• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Optimized application (crunching program) for Seti@Home

petrusbroder

Elite Member
I have tested the Altivec optimized science application for Seti@Home - BOINC for PowerMac with 2 G5 processors @ 2 GHz:

Standard seti-application: 17 WUs average: 11673 seconds = 194 minutes 33 seconds = 3 hours 14 minutes 33 seconds
Optimized application (Altivec): 17 WUs average: 4522 seconds = 75 minutes 22 seconds = 1 hour 15 minutes 22 seconds

The optimized application takes only 38.7% of the time the standard client uses. Considering that the mac crunches 2 WUs at a time the mac crunches 36 - 39 WUs per day (depending on the WUs) using the optimized application, compared to 13 - 15 WUs using the standard application.

I have run the optimized application for 2 days now and all above WUs have been validated. No computational errors have occurred. Since the crunching times are among the shorter I have - due to the way the credits are calculated (discard the fastest and the slowest value, grant the average of the two remaining) - got more credits than claimed (claimed approx. 12 - 13 credits, granted 18 - 29 credits). Good for science, good for the stats!

The discussion at MacNN can be found here!
The optimized applications can be found here and at the forum above.
 
I, of all people, was just recently considering a Dual G4 rig.. still thinking about it.. perhaps in spring after taxes. 🙂
 
SirUlli: did the optimized client you were testing a few weeks ago ever get released for public use? I'd really like to get my hands on it 😀
 
I'm really puzzled now. I tried the Crunche3r optimized client for Athlon64 (SSE2) on my X2. (5.2.13)

I went from 2227 Whet / 4180 Dhry to 3280 Whet / 10078 Dhry. Is that even possible? I mean, Dhry more than doubled!

Any ideas, comments?
 
it is a great work what the People have done

my Athlon64 3.200+

30.12.2005 01:02:55||Benchmark results:
30.12.2005 01:02:55|| Number of CPUs: 1
30.12.2005 01:02:55|| 2673 double precision MIPS (Whetstone) per CPU
30.12.2005 01:02:55|| 8740 integer MIPS (Dhrystone) per CPU
30.12.2005 01:02:55||Finished CPU benchmarks


for Info

Sir Ulli
 
Pardon me, but I'm feeling lazy because I'm at home on a laptop and a slow net connection (though having a 21" LCD hooked up to the laptop mitigates the pain a bit)... 😉

Anyway, does BOINC for MacOS X behaves itself well on laptops? I may just try this. Basically, I'd like it to shut down and/or slow down when I go off AC, if possible. Clients that have to be manually shut down and restarted to implement this tend to get forgotten (sorry F@H, you're missing out on a G4 1.5GHz because of this). 😱
 
Originally posted by: ProviaFan
Pardon me, but I'm feeling lazy because I'm at home on a laptop and a slow net connection (though having a 21" LCD hooked up to the laptop mitigates the pain a bit)... 😉

Anyway, does BOINC for MacOS X behaves itself well on laptops? I may just try this. Basically, I'd like it to shut down and/or slow down when I go off AC, if possible. Clients that have to be manually shut down and restarted to implement this tend to get forgotten (sorry F@H, you're missing out on a G4 1.5GHz because of this). 😱

BOINC behaves very well on OS X Panther and Tiger (have tried both). Works well on the Mini, iBook, PowerBook and PowerMac. It is the applications which sometimes are lacking. I have tried the following projects on Macs: Seti, Einstein, Rosetta, Predictor and CPDN. Rosetta is OK (I have had some probs with the version 4.79, previous version was flawless), Seti, Einstein, Predictor and CPDN (have not tried the new sulphur model) work just great. There is no application for LHC@Home, SIMAP and some others. SIMAP has promised a application in January.

The settings are simple: In your preferences (go on the 'net to the project, log in to your account, choose view or edit General preferences, then go (on the bottom of the page) to edit preferences: set the top preference to no. Then BOINC will run only when the comp is on AC.

If you have any questions, just ask ... 😀 Good luck, happy crunching and a successful New Year 2006! 🙂
 
Thanks so much, I'll see what I can do about getting set up in the next few days. 🙂

Happy new year to you as well! :thumbsup:
 
I don't know - but - as in a discussion at MacNN - if the results are - and get - validated and if they are similar enough to what other crunchers produce, then they do not harm the science at all.
Normally the code is not changed at all. What most - if not all - of the" optimizers" is that they use better (optimized) compilers, which use SSE2 or SSE3 or other enhancements of the processors to crunch faster. That is why you see one binary for the type of processor: PII, PIII, P4, AMD Athlon Thunderbird, Athlon XP, A64, G4, G5, and for each operating system. Often also newer version of the compilers are used - with better results. Before being released test WUs are run and the files are being compared to see any differences - there are none reported for the apps I am using.
 
Originally posted by: Assimilator1
Peter
These clients sound good 🙂 ,are they approved by Berkley though?


yes, you have also to look here

http://boinc.berkeley.edu/download_other.php

and for Info

BOINC applications, and the BOINC core client, are native-mode programs, so different versions are required for each platform (a 'platform' is the combination of an operating system and a processor type: e.g., Linux and Intel/x86).

