davidsen@steinmetz.steinmetz.ge.com (William E. Davidsen Jr) (04/02/88)
In article <1975@pasteur.Berkeley.Edu> nfong@cory.Berkeley.EDU (Norman Fong) writes: >Can someone comment on A/UX's performance? Is it comparable to say a >Sun 3/50's performance? Does Apple's current implementation utilize >most of a disk's bandwidth or will future scsi drivers improve performance? >Also does A/UX have 4.3BSD enhancements to the file system? (cylinder grps etc) > >nfong@cory.Berkeley.EDU.UUCP I posted this a few days ago, but no one seems to have gotten it at other sites, so here it comes again. If this appears twice at some sites, I apologize, it did not seem to get out. This is the output of tbl and troff, edited to take out the control sequences. You may have to print it to make it clear. At this time I can't supply the original benchmarks. Test Vax UNISYS386+387 Sun3/280S Mac II ______________________________________________________________________________________ test conditions cpu/speed | 11/780 | 80386:16 | 68020:25 | 68020:16 fp chip | FPU | 80387:16 | 68881/25 | 68881:16 memory (mb) | 8 | 1 | 16 | 5 disk type | dec | --- | --- | quantum80 rated seek (ms) | <20 | 20 | <20 | 28 O/S | Ultrix | Xenix/386 | SunOS 3 | A/UX ______________________________________________________________________________________ math speed (k ops/sec) short | 308.4 | 1193.6 | 1114.5 | 891.9 long | 503.1 | 1227.8 | 1787.9 | 905.6 float | 128.0 | 341.1 | 181.8 | 145.7 double | 181.5 | 279.1 | 180.2 | 138.4 ______________________________________________________________________________________ Compare and branch (k ops/sec) integer | 293.0 | 827.8 | 1666.7 | 646.6 float | 115.8 | 144.1 | 74.3 | 116.3 ______________________________________________________________________________________ Transcendental functions (k ops/sec) circle points | 537 | 1295 | 4615 | 1786 trig functions | 2550 | 4250 | 7650 | 7650 ______________________________________________________________________________________ Standard benchmarks Dhrystone | 1453 | 5319 | 7444 | 3226 Dhampstone | 1430 | 6207 | 8640 | 3661 ______________________________________________________________________________________ bit ops. (vals/sec) Turing machine | 48.0 | 160.1 | 212.3 | 113.6 BTG | 10084 | 68966 | 114286 | 54545 GTB | 2652 | 14085 | 27273 | 13953 Sieve | 801 | 3621 | 3650 | 2122 ______________________________________________________________________________________ Sort (sec) integer | 1.073| 0.413| 0.243| 0.510 float | 1.117| 0.760| 0.843| 1.567 ______________________________________________________________________________________ disk performance (kb) write | 20.4 | 117.0 | 687.7 | 41.9 read | 22.4 | 54.7 | 613.2 | 45.2 access (ms) | 123.5 | 6.6 | 34.9 | 17.1 ______________________________________________________________________________________ System performance (k/sec) xfer via pipes | 269.5 | 1066.7 | 614.4 | 558.5 funct call | 46.9 | 277.8 | 500.0 | 200.0 system call | 4.7 | 11.3 | 17.4 | 7.8 ______________________________________________________________________________________ Compilation speed 20 small real | 479.5 | 277.2 | 99.4 | 548.8 CPU | 144.5 | 100.9 | 45.0 | 142.6 3 large | | | | real | 646.8 | 90.6 | 31.4 | 170.2 CPU | 76.3 | 39.7 | 21.5 | 93.4 -- bill davidsen (wedu@ge-crd.arpa) {uunet | philabs | seismo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me
gnu@hoptoad.uucp (John Gilmore) (04/03/88)
davidsen@steinmetz.steinmetz.ge.com (William E. Davidsen Jr) wrote: > At this time I > can't supply the original benchmarks. My advice is: ignore this benchmark. If you can't get the sources, it's hard to tell whether the benchmark is representative of what you really want to measure. Some of the numbers look pretty questionable to me, e.g. Sun-3/280 can do 7,650,000 trig functions per second, but only 74,300 float compare-and-branches? Also note that the Sun being compared is the highest end 68020 based Sun (25MHz, big cache, pricey). -- {pyramid,pacbell,amdahl,sun,ihnp4}!hoptoad!gnu gnu@toad.com "Don't fuck with the name space!" -- Hugh Daniel
