Hey guys,
Thanks for the info on that barton core. I ran an athlon 64 3000+ last year and got a lot of overclock on it, my 3700+ doesn't seem to want to overclock much, have it running at 2.45 instead of 2.4, had no idea how easy and how much you could get out of the barton. Neither one of those chips showed up as anything diff in windows, their speeds were faster, but the chip designation never changed. With the barton being recognized as a 3200+ now, guess that's what the chip "really" is? I fiddled with the fsb a little bit more yesterday, but to no avail. I really don't feel like messing with the voltages, I did bump voltages when running up my 3000+, both cpu and mem, I have a feeling any more increases from here will be incremental, not as drastic as what I have already gotten out of it.
Thanks again you Gary for the info on that chip. The machine is now doing about 4 wu's a day instead of @3!! May be insignificant numbers when looking at the entire project, but every wu crunched helps. I just wish I were able to run my machines 24/7 :-(
I have read it else where, and I will repeat it, too bad we can't take advantage of the processing power of our video cards. My 3700+ machine used to play games, it has a 6800gt OC with 256mg of ddr3 on it, and I have been able to overclock it quite a bit. Now that card just sets there idle, no games anymore, machine just crunches. I do realize that we can't use that power, but what a shame to have a $400 video card just setting there idle.
Not exactly a WU speed, but closer to this topic than to Milestones. I've just last night finished crunching 24 WUs in 5 days (12/10-12/14), on a single Barton-core Athlon.
microcraft
"The arc of history is long, but it bends toward justice" - MLK
This "Best 3150" laptop from 1999 had reached the end of its useful life in production use. I replaced the defunct Windows 98 with Debian and now the machine serves as a surfboard@home and crunches for Einstein@home. Recent average credit usually varies between 30-40. CPU time per WU is 160...165 ks and claimed credit about 90.
[pre]
Created 23 Sep 2005 18:08:18 UTC
Total Credit 2,919.82
Recent average credit 34.55
CPU type GenuineIntel Celeron (Mendocino)
Number of CPUs 1
Operating System Linux 2.6.12-1-686
Memory 187.47 MB
Cache 976.56 KB
Swap space 184.53 MB
Total disk space 3.71 GB
Measured floating point speed 389.29 million ops/sec
Measured integer speed 528.72 million ops/sec
Average upload rate 0.56 KB/sec
Average download rate 34.52 KB/sec
Average turnaround time 2.14 days
Number of times client has contacted server 85
Last time contacted server 31 Dec 2005 17:36:20 UTC
% of time BOINC client is running 99.977 %
While BOINC running, % of time work is allowed 99.2703 %
Average CPU efficiency 0.932255
Result duration correction factor 1.306171
[/pre]
It did average 42,000sec, then I did a defrag & reinstall everthing but Office Pro, onto my computer. Now my Average time is 9,100-8,800 sec.
Running Setie @ the same time @ 50%
Running Einstein @ 66.67% (using VR RAM)
P4 3ghz 2gig RAM + 1064gig VR-RAM
XP Pro
Laptop
I want pancakes! God, do you people understand every language except English? Yo quiero pancakes! Donnez moi pancakes! Click click bloody click pancakes!
-Stewie Griffin, Family Guy
I would like to share comparisons of two Athlon XP 2800's.
host 422996 runs with Window XP Home and on Einstein averaged 22239.78 sec, 69.93 claimed credit and 71.74 granted credit.
host 443142 runs Linspire linux 2.6.10 and on Einstein averaged 28039.41 sec, 88.93 claimed and 76.98 granted.
Windows took 6.19 hours and Linux took 7.79 hours and these were my best Einstein crunchers.
But now with Albert I have this
422996 does 14557.01 sec, 50.65 claimed, 46.13 granted
443142 does 5335.54 sec, 16.94 claimed, 15.63 granted
The alberts with Linux are taking are taking 37 percent of the time of the Windows ones and getting 34 percent of the credit of the Windows machine.
