[RC5] Mac G5 Benches

Jonathan Smith jsmith at cygnacom.com
Tue Sep 9 14:51:58 EDT 2003


According to an interview I read, but can't seem to locate at the moment,
the person who develops the G4/G5 cores thinks that ultimately the G5 will
be slightly faster per MHz than the G4 in rc5-72.  However since he didn't
yet have access to a G5 to do development and code profiling on his first
attempt at a G5 core was worse that just using the older G4e core.

Notice that the KKS 7450 (G4e) core produced noticeably faster results than
the KKS 970 (G5) core.

So the problem doesn't appear to be fundamental, just the fact that the
coder doesn't (didn't) have a machine to optimize on.

	Jonathan Smith

-----Original Message-----
From: Stephane Leon [mailto:StephaneLeon_ml at Bluewin.ch]
Sent: Tuesday, September 09, 2003 1:02 PM
To: Thorsten Wolf; D.net Discussion
Subject: Re: [RC5] Mac G5 Benches


Nice !
But i am wondering why its slower than G4 stats :

http://n0cgi.distributed.net/speed/ 
query.php?cputype=99&cpumhz=1250&recordid=1&contest=rc572&multi=0

Strange ?


Le Mardi, 9 sep 2003, à 16:06 Europe/Zurich, Thorsten Wolf a écrit :

> Hello Everyone.
>
> I don't know if anyone has postet G5 bench results yet, but here are
> some:
>
> Hardware: G5 1.6 GHz  Single CPU
>
>
> distributed.net client for Mac OS X Copyright 1997-2003,  
> distributed.net
> RC5-72 PowerPC assembly by Malcolm Howell and Didier Levet
> Enhancements for 604e CPUs by Roberto Ragusa
> RC5-72 Altivec assembly by Didier Levet
> Please visit http://www.distributed.net/ for up-to-date contest  
> information.
>
>
> dnetc v2.9006-485-CTR-03072100 for Mac OS X (Darwin 6.7.5).
> Please provide the *entire* version descriptor when submitting bug  
> reports.
> The distributed.net bug report pages are at  
> http://www.distributed.net/bugs/
>
> [Sep 09 13:47:23 UTC] Automatic processor type detection found
>                       a PowerPC 970 (G5) processor.
> [Sep 09 13:47:23 UTC] OGR: using core #0 (GARSP 5.13 Scalar).
> [Sep 09 13:47:42 UTC] OGR: Benchmark for core #0 (GARSP 5.13 Scalar)
>                       0.00:00:16.67 [15,079,652 nodes/sec]
> [Sep 09 13:47:42 UTC] OGR: using core #2 (GARSP 5.13 Vector).
> [Sep 09 13:48:01 UTC] OGR: Benchmark for core #2 (GARSP 5.13 Vector)
>                       0.00:00:16.51 [8,919,441 nodes/sec]
> [Sep 09 13:48:01 UTC] RC5-72: using core #3 (MH 2-pipe).
> [Sep 09 13:48:20 UTC] RC5-72: Benchmark for core #3 (MH 2-pipe)
>                       0.00:00:16.86 [3,490,094 keys/sec]
> [Sep 09 13:48:20 UTC] RC5-72: using core #4 (KKS 2-pipe).
> [Sep 09 13:48:39 UTC] RC5-72: Benchmark for core #4 (KKS 2-pipe)
>                       0.00:00:16.58 [3,750,378 keys/sec]
> [Sep 09 13:48:39 UTC] RC5-72: using core #5 (KKS 604e).
> [Sep 09 13:48:58 UTC] RC5-72: Benchmark for core #5 (KKS 604e)
>                       0.00:00:16.05 [3,541,631 keys/sec]
> [Sep 09 13:48:58 UTC] RC5-72: using core #6 (KKS 7400).
> [Sep 09 13:49:17 UTC] RC5-72: Benchmark for core #6 (KKS 7400)
>                       0.00:00:16.06 [9,985,151 keys/sec]
> [Sep 09 13:49:17 UTC] RC5-72: using core #7 (KKS 7450).
> [Sep 09 13:49:36 UTC] RC5-72: Benchmark for core #7 (KKS 7450)
>                       0.00:00:16.07 [11,948,040 keys/sec]
> [Sep 09 13:49:36 UTC] RC5-72: using core #8 (MH 1-pipe).
> [Sep 09 13:49:54 UTC] RC5-72: Benchmark for core #8 (MH 1-pipe)
>                       0.00:00:16.07 [3,015,752 keys/sec]
> [Sep 09 13:49:54 UTC] RC5-72: using core #9 (MH 1-pipe 604e).
> [Sep 09 13:50:12 UTC] RC5-72: Benchmark for core #9 (MH 1-pipe 604e)
>                       0.00:00:16.08 [3,005,728 keys/sec]
> [Sep 09 13:50:12 UTC] RC5-72: using core #10 (KKS 970).
> [Sep 09 13:50:31 UTC] RC5-72: Benchmark for core #10 (KKS 970)
>                       0.00:00:16.32 [10,077,520 keys/sec]
>
> Sincerely,
>
> t_wolf
>
> _______________________________________________
> rc5 mailing list
> rc5 at lists.distributed.net
> http://lists.distributed.net/mailman/listinfo/rc5
>

_______________________________________________
rc5 mailing list
rc5 at lists.distributed.net
http://lists.distributed.net/mailman/listinfo/rc5


More information about the rc5 mailing list