My A64x2 2.55gig has gone from 22credit/hour to 26credit/hour, which is only a 15% speedup. OTOH my box had a DCF of .5 using the stock app, which would imply that it's limitations hurt p4's more severely to begin with.
Bah! My PC decided to reset it's clocking during the reboot I made right after installing the new app. Running with it at normal speed I'm seeing the expected 35% drop in time. Or to put it the other way, it's a 50% increase in throughput.
With the S5 stock app crunching about 3.5x as fast as the stock s4 one this puts 4.24 at a 5.25x speedup over the s4 albert app, which means it's closed roughly half the remaining gap between 4.02 and akos's best s4 effort.
First result (a long WU) with 4.24 beta completed successfully, but is still awaiting validation. Completion time on my Athlon MP 2000+ (dual CPU) was just over 8 hours, compared to just over 12 hours with 4.02. Nice work!
-- Tony
Just thought I'd post an update. The above result has validated successfully, as has this one. Crunch time reduction is holding at ~33%.
I just switched over a P3/500 machine that's doing short WUs in about 5:40. I'll report results as soon as it completes some WUs.
i was really glad to see that something's going on to speed things up.
it was early in the morning, i read through bernd machenschalk's "Switching Apps half way trough a result" and did it that way. no probs at all...
then i started to think...that's when things went down the tube.
i decided to re-rename the beta app to 4.24 again, like "the app_info.xml will do the job". yeah right, end of story: six big fat compute errors.
sadly, i still do not know which malfunctioning part of my brain is to be held responsible for thinking that way.
so to all noobs like me, take your time, stick to the advice given here or set your manager to "no new work" and wait for your cache to drown...and then give it a try.
i'm so sorry for the client errors i caused...let's have a (written) speech act: i swear i won't mess things up the next time.
I only copied the application, the xml and the debugfile into the appropriate project folder and restarted the client with all the work in process, WUs in cache were finished with the old 4.02 application, all new queried WUs are now worked at with the new 4.24 application.
What have you done wrong?
Quote:
AMD Opteron 144@1800; 1024MB RAM; Windows XP SP2
WU: 178,76 Credits
4.02: ~ 39562 s
4.24: ~ 26343 s => 33.4% shorter time
AMD Athlon 64 X2 3800+@2000; 512MB RAM; Windows XP SP2
WU: 176,90 Credits
4.02: ~ 34720 s
4.24: 23648 s => 31.9% shorter time
Intel Pentium M 1800; 1024MB RAM; Windows XP SP2
WU: 19.58 Credits
4.02: ~ 5378 s
4.24: ~ 3687 s => 31.4% shorter time
The first two clients are einstein@home only, the Pentium M is also used for other purposes i.e. internet surfing, software development, gaming.
Another one comes in:
Intel Pentium M LV 1200; 768MB RAM; Windows XP SP2
WU: 178.21 Credits
4.02: ~ 59654.69 s
4.24: 39516.14 s => 33.8% shorter time
This client now has the same WU-runtime as the Opteron 144 with the 4.02 client.
The speed gained with the 4.24-application is awesome!
As soon as i get to my server i would also enable the 4.24-application for my Pentium D 805 to see what it gets there, but i am in a good mood.
Execution time ratio (4.24/4.02)
Banias 0.64
Gallatin HT with Simon's SETIenh 0.72
Coppermine #1 0.74
Coppermine #2 0.82 (estimate from result 51% complete so far)
With more data, some revision:
Banias 0.63
Gallatin HT 0.71
Coppermine #1 0.73
Coppermine #2 0.69
Both Coppermines are on Win98 SE systems. I failed to notice that my weekly antivirus scan had badly distorted reported CPU time on the first result on Coppermine #2--a standard Win98 BOINC problem.
Is "Found app_info.xml; using anonymous platform" in your messages? It should show up in the group of messages written right after restarting BOINC. If it's not there, something went wrong.
Quote:
Hi folks,
a short question. I unzipped the 4.24-files to the einstein-folder and allowed new work to einstein.
