Re: What now, tech? (Bill Schottstaedt )


Subject: Re: What now, tech?
From:    Bill Schottstaedt  <bil(at)CCRMA.STANFORD.EDU>
Date:    Fri, 20 Dec 1996 09:46:41 -0800

> I would like to know how fast the O2 really is. We recently got a couple O2's and I had a chance today to run my timing tests. These are CLM-related, so insert all the usual benchmark qualifications and caveats here. The first group relates to a simple C program doing IO, ffts, table lookups; the second relates to CLM running some standard tests. ---- * machine read/write: fread/fwrite: arithmetic: * 256 512 8192 65536 same sizes tbl bigfft sffts * * NeXT 68040 (32MB): 11575 10514 10256 9943 11951 11923 12358 12259 10478 108122 26622 * NeXT Turbo (16MB): 8329 7760 6933 6833 9216 8742 9416 9238 7825 121591 19495 * HP 90MHz Pentium NextStep: 11970 10069 9840 9920 11930 11209 11399 11540 1930 46389 4019 * Mac 8500 120 MHz PPC: 21733 15416 5000 2916 9566 9550 9733 9850 * SGI R4600 132 MHz Indy (32MB): 2412 1619 959 1045 1172 1174 1111 1126 1224 30825 3490 * SGI R5000 150 MHz Indy (32MB): 1067 846 684 737 847 817 734 791 885 25878 1591 * HP 166 MHz Pentium Linux: 983 727 811 878 1060 1073 1191 1181 878 * SGI R5000 180 MHz O2 (64MB): 1359 788 431 446 1919 1944 1891 1885 828 24658 1390 ---- machine 1. 2. 3. 4. 5. 6. 7. 8. 68040 NeXT 56k gens: 4.9 9.5 2.4 1.3 116 482 10.3 68040 NeXT with Ariel QP (2 QPs): 4.6 6.9 2.4 1.3 29 199 (117) 9.5 68040 NeXT C gens: 3.8 4.7 2.0 1.3 94 298 9.5 19644 166 MHz Pentium GCL NextStep: .47 .8 .3 .18 13 52 .9 PPC 8500/120, 32 MB, MacOS 7.5.3: .78 1.1 .4 1.5 13.5 61 1.4 R4600SC SGI Indy GCL (133 Mhz): 1.4 .7 .3 .3 13 38 1.1 2328 R5000 SGI Indy ACL (150 Mhz): .34 .54 .24 .18 10.3 31 .86 2065 200 MHz Pentium Pro/Openstep/GCL: .23 .55 .24 .21 9.0 43 .65 R5000 SGI O2 ACL (180 MHz): .26 .49 .22 .19 7.9 24 .63 1526 1. sample rate conversion and "granular synthesis" 2. reverberated fm note 3. just the single fm note 4. a few short sound file fragments mixed 5. a bunch of fm notes 6. a portion of the piece "idyll" (fm+reverb) 7. spectral analysis followed by additive synthesis 8. fifth movement of "Colony" ---- The 200MHz pentium case above does not coincide completely with the O2 case -- I hope to get Linux and ACL running soon on such a machine for a fairer comparison. In any case, everything I've tried has run fine on the O2, and I'm pleased with its performance.


This message came from the mail archive
http://www.auditory.org/postings/1996/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University