3 Different Versions?
20 Nov 2004 20:44:33 UTC
Topic 187125
(moderation:
I currently have einstein 4.35. Then BOINC downloaded 4.42, and I checked the E@H applications section and it said that. Then it downloaded 4.40 and the section says 4.35. What is going on?
--- - 2004-11-20 11:15:15 - May run out of work in 0.50 days; requesting more
Einstein@Home - 2004-11-20 11:15:15 - Requesting 6.59 seconds of work
Einstein@Home - 2004-11-20 11:15:15 - Sending request to scheduler: http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi
Einstein@Home - 2004-11-20 11:15:19 - Scheduler RPC to http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi succeeded
Einstein@Home - 2004-11-20 11:15:20 - Started download of einstein_4.42_windows_intelx86.exe
Einstein@Home - 2004-11-20 11:15:33 - Finished download of einstein_4.42_windows_intelx86.exe
Einstein@Home - 2004-11-20 11:15:33 - Throughput 122829 bytes/sec
--- - 2004-11-20 12:38:36 - May run out of work in 0.50 days; requesting more
Einstein@Home - 2004-11-20 12:38:36 - Requesting 5.30 seconds of work
Einstein@Home - 2004-11-20 12:38:36 - Sending request to scheduler: http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi
Einstein@Home - 2004-11-20 12:38:41 - Scheduler RPC to http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi succeeded
Einstein@Home - 2004-11-20 12:38:42 - Started download of einstein_4.40_windows_intelx86.exe
Einstein@Home - 2004-11-20 12:38:55 - Finished download of einstein_4.40_windows_intelx86.exe
Einstein@Home - 2004-11-20 12:38:55 - Throughput 122961 bytes/sec
3 Different Versions?
)
> I currently have einstein 4.35. Then BOINC downloaded 4.42, and I checked the
> E@H applications section and it said that. Then it downloaded 4.40 and the
> section says 4.35. What is going on?
The 4.42 and 4.40 apps had problems with graphics so we went back to 4.35.
One of the programers should be able to give you a better answer.
Well, no better answer yet.
)
Well, no better answer yet. We found to have trouble with the 4.4x binaries, so we switched back to 4.35 until we have found and fixed the problem. Thanks for participating in this alpha-test!
BM
BM
I have Einstein 4.44 (Windows
)
I have Einstein 4.44 (Windows client) running and it appears to be without any problem so far... (yes, I knocked on wood).
- Marco
> Well, no better answer yet. We found to have trouble with the 4.4x binaries,
> so we switched back to 4.35 until we have found and fixed the problem. Thanks
> for participating in this alpha-test!
>
> BM
>
- Marco
Team Canada
Good to hear. Yes, we found
)
Good to hear. Yes, we found and fixed the problem. The 4.44 ist out since yesterday and besides containing some fixes behind the scenes should be 20-25% faster than the 4.35.
BM
BM
> Good to hear. Yes, we found
)
> Good to hear. Yes, we found and fixed the problem. The 4.44 ist out since
> yesterday and besides containing some fixes behind the scenes should be 20-25%
> faster than the 4.35.
>
> BM
The new version seems very AMD friendly and Intel hostile. My home machine (Athlon XP 3200+) dropped from 43min per work unit to 27min. My work machine (P4 2.66Ghz) went from from 53min/unit to 72min/unit. Looks like the new core may be hiting a sweet spot on the XP and throwing a bubble somewhere in the P4's long pipeline. Anyone else with similar results?
Looks like its happening to
)
Looks like its happening to me too. :) My AMD usually takes just over 60min to do one WU. Now around 40min.
Just checked my results
)
Just checked my results again, and the new version seems to be running about twice as fast. Running a Celeron 1Ghz.
senator2 > The new version
)
senator2
> The new version seems very AMD friendly and Intel hostile.
> My home machine (Athlon XP 3200+) dropped from 43min per work unit to 27min.
> My work machine (P4 2.66Ghz) went from from 53min/unit to 72min/unit. Looks
> like the new core may be hiting a sweet spot on the XP and throwing a bubble
> somewhere in the P4's long pipeline. Anyone else with similar results?
AMD friendly yes, but I am not seeing any negative impacts with the Pentiums at all. Everything has improved substantially.
P4 3.4mhz 1:12 down to :55 (with hyperthreading on) per thread.
P4 3.0mhz 1:20 down to 1:04 (with hyperthreading on) per thread.
Athlon 2400 :46 down to :31
AMD 64 3200 :54 down to :29.5
The AMD 64 had been a real laggard performing even behind my Athlon 2400, but now it is finally crunching as it should. It showed the largest improvement at 80%. The 2400 is amazing, but don't ask me why, it is not overclocked. Not complaining though.
So, you've improved the Linux client and the Windows client, when will the Mac client see some improvement? ;-)
Team MacNN - The best Macintosh team ever.
> AMD friendly yes, but I am
)
> AMD friendly yes, but I am not seeing any negative impacts with the Pentiums
> at all. Everything has improved substantially.
>
> P4 3.4mhz 1:12 down to :55 (with hyperthreading on) per thread.
> P4 3.0mhz 1:20 down to 1:04 (with hyperthreading on) per thread.
> Athlon 2400 :46 down to :31
> AMD 64 3200 :54 down to :29.5
Hmm, I don't feel bad about a 2.66Ghz P4 doing 1:12 if your 3.4Ghz (I hope it's not 3.4Mhz) is doing 0:55, but I was doing 0:53 on the old core. It's a laptop so I wonder if the larger cache on the Mobile was helping before but has been exceeded with the new core. Looks like the P4 cache is getting clobbered.
> The AMD 64 had been a real laggard performing even behind my Athlon 2400, but
> now it is finally crunching as it should. It showed the largest improvement
> at 80%. The 2400 is amazing, but don't ask me why, it is not overclocked.
> Not complaining though.
Sounds about right. My 3200+ is 2.2Ghz, your 2400+ would be 2.0Ghz (just lower FSB speed) if it's running in Cache it should be about 10% slower than mine (26'53" last unit)
Are all these values posted
)
Are all these values posted here taken from einstein_4.44 on Windows, or are there some Linux ones between them? On our systems (mainly Intel) the speedup under Linux (gcc) was larger than that of Windows (MSVC), but we've never ever seen a slowdown. The two (Windows) machines I'm sitting in front of (2,6GHz P4 and 1,8GHz Centrino) consistently show a speedup of about 20%. On Linux Systems we've seen up to 80%.
We are working to get the Mac graphics running, the next Mac app will have the faster "science code", too.
BM
BM