BOINC-based projects compile program versions for common platforms (Windows/Intel, Linux/Intel, Mac OS/X. etc.), and place them on its servers. A participant downloads the core client for his platform (assuming that platform is supported by the project). When the core client requests work from the project's scheduling server, the client tells the server its platform, and the server instructs it to download the appropriate executables.

http://boinc.berkeley.edu/anonymous_platform.php

Sir Ulli
 
I see no differance in the time it take to process the WU I did see a HUGE differance in the Benchmarks on my Opteron System
OLD
2155 Double Precision MIPS Whetstones
4018 Integer MIPS Dhrystone
NEW
3098 double precision MIPS (Whetstone) per CPU
9420 integer MIPS (Dhrystone) per CPU

It still takes 1.45 min or so to process on both my systems any idea why or what I did wrong?

I have a Opteron 165 1.8GHz OC to 2.3GHz 1 GB GEIL ram
Pentium 3.0 GHz Prescott - 1 GB ram

Merlin
 
This sounds interesting.
Could it be so that you have installed an optimized BOINC-client (that is the BOINC-stuff, such as BOINC-manager, etc) and not the science application, which runs under BOINC as the "project"? The fact that you get higher benchmarks indicates that it is BOINC itself that has been changed - because the benchmarks are measured by the manager and its files and not by each project application. If you had exchanged the science application - the crunching program - then you would see shorter times ...

As far as I know there are no optimized applications for Opterons for seti@home or other projects - but I may be worng. OTOH as long as the processor has SSE2 or SSE3 enabled it should be able to run the seti-application much faster ... with an optimized application (if there is one) you should be able to crunch a WU in about 45 minutes - 1 h 15 minutes or so ... - I may however be wrong there. 😉

For us to help you, could you please post the files you have downloaded and from where you downloaded them.
 
Originally posted by: petrusbroder
This sounds interesting.
Could it be so that you have installed an optimized BOINC-client (that is the BOINC-stuff, such as BOINC-manager, etc) and not the science application, which runs under BOINC as the "project"? The fact that you get higher benchmarks indicates that it is BOINC itself that has been changed - because the benchmarks are measured by the manager and its files and not by each project application. If you had exchanged the science application - the crunching program - then you would see shorter times ...

As far as I know there are no optimized applications for Opterons for seti@home or other projects - but I may be worng. OTOH as long as the processor has SSE2 or SSE3 enabled it should be able to run the seti-application much faster ... with an optimized application (if there is one) you should be able to crunch a WU in about 45 minutes - 1 h 15 minutes or so ... - I may however be wrong there. 😉

For us to help you, could you please post the files you have downloaded and from where you downloaded them.

Opterons are essentially the exact same chip as an Athlon 64 (except they endure more stringent quality control tests and the socket 940 varients can use registered ram), so the A64 client should run fine on them.
 
I must be installing it wrong you download it from the Seti@home site > install it then shut down BONIC and extract the Pentium4_SSE2.zip files into the "../projects/setiathome.berkeley.edu folder and over right any files Are you setting BONIC up as a service?
 
Originally posted by: merlin230b
I must be installing it wrong you download it from the Seti@home site > install it then shut down BOINC and extract the Pentium4_SSE2.zip files into the "../projects/setiathome.berkeley.edu folder and over right any files Are you setting BOINC up as a service?

No I have not set up BOINC as a service. That should not matter at all.
Hmm, Why don't you look at this thread? There I have some comparisons between different processors: AMD Athlons of different kinds and Intel Pentiums.
Then I wonder if you would not be better off using a optimization prepared for AMD A64. You can find it here. There is a difference between optimizations for the P4-processor and for the A64-processors with SSE2. I think the file can be found here, just click here and it should start downloading.
I hope this does help you ... 🙂
 
Well I got my Opteron 165 system working it is compleating 2 every 46 min though my Pentium 4 3.00GHz Prescott is doing 2 in 3.00 hours I did do the benchmarks and it did change the before and after results. I put the new Optimized files in.

Benchmarks after install of the new files
1887 double precision MIPS (Whetstone) per CPU
2904 integer MIPS (Dhrystone) per CPU

I installed the following files

Boinc_5.2.13_P4_Athlon64_SSE2.zip that contains

boinc.dll
boinc.exe
Microsoft.VC80.CRT.manifest
msvcm80.dll
msvcp80.dll
msvcr71.dll
msvcr80.dll

I do see one labled Prescott_SSE3.zip and it contains these files but I did not install because the names are differnt the one in the Bionc folder is labled setiathome_4.18_windows_intelx86.pdb the ones downloaded are

app_info.xml
setiathome_4.11_windows_intelx86.exe
setiathome_4.11_windows_intelx86.pdb

 
Originally posted by: merlin230b
Well I got my Opteron 165 system working it is compleating 2 every 46 min though my Pentium 4 3.00GHz Prescott is doing 2 in 3.00 hours I did do the benchmarks and it did change the before and after results. I put the new Optimized files in.

Benchmarks after install of the new files
1887 double precision MIPS (Whetstone) per CPU
2904 integer MIPS (Dhrystone) per CPU

I installed the following files

Boinc_5.2.13_P4_Athlon64_SSE2.zip that contains

boinc.dll
boinc.exe
Microsoft.VC80.CRT.manifest
msvcm80.dll
msvcp80.dll
msvcr71.dll
msvcr80.dll

I do see one labled Prescott_SSE3.zip and it contains these files but I did not install because the names are differnt the one in the Bionc folder is labled setiathome_4.18_windows_intelx86.pdb the ones downloaded are

app_info.xml
setiathome_4.11_windows_intelx86.exe
setiathome_4.11_windows_intelx86.pdb

Stop BOINC. Move the 4.18 files to another (save just in case) directory and insert all the optimized files including the 4.11 files. Then Start BOINC ... WOW!

 
Back
Top