Well, the 4.27 hase been stripped off all SSE2 instructions that made it fast on machines capable of SSE2. I was hoping that the faster sin/cos code would make up for this, but whether it actually does or not depends on the particular CPU. If it isn't slower than 4.21, then the bugfixes and the broader variety of CPUs it runs on should be worth it anyway.
I didn't think I had enough data, but I made up a chart just to prove it. perhaps someone else can know for sure if it's faster. Someone who knows what the waveform should continue out to look like. My guess is that it's faster than 4.24 but without more 4.24 data from the same freq, I'm stuck. Well, 4.27 does run and I suppose that's part of the beta test. Hopefully, I'll get enough samples with 4.27 to compare the next app.
Archae86s chart doesn't get up to 775 but the 712 freq on the 6000 should put the "periodocity"??? at 105ish, so the 102 task done should be about peak. Meaning the cycle would run from zero to 105ish. Still doesn't help me figure if it's faster. The 5200 is unknown, and it's moved on to a new freq, so no more data will be forthcoming for that one (unless einstein repeats).
But that windows stuff in the app_info.xml is funny.
Sorry, the app_info.xml was derived from the Windows one. Maybe I should delete the tag altogether. All people having trouble on 64 bit: give it a try. Remove the line manually.
BM
This doesn't just apply to 64-bit versions. I've also had to edit the app_info file for my 32-bit machines. (They were giving me the same error message.)
Yes, it seems there is no speedup also on Core2 arch. Im still working on some numbers for 4.21 vs 4.27 and will try to post a graph tonight (need more results), so far it doesnt look too good.
Wow, your "Intel(R) Xeon(R) CPU 3060 @ 2.40GHz" is crunching away the WUs at around 12,000 CPU seconds / unit !!!! That's quite impressive and looks good to me :-). May I ask if this one is overclocked?? This dual core would produce >> 3000 credits/day.
Yes, its the new WUs in the 800 range, my times were around 17,000 secs on the "common" WUs.
The xeon3060 is at 3.64GHz, recently upgraded from 975X to X38 mainboard to be ready for Wolfdale/Yorkfield, has seen a performance improvement in E@H with the new board, X38 obviously has some good MCH timing improvements and stuff. Not many seconds but still an improvement. E8400 Wolfdale is in the mail and E8500 on order, 12,000sec is gonna look slow when those arrives :D
Performance on Core2, 4.21 vs 4.27... Too big variations to say for sure, need more results with both apps, currently im getting WUs from a mix of data files, cant compare them, would be too unreliable.
Heres some results from h1_0825.90_S5R2__415_S5R3a to h1_0825.90_S5R2__403_S5R3a, not enough to call a winner, too much variation in crunching times.
Its not so interesting anyway... the Signal 11 error is what counts, failing WUs is a bigger hazard to my RAC than slightly lower/higher performance, im just being too much curious here and dont have anything better to do than watching Boinc do its thing while waiting for the Wolfdale (Thursday is the big day)
I'm running Ubuntu 8.04 64 bit with the 5.10.30 boinc-client.
The "Show Graphics" button in boinc manager doesn't produce any output.
But if I open a terminal and run the einstein_S5R3_4.27_graphics_i686-pc-linux-gnu program with a workunit from slots/0 as parameter I get a perfect graphics window.
From stderr:
No protocol specified
GLUT: Fatal Error in BOINC: could not open display: :1.0
I'm running Ubuntu 8.04 64 bit with the 5.10.30 boinc-client.
The "Show Graphics" button in boinc manager doesn't produce any output.
But if I open a terminal and run the einstein_S5R3_4.27_graphics_i686-pc-linux-gnu program with a workunit from slots/0 as parameter I get a perfect graphics window.
From stderr:
No protocol specified
GLUT: Fatal Error in BOINC: could not open display: :1.0
Open a terminal and type "echo $DISPLAY". If you get something different than ":1" or ":1.0", terminate the Manager and the client ("killall boinc") and run "BOINC/run_manager" again. Having a DISPLAY :1 is slightly unusual if you have only one screen and are not logged in remotely.