Both machines have BOINC 5.2.13 but I also use optimized BOINC's by Crunch3r and both machines do Seti@home also. The linux machine does LHC when it's available but the Windows machine could never get a valid LHC result. I don't know whether to blame the Kingston ram I put in it or the refurb Gigabyte mobo.
422996 does 14557.01 sec, 50.65 claimed, 46.13 granted
443142 does 5335.54 sec, 16.94 claimed, 15.63 granted
The alberts with Linux are taking are taking 37 percent of the time of the Windows ones and getting 34 percent of the credit of the Windows machine.
Both machines have BOINC 5.2.13 but I also use optimized BOINC's by Crunch3r and both machines do Seti@home also. The linux machine does LHC when it's available but the Windows machine could never get a valid LHC result. I don't know whether to blame the Kingston ram I put in it or the refurb Gigabyte mobo.
Regards,
Pam
>Hi, Pam.....it's possible you are comparing apples and oranges with the 'Alberts'. There is an variety of these new WUs and they don't have consistant processing times like the old 'Einstein' WUs. They are all called 'Alberts' but the low frequency versions complete in approximately 30% of the time of the 'Einsteins' and the higher frequency ones closer to 'Einstein' times. It's possible one machine (443142) has the faster WU and the other the ones that take more processing. Re. the Kingston RAM: I've had no problems with that brand. Hope this helps....Cheers, Rog.
>Hi, Pam.....it's possible you are comparing apples and oranges with the 'Alberts'. There is an variety of these new WUs and they don't have consistant processing times like the old 'Einstein' WUs. They are all called 'Alberts' but the low frequency versions complete in approximately 30% of the time of the 'Einsteins' and the higher frequency ones closer to 'Einstein' times. It's possible one machine (443142) has the faster WU and the other the ones that take more processing. Re. the Kingston RAM: I've had no problems with that brand. Hope this helps....Cheers, Rog.
Hi, Rog, thanks for pointing that out. I remembered that there were different processing times but I forgot a host gets work units from the same set of data until that set is all sent out. Or something like that. I guess eventually, the two machines could switch types of Alberts.
I remembered that there were different processing times but I forgot a host gets work units from the same set of data until that set is all sent out. Or something like that. I guess eventually, the two machines could switch types of Alberts.
>Right on, Pam.....have a great New Year....Cheers, Rog.
I remembered that there were different processing times but I forgot a host gets work units from the same set of data until that set is all sent out. Or something like that. I guess eventually, the two machines could switch types of Alberts.
I have work on the G5 that are 51 minutes, and now some that are listed at 3.5 hours ... have not completed one so not sure how long it will be really ...
Though I got another selection of 51 minute ones agian ... it will be interesting to see the mixes and what the timings are like. I am seeing a much wider variance on the WIndows machines with times from my prior ~12 hours down to 4 hours ...
Hey guys, Thanks for the
)
Hey guys,
Thanks for the info on that barton core. I ran an athlon 64 3000+ last year and got a lot of overclock on it, my 3700+ doesn't seem to want to overclock much, have it running at 2.45 instead of 2.4, had no idea how easy and how much you could get out of the barton. Neither one of those chips showed up as anything diff in windows, their speeds were faster, but the chip designation never changed. With the barton being recognized as a 3200+ now, guess that's what the chip "really" is? I fiddled with the fsb a little bit more yesterday, but to no avail. I really don't feel like messing with the voltages, I did bump voltages when running up my 3000+, both cpu and mem, I have a feeling any more increases from here will be incremental, not as drastic as what I have already gotten out of it.
Thanks again you Gary for the info on that chip. The machine is now doing about 4 wu's a day instead of @3!! May be insignificant numbers when looking at the entire project, but every wu crunched helps. I just wish I were able to run my machines 24/7 :-(
I have read it else where, and I will repeat it, too bad we can't take advantage of the processing power of our video cards. My 3700+ machine used to play games, it has a 6800gt OC with 256mg of ddr3 on it, and I have been able to overclock it quite a bit. Now that card just sets there idle, no games anymore, machine just crunches. I do realize that we can't use that power, but what a shame to have a $400 video card just setting there idle.
