for us users with old PIII machines ....
If I've lived this long - I gotta be that old!
RE: My first WU done with
)
Oh - faster? What machine (CPU)?
BM
BM
Okay, I just ended crunching
)
Okay, I just ended crunching my one unit with the normal client. Will now crunch one with the 0.03 client and check results afterwards. It's nice to see that the graphics are working again, but I do miss the interactive keys we had in Pirates. :(