Still my result-summarize shows app 4.02 and in boinc's work-tab as well.
Is this correct?
4343,9 sec CPU-Time for a short result on an XP2800+.
RE: My A64x2 2.55gig has
)
Bah! My PC decided to reset it's clocking during the reboot I made right after installing the new app. Running with it at normal speed I'm seeing the expected 35% drop in time. Or to put it the other way, it's a 50% increase in throughput.
With the S5 stock app crunching about 3.5x as fast as the stock s4 one this puts 4.24 at a 5.25x speedup over the s4 albert app, which means it's closed roughly half the remaining gap between 4.02 and akos's best s4 effort.
RE: First result (a long
)
Just thought I'd post an update. The above result has validated successfully, as has this one. Crunch time reduction is holding at ~33%.
I just switched over a P3/500 machine that's doing short WUs in about 5:40. I'll report results as soon as it completes some WUs.
-- Tony
oh my god, i screwed it
)
oh my god, i screwed it up...
i was really glad to see that something's going on to speed things up.
it was early in the morning, i read through bernd machenschalk's "Switching Apps half way trough a result" and did it that way. no probs at all...
then i started to think...that's when things went down the tube.
i decided to re-rename the beta app to 4.24 again, like "the app_info.xml will do the job". yeah right, end of story: six big fat compute errors.
sadly, i still do not know which malfunctioning part of my brain is to be held responsible for thinking that way.
so to all noobs like me, take your time, stick to the advice given here or set your manager to "no new work" and wait for your cache to drown...and then give it a try.
i'm so sorry for the client errors i caused...let's have a (written) speech act: i swear i won't mess things up the next time.
I only copied the
)
I only copied the application, the xml and the debugfile into the appropriate project folder and restarted the client with all the work in process, WUs in cache were finished with the old 4.02 application, all new queried WUs are now worked at with the new 4.24 application.
What have you done wrong?
Another one comes in:
Intel Pentium M LV 1200; 768MB RAM; Windows XP SP2
WU: 178.21 Credits
4.02: ~ 59654.69 s
4.24: 39516.14 s => 33.8% shorter time
This client now has the same WU-runtime as the Opteron 144 with the 4.02 client.
The speed gained with the 4.24-application is awesome!
As soon as i get to my server i would also enable the 4.24-application for my Pentium D 805 to see what it gets there, but i am in a good mood.
RE: Execution time ratio
)
With more data, some revision:
Banias 0.63
Gallatin HT 0.71
Coppermine #1 0.73
Coppermine #2 0.69
Both Coppermines are on Win98 SE systems. I failed to notice that my weekly antivirus scan had badly distorted reported CPU time on the first result on Coppermine #2--a standard Win98 BOINC problem.
Pentium 4 (2.8) stock
)
Pentium 4 (2.8)
stock average: 43,500 - 45,000 sec.
beta: 27,800 - 29,100 sec. on 8 granted and 1 pending
Here are two results: -
)
Here are two results:
- Win XP Pro sp2
Pentium 4 3.0 HT
RAM 1Gb
Calculation time : 42750 sec
- Win XP Home sp2
Pentium D 2.8
RAM 1Gb
Calculation time : 30934 sec
[pre] [/pre]
Intel I7 930 - GTX 480 - Windows 7 64
Join BOINC Synergy, the best team in the galaxy!
Something went very wrong
)
Something went very wrong here few hours in.
resultid; 38617164
on this host.
hostid; 544023
Hi folks, a short
)
Hi folks,
a short question. I unzipped the 4.24-files to the einstein-folder and allowed new work to einstein.
Still my result-summarize shows app 4.02 and in boinc's work-tab as well.
Is this correct?
4343,9 sec CPU-Time for a short result on an XP2800+.
Thx + happy crunchin'
Groundhog
Is "Found app_info.xml; using
)
Is "Found app_info.xml; using anonymous platform" in your messages? It should show up in the group of messages written right after restarting BOINC. If it's not there, something went wrong.