Bill
Not exactly a WU speed, but
)
Not exactly a WU speed, but closer to this topic than to Milestones. I've just last night finished crunching 24 WUs in 5 days (12/10-12/14), on a single Barton-core Athlon.
microcraft
"The arc of history is long, but it bends toward justice" - MLK
This "Best 3150" laptop from
)
This "Best 3150" laptop from 1999 had reached the end of its useful life in production use. I replaced the defunct Windows 98 with Debian and now the machine serves as a surfboard@home and crunches for Einstein@home. Recent average credit usually varies between 30-40. CPU time per WU is 160...165 ks and claimed credit about 90.
[pre]
Created 23 Sep 2005 18:08:18 UTC
Total Credit 2,919.82
Recent average credit 34.55
CPU type GenuineIntel Celeron (Mendocino)
Number of CPUs 1
Operating System Linux 2.6.12-1-686
Memory 187.47 MB
Cache 976.56 KB
Swap space 184.53 MB
Total disk space 3.71 GB
Measured floating point speed 389.29 million ops/sec
Measured integer speed 528.72 million ops/sec
Average upload rate 0.56 KB/sec
Average download rate 34.52 KB/sec
Average turnaround time 2.14 days
Number of times client has contacted server 85
Last time contacted server 31 Dec 2005 17:36:20 UTC
% of time BOINC client is running 99.977 %
While BOINC running, % of time work is allowed 99.2703 %
Average CPU efficiency 0.932255
Result duration correction factor 1.306171
[/pre]
Very little variation from ~
)
Very little variation from ~ 8hrs 20 min on EAH wu on my P4 2.8 ghz 512kb L2 with the einstein 4.79.
It did average 42,000sec,
)
It did average 42,000sec, then I did a defrag & reinstall everthing but Office Pro, onto my computer. Now my Average time is 9,100-8,800 sec.
Running Setie @ the same time @ 50%
Running Einstein @ 66.67% (using VR RAM)
P4 3ghz 2gig RAM + 1064gig VR-RAM
XP Pro
Laptop
I want pancakes! God, do you people understand every language except English? Yo quiero pancakes! Donnez moi pancakes! Click click bloody click pancakes!
-Stewie Griffin, Family Guy
I would like to share
)
I would like to share comparisons of two Athlon XP 2800's.
host 422996 runs with Window XP Home and on Einstein averaged 22239.78 sec, 69.93 claimed credit and 71.74 granted credit.
host 443142 runs Linspire linux 2.6.10 and on Einstein averaged 28039.41 sec, 88.93 claimed and 76.98 granted.
Windows took 6.19 hours and Linux took 7.79 hours and these were my best Einstein crunchers.
But now with Albert I have this
422996 does 14557.01 sec, 50.65 claimed, 46.13 granted
443142 does 5335.54 sec, 16.94 claimed, 15.63 granted
The alberts with Linux are taking are taking 37 percent of the time of the Windows ones and getting 34 percent of the credit of the Windows machine.
Both machines have BOINC 5.2.13 but I also use optimized BOINC's by Crunch3r and both machines do Seti@home also. The linux machine does LHC when it's available but the Windows machine could never get a valid LHC result. I don't know whether to blame the Kingston ram I put in it or the refurb Gigabyte mobo.
Regards,
Pam
RE: But now with Albert I
)
RE: >Hi, Pam.....it's
)
Hi, Rog, thanks for pointing that out. I remembered that there were different processing times but I forgot a host gets work units from the same set of data until that set is all sent out. Or something like that. I guess eventually, the two machines could switch types of Alberts.
Regards,
Pam
RE: RE: I remembered
)
RE: I remembered that there
)
I have work on the G5 that are 51 minutes, and now some that are listed at 3.5 hours ... have not completed one so not sure how long it will be really ...
Though I got another selection of 51 minute ones agian ... it will be interesting to see the mixes and what the timings are like. I am seeing a much wider variance on the WIndows machines with times from my prior ~12 hours down to 4 hours ...