jwhitnel@csi.UUCP (Jerry Whitnell) (04/05/88)
In article <10209@steinmetz.steinmetz.ge.com> davidsen@kbsvax.steinmetz.UUCP (William E. Davidsen Jr) writes: >This is the output of tbl and troff, edited to take out the control >sequences. You may have to print it to make it clear. At this time I >can't supply the original benchmarks. A little cleanup work with the editor... > Test Vax UNISYS386+387 Sun3/280S Mac II ____________________________________________________________________________ test conditions cpu/speed | 11/780 | 80386:16 | 68020:25 | 68020:16 fp chip | FPU | 80387:16 | 68881/25 | 68881:16 memory (mb) | 8 | 1 | 16 | 5 disk type | dec | --- | --- | quantum80 rated seek (ms) | <20 | 20 | <20 | 28 O/S | Ultrix | Xenix/386 | SunOS 3 | A/UX _______________________________________________________________________________ math speed (k ops/sec) short | 308.4 | 1193.6 | 1114.5 | 891.9 long | 503.1 | 1227.8 | 1787.9 | 905.6 float | 128.0 | 341.1 | 181.8 | 145.7 double | 181.5 | 279.1 | 180.2 | 138.4 _______________________________________________________________________________ Compare and branch (k ops/sec) integer | 293.0 | 827.8 | 1666.7 | 646.6 float | 115.8 | 144.1 | 74.3 | 116.3 _______________________________________________________________________________ Transcendental functions (k ops/sec) circle points | 537 | 1295 | 4615 | 1786 trig functions | 2550 | 4250 | 7650 | 7650 _______________________________________________________________________________ Standard benchmarks Dhrystone | 1453 | 5319 | 7444 | 3226 Dhampstone | 1430 | 6207 | 8640 | 3661 ________________________________________________________________________________ bit ops. (vals/sec) Turing machine | 48.0 | 160.1 | 212.3 | 113.6 BTG | 10084 | 68966 | 114286 | 54545 GTB | 2652 | 14085 | 27273 | 13953 Sieve | 801 | 3621 | 3650 | 2122 _______________________________________________________________________________ Sort (sec) integer | 1.073| 0.413| 0.243| 0.510 float | 1.117| 0.760| 0.843| 1.567 ________________________________________________________________________________ disk performance (kb) write | 20.4 | 117.0 | 687.7 | 41.9 read | 22.4 | 54.7 | 613.2 | 45.2 access (ms) | 123.5 | 6.6 | 34.9 | 17.1 ________________________________________________________________________________ System performance (k/sec) xfer via pipes | 269.5 | 1066.7 | 614.4 | 558.5 funct call | 46.9 | 277.8 | 500.0 | 200.0 system call | 4.7 | 11.3 | 17.4 | 7.8 ________________________________________________________________________________ Compilation speed 20 small real | 479.5 | 277.2 | 99.4 | 548.8 CPU | 144.5 | 100.9 | 45.0 | 142.6 3 large | | | | real | 646.8 | 90.6 | 31.4 | 170.2 CPU | 76.3 | 39.7 | 21.5 | 93.4 >-- bill davidsen (wedu@ge-crd.arpa) > {uunet | philabs | seismo}!steinmetz!crdos1!davidsen >"Stupidity, like virtue, is its own reward" -me
benoni@ssc-vax.UUCP (Charles L Ditzel) (04/05/88)
In article <4309@hoptoad.uucp>, gnu@hoptoad.uucp (John Gilmore) writes: > davidsen@steinmetz.steinmetz.ge.com (William E. Davidsen Jr) wrote: > > At this time I > > can't supply the original benchmarks. I am always(!) suspicious of people that don't supply the benchmark code. > Some of the numbers look pretty questionable to me, e.g. Sun-3/280 can > do 7,650,000 trig functions per second, but only 74,300 float > compare-and-branches? Also note that the Sun being compared is the > highest end 68020 based Sun (25MHz, big cache, pricey). Actually some benchmarks have already emanated from