Re: 500mhz alpha

Jim Nance (jlnance@avanticorp.com)
Thu, 5 Dec 1996 11:38:01 -0500 (EST)

Messages sorted by: [ date ][ thread ][ subject ][ author ]
Next message: Stephen J. Gaudet: "Re: 500mhz alpha"
Previous message: Robert G. Smith: "Re: Using Your Jaz Drive on AXP"
In reply to: Miguel Avillez: "is there any problem with tar?"

> I would like to see the Fhourstones benchmark on your 500mhz
> machines with 1mb and 2mb cache to see the performance difference.
> This benchmark is available from
> ftp://ftp.nosc.mil/pub/aburto/c4.shar
> and results are posted in ftp://ftp.nosc.mil/pub/aburto/c4.tbl

I just got a new AlphaServer 4100, so I decided to try that out. The results
are included below. It should be interesting to compare this to the the
500 MHz systems. They have an extra 20% clock speed, but I have more cache.
These were compiled under Digital Unix using the cc compiler. I have
also included results using the gcc compiler. Looks like gcc is about 13%
slower, which is not too bad, but kind of annoying. The last one has been
compiled with the -cord flag and feedback information from a prior run.

zoom> ./c4 < input
Fhourstones 1.0
options:
Solving 12-ply position after 444333377755 . . .
score = 7 (+) work = 21
3871322 pos / 8.1 sec = 479.1 Kpos/sec
store rate = 0.914
- 0.211 < 0.422 = 0.059 > 0.054 + 0.254
56765 376940 173046 174816 121323 70603 38079 19252
9741 4853 2393 1147 537 291 116 54
30 16 3 2 3 1 0 0
0 0 0 0 0 0 0 0

zoom> ./c4.64 < input
Fhourstones 1.0
options: 64-bit modulo
Solving 12-ply position after 444333377755 . . .
score = 7 (+) work = 21
3871322 pos / 7.6 sec = 507.3 Kpos/sec
store rate = 0.914
- 0.211 < 0.422 = 0.059 > 0.054 + 0.254
56765 376940 173046 174816 121323 70603 38079 19252
9741 4853 2393 1147 537 291 116 54
30 16 3 2 3 1 0 0
0 0 0 0 0 0 0 0

zoom> ./c4.64.gcc <input
Fhourstones 1.0
options: 64-bit modulo
Solving 12-ply position after 444333377755 . . .
score = 7 (+) work = 21
3871322 pos / 8.6 sec = 449.1 Kpos/sec
store rate = 0.914
- 0.211 < 0.422 = 0.059 > 0.054 + 0.254
56765 376940 173046 174816 121323 70603 38079 19252
9741 4853 2393 1147 537 291 116 54
30 16 3 2 3 1 0 0
0 0 0 0 0 0 0 0

zoom> ./c4.64.cord <input
Fhourstones 1.0
options: 64-bit modulo
Solving 12-ply position after 444333377755 . . .
score = 7 (+) work = 21
3871322 pos / 7.4 sec = 524.9 Kpos/sec
store rate = 0.914
- 0.211 < 0.422 = 0.059 > 0.054 + 0.254
56765 376940 173046 174816 121323 70603 38079 19252
9741 4853 2393 1147 537 291 116 54
30 16 3 2 3 1 0 0
0 0 0 0 0 0 0 0

Jim

--
To unsubscribe: send e-mail to axp-list-request@redhat.com with
'unsubscribe' as the subject.  Do not send it to axp-list@redhat.com

Next message: Stephen J. Gaudet: "Re: 500mhz alpha"
Previous message: Robert G. Smith: "Re: Using Your Jaz Drive on AXP"
In reply to: Miguel Avillez: "is there any problem with tar?"