Well, the 4.27 hase been
)
Well, the 4.27 hase been stripped off all SSE2 instructions that made it fast on machines capable of SSE2. I was hoping that the faster sin/cos code would make up for this, but whether it actually does or not depends on the particular CPU. If it isn't slower than 4.21, then the bugfixes and the broader variety of CPUs it runs on should be worth it anyway.
BM
BM
I didn't think I had enough
)
I didn't think I had enough data, but I made up a chart just to prove it. perhaps someone else can know for sure if it's faster. Someone who knows what the waveform should continue out to look like. My guess is that it's faster than 4.24 but without more 4.24 data from the same freq, I'm stuck. Well, 4.27 does run and I suppose that's part of the beta test. Hopefully, I'll get enough samples with 4.27 to compare the next app.
Archae86s chart doesn't get up to 775 but the 712 freq on the 6000 should put the "periodocity"??? at 105ish, so the 102 task done should be about peak. Meaning the cycle would run from zero to 105ish. Still doesn't help me figure if it's faster. The 5200 is unknown, and it's moved on to a new freq, so no more data will be forthcoming for that one (unless einstein repeats).
RE: RE: But that windows
)
This doesn't just apply to 64-bit versions. I've also had to edit the app_info file for my 32-bit machines. (They were giving me the same error message.)
RE: Yes, it seems there is
)
Wow, your "Intel(R) Xeon(R) CPU 3060 @ 2.40GHz" is crunching away the WUs at around 12,000 CPU seconds / unit !!!! That's quite impressive and looks good to me :-). May I ask if this one is overclocked?? This dual core would produce >> 3000 credits/day.
Bikeman
There are new WUs from higher
)
There are new WUs from higher frequencies out in the wild that take about half the time of the longest WUs.
Opteron 185:
http://einsteinathome.org/host/887182/tasks
cu,
Michael
Yes, its the new WUs in the
)
Yes, its the new WUs in the 800 range, my times were around 17,000 secs on the "common" WUs.
The xeon3060 is at 3.64GHz, recently upgraded from 975X to X38 mainboard to be ready for Wolfdale/Yorkfield, has seen a performance improvement in E@H with the new board, X38 obviously has some good MCH timing improvements and stuff. Not many seconds but still an improvement. E8400 Wolfdale is in the mail and E8500 on order, 12,000sec is gonna look slow when those arrives :D
Team Philippines
Ah, I see, I didn't know
)
Ah, I see, I didn't know about the new WUs. Still, impressive performance!
Bikeman
Performance on Core2, 4.21 vs
)
Performance on Core2, 4.21 vs 4.27... Too big variations to say for sure, need more results with both apps, currently im getting WUs from a mix of data files, cant compare them, would be too unreliable.
Heres some results from h1_0825.90_S5R2__415_S5R3a to h1_0825.90_S5R2__403_S5R3a, not enough to call a winner, too much variation in crunching times.
415 - 12882.92 - 237.71 - app 4.27
414 - 12028.44 - 237.71 - app 4.21
413 - 11991.54 - 237.71 - app 4.21
412 - 12920.52 - 237.71 - app 4.21
411 - 13335.43 - 237.71 - app 4.21
410 - 11774.68 - 237.71 - app 4.27
409 - 11773.88 - 237.71 - app 4.27
408 - 12011.39 - 237.71 - app 4.27
407 - 11915.32 - 237.71 - app 4.27
406 - 12306.05 - 237.71 - app 4.27
405 - 12362.14 - 237.71 - app 4.27
404 - 11465.55 - 237.71 - app 4.21
403 - 11401.08 - 237.71 - app 4.21
Its not so interesting anyway... the Signal 11 error is what counts, failing WUs is a bigger hazard to my RAC than slightly lower/higher performance, im just being too much curious here and dont have anything better to do than watching Boinc do its thing while waiting for the Wolfdale (Thursday is the big day)
Team Philippines
RE: I'm running Ubuntu 8.04
)
From stderr:
http://einsteinathome.org/task/90982448
RE: RE: I'm running
)
Open a terminal and type "echo $DISPLAY". If you get something different than ":1" or ":1.0", terminate the Manager and the client ("killall boinc") and run "BOINC/run_manager" again. Having a DISPLAY :1 is slightly unusual if you have only one screen and are not logged in remotely.
BM
BM