the Mac group and they compared the lowest end Sun 3/50...they showed an A/UX Mac II do be slow. These were posted by : fnf@fishpond.UUCP (Fred Fish) *Ok, with all the speculation about disk I/O, and the advantages/disadvantages *of DMA, I decided to drag out and dust of a disk performance benchmark written *by Rick Spanbauer a LONG time ago and used to test Amiga hard disks when they *first became available. Since I already had Sun 3/50 timings, all I had to *do were the A/UX ones. Here are my measured results using the diskperf.c *program, and ballpark verified using the Unix dd program: *Performance timings using Rick Spanbauer's diskperf.c program. * Amiga Amiga Mac-II Sun * Floppy CLtd A/UX 3/50 * df1: dh0: HD80SC * *File creations (files/sec) <=1 7 6 6 *File deletions (files/sec) 1 15 8 11 *Directory scan (entries/sec) 38 50 371 350 *Seek+read (seek+read/sec) 2 40 110 298 *Read speed, 512 buffer (byte/sec) 11014 17133 55168 240499 *Read speed, 4096 buffer (byte/sec) 12024 17133 53708 234057 *Read speed, 8192 buffer (byte/sec) 12080 17133 54013 233189 *Read speed, 32768 buffer (byte/sec) 12136 17133 53644 236343 *Write speed, 512 buffer (byte/sec) 4974 12603 44181 215166 *Write speed, 4096 buffer (byte/sec) 5180 13512 47211 182466 *Write speed, 8192 buffer (byte/sec) 5170 13653 46832 179755 *Write speed, 32768 buffer (byte/sec) 5190 13797 46930 187580 * *Notes: * (1) All Amiga tests done under 1.2 release 33.46. * (2) df1: tests done after "addbuffers 32" & fresh formatted disk * (3) All Amiga and Mac-II timings done by Fred Fish. * (4) Sun-3/50 timings by Rick Spanbauer. As can be seen the low-end 3/50 outperforms with relative ease the Mac II. Of course, this is not the whole story. These are the I/O benchmarks. Other benchmarks would shed a little more light on the subject. (drystone, whetstone, etc.). Given Apple's rather cavalier attitude toward Unix and it's distribution... If you don't care about Mac software both the Sun 3/50 and 3/60 make better Unix platforms and cost less or about the same...in addition the X11/NeWS environment will soon be a standard part of Suns. ----------------------- Naturally My Opinions Are My Own.
davidsen@steinmetz.steinmetz.ge.com (William E. Davidsen Jr) (04/06/88)
In article <4309@hoptoad.uucp> gnu@hoptoad.uucp (John Gilmore) writes: >Some of the numbers look pretty questionable to me, e.g. Sun-3/280 can >do 7,650,000 trig functions per second, but only 74,300 float Obviously one of the values I marked in k isn't... I went in by hand and edited the printer escape sequences out of the troff output so people could read it. When there were a lot of escapes I retyped things by hand. >compare-and-branches? Also note that the Sun being compared is the >highest end 68020 based Sun (25MHz, big cache, pricey). >-- If you'll send me the Sun you want tested I'll do it. The machines of interest here are the 386 vs. 68020, Xenix/386 vs. AU/X performance issues. Not which is "better" but what the strengths of each may be. I included the VAX and Sun as examples of other common machines. I'm sorry if you were unable to gain information from the data, several other people have sent mail saying that it was of interest. I didn't publish the entire manual with one table, not do I intend to. -- bill davidsen (wedu@ge-crd.arpa) {uunet | philabs | seismo}!steinmetz!crdos1!davidsen "Stupidity, like virtue, is its own reward" -me
joe@tekbspa.UUCP (Joe Angelo) (04/08/88)
