MacOS PPC Beta Test App 4.26 available
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4313
Credit: 250740031
RAC: 34773
25 Aug 2006 11:36:01 UTC
Topic 191736
(moderation:
I put a new MacOS App for testing on the Beta Test Page.
This features the AltiVec code that's already in the 4.27 G5 App. However in the 4.26 this is compiled for backward compatibility with G4 (and G3), so the 4.27 should be faster on G5 Macs.
This code should show a noticable speedup over the current 4.12 App on G4. How much exactly is left to you to find out.
The "generic" code ist more or less the same, so there shouldn't be much of a change for G3. Please test it on them anyway, to make sure it still runs on them.
Happy crunching!
BM
BM
MacOS PPC Beta Test App 4.26 available
)
I have tried the new app on my iBook G4 933 MHz. All five results valided successfully. Times went done from around 9600 secs to 7100 secs for a small work unit, which makes my iBook faster than my Pentium III 733 (needing 8300 secs) again. And puts the world back into the right order. ;-)
Do I need to worry about the good old PowerPCs, as there hasn't been a post to this thread for 24 hours? Anyway, thanks for the speedup and power-saving...
Cheers!
Hello, i just launch this
)
Hello, i just launch this Beta application on my Powermac G5 1,8Ghz (single proc)
Before using the 4.12 version; it worked WU in 12 000 seconds (CPU TIME) and credit 20-40 credits ;since few weeks it works WU in 44 000 seconds(so bigs WU) and credit 178 or 120 credits;
How can i know the exact "size" of each WU ?
i give you results/CPU time/bugs tomorrow for the first/second WU computed.
Cristobool,
PS:
my results at this time(if the link works):
http://einsteinathome.org/account/tasks
my login is [AF]CRISTOBOOL
team: L'alliance Francophone http://einstein.phys.uwm.edu/team_display.php?teamid=17
PS2: sorry for my english language !
RE: […] How can i know
)
Look at the second part of its name: if the number is greater than 400, it’s a “long� WU worth about 120 credits; otherwise it’s a “short� that’ll earn you around 14. For example, the last WU my G5 returned was named “h1_0927.0_S5R1__336_S5R1a�; since 927 > 400 it’s clearly a long WU.
I notice you’re running an old BOINC client, v4.44, which is trying to claim credit according to benchmarks, instead of using the new system; the earliest version that supports E@h’s method (and SETI@home’s Flop-counting) is v5.2.6, and I believe the standard current version is v5.4.11.
Voir la deuxième partie du nom. Si le numéro est plus grand que 400, c’est une unité de travail « longue » qui vaut environ 120 crédits, autrement c’est « courte », vous gagnant autour 14. Par exemple, la dernière que mon G5 a retourné s’appelait “h1_0927.0_S5R1__336_S5R1a� ; puisque 927 > 400 c’est clairement une unité longue.
Vous exécutez un client BOINC vieilli, v4.44, qui essaye de demander des crédits selon les calculs de référence, au lieu d’utiliser le nouveau système ; la première version qui supporte la methode d’E@h (et cela de SETI@home, comptant les exécutions) est v5.2.6, et je croie que la version standard en cours est v5.4.11.
(Pardonnez mon français faible.)
RE: Hello, i just launch
)
Thanks for trying this. If you care about speed, for a G5 I would, however, recommend the 4.27 App. It has been compiled with special settings for the G5 that I didn't get to work in a compatible App such as the 4.26, and thus should run a little faster there. I would be curious to know how large the speedup actually is between 4.26 and 4.27 - should be around 20%.
That has nicely been answered (as far as my French goes) by Odysseus. There is still a variation in the order of +-5% between Workunits with the same "basename" (or data file name), which, in his example, is "h1_0927.0". This should correspond to a variation of the (granted) credit - you would also see this in the claimed credit if you were using a newer BOINC Core Client.
It's not possible for someone else to see your result page (that contains the results of all your machines) directly, but the results for a single computer are visible. In your case this is http://einsteinathome.org/host/474982/tasks
BM
BM
RE: I put a new MacOS App
)
Assuming this passes beta test will it eventually be pushed out automatically?
RE: RE: Hello, i just
)
thanks for your answers Bernd and odysseus,
i'm a little bit surprise because WU computed on my computer (with 4.12) are all "1400" of size computed in 44000 seconds however odysseus's WU "927" of size have been computed in 54 000 seconds on his Powermac 11,2 (a quad G5 2,5Ghz)
So the new BOINC have a new system for credits ? it's better ? will it works with Mac OS X 10.3.9 (panther) version i do not have Tiger.
I use a special version of BOINC (superbench of team MACNN)to have good credits
with the SETI@Home optimized application of MACNN;but i replace SETI by Enstein since months.
Will the 4.26 replace 4.12 for all einstein MACusers if the test of 4,26 is conclusive ?
i continue the beta testing then i will use the 4.27 and new BOINC(if it works on panther) thanks.
For the beta testing: the first WU is not finish,it seems to work.
RE: i'm a little bit
)
Only a dual-core 2.3-GHz. I couldn’t justify the extra $1000 for the quad … But the number of CPUs doesn’t make any difference to crunching times, because each WU is only computed by one processor at a time. The system was averaging about 70000 seconds (11.5 h) per long WU with v4.12.
BOINC v5.4.9 runs fine under Panther on my G4s.
I installed E@h v4.26 on my G4/733 on Friday afternoon. It hasn’t reported any results yet; it still had some work that was already assigned to v4.12.
70 0000 seconds for 2,3Ghz
)
70 0000 seconds for 2,3Ghz with the 4.12 against 44 000 for 1,8Ghz with 4.12 too ???
the first WU with the 4.26 is finish ,success and in "pending"
h1_1412.5_S5R1__3068_S5R1a so "1412,5" 36910 seconds it's good about 20% faster than 4.12 version.
the first WU is valid ,
)
the first WU is valid , granted credit 120,63;
i see the other computer who have done this WU is a Pentium3 1,4Ghz in 37000 seconds; so a Pentium3 is more faster than a G5 on einstein ??? it's amazing !
in other hand another WU have been computed in 42000 seconds on a Pentium4 3Ghz
and another in 60 000 seconds on a Pentium4 3,2Ghz so i don't understand ???
I just upgrade my Boinc
)
I just upgrade my Boinc software with the 5.4.9 version before i use the 4.44 version but i have got a problem ; there is a duplication of my computer in the list of my computers i have two computer ID for the same computer how to combine them ?
http://einstein.phys.uwm.edu/hosts_user.php