in article <1837@ssc-vax.UUCP>, benoni@ssc-vax.UUCP (Charles L Ditzel) says: > > As can be seen the low-end 3/50 outperforms with relative ease the > Mac II. Of course, this is not the whole story. These are the I/O > benchmarks. ... I've yet had time to run real bmarks on our MacII A/UX, but from just tinkering and fussing around ... i've can instinctively say that MacII A/UX is horribly slow in regards to process switching and disk utilitzation in ''init 2'' (ie: when more then 3 procs. are running; I betcha those Apple bmarks were compiled in init 1 as well!! Strangely enough, I find the performance of an NFS mount point (from a SUN 3/280 to a MacII) rather acceptable. Why do I have a feeling that them apples and oranges (emun) programmers did all of thier work via NFS? (I usually test NFS perf. by dumping an NFS mounted dir struct to /dev/null on our main machine; yes, i take into account the current network load.) I've only had A/UX up (here) for the past day and already have a six page ``feature'' list. This entire A/UX thing has really blown my mind. There is a VERY strange mixture of OS's and commands and a large number of commands don't generate the output one would expect. This is MY personal opp. -> I'm not impressed Apple. But who cares? Surely I won't blindly buy 300 of them for some phantom accounting department... -- "I'm trying Joe Angelo -- Senior Systems Engineer/Systems Manager to think at Teknekron Software Systems, Palo Alto 415-325-1025 but nothing happens!" uunet!tekbspa!joe -OR- tekbspa!joe@uunet.uu.net
jackie@Apple.COM (Hernan'Jackie' Macapanpan) (04/08/88)
In article <171@tekbspa.UUCP>, joe@tekbspa.UUCP (Joe Angelo) writes: > I've yet had time to run real bmarks on our MacII A/UX, but from just > tinkering and fussing around ... i've can instinctively say that > MacII A/UX is horribly slow in regards to process switching and disk > utilitzation in ''init 2'' How much physical memory are you running with ? > I've only had A/UX up (here) for the past day and already have > a six page ``feature'' list. This entire A/UX thing has really > blown my mind. There is a VERY strange mixture of OS's and > commands and a large number of commands don't generate the output > one would expect. Strange? Are you a BSD or SysV user? What kinda system are you used to running on anyway? :-) ---------------------------------------------------------------------------- Disclaimer: I works heres, buts theys don'ts knows I's cans types. Hernan 'Jackie' Macapanpan amdahl \ Technical Communications/A/UX Hotline pyramid!sun - apple!jackie Apple Computer, Inc. (408) 996-1010 decwrl / ----------------------------------------------------------------------------
rob@uokmax.UUCP (Robert K. Shull) (04/09/88)
In article <171@tekbspa.UUCP> joe@tekbspa.UUCP (Joe Angelo) writes: >in article <1837@ssc-vax.UUCP>, benoni@ssc-vax.UUCP (Charles L Ditzel) says: >> >> As can be seen the low-end 3/50 outperforms with relative ease the > >a six page ``feature'' list. This entire A/UX thing has really >This is MY personal opp. -> I'm not impressed Apple. But who cares? > happens!" uunet!tekbspa!joe -OR- tekbspa!joe@uunet.uu.net This isn't a flame at anyone in particular, just something I thought needed to be said: Give it some time! You're using the first release of a completely new OS. Did anyone use SunOS 1.2? The "feature" list could have been more like 60 or 70 pages. It's taken a BUNCH of releases to make the fixes, fix the fixes, fix those fixes, etc. What made you expect that A/UX would be perfect? Have you ever used a major piece of software that was perfect in its first release? I'm not saying we should blindly accept everything, but the tone of most of these message seems VERY accusing. By the way, I'd love to see the source for those benchmarks that were posted recently. I can't come close to the Dhrystone numbers on our Sun 3's. -- Robert K. Shull University of Oklahoma, Engineering Computer Network ihnp4!occrsh!uokmax!rob CIS 73765,1254 Delphi RKSHULL Opinions contained herein in no way reflect those of the University of Oklahoma.