shawn@jdyx.UUCP (Shawn Hayes) (03/12/90)
Since everyone is interested in comparing the RS 6000 family to the other families of machines that are available here is my contribution to muddy the waters. The following numbers are the results of the Byte Unix Benchmark suite running on the 8943Q release of AIX 3.0 using a RS 6000/320. These results are very close to what I got using the 9005 release so I don't expect much change in the release version of AIX. Everex Next DEC 3100 HP 370 RS 320 HP 360 dhry2 13847 5921 23077 16073 44021 10558 dhry2reg 14634 5932 23077 16003 44050 10652 arithoh 0.72 0.53 0.22 0.21 0.10 0.27 register 2.92 6.83 3.52 4.92 1.32 6.55 short 3.52 6.13 3.52 4.92 1.42 6.53 int 3.12 6.82 3.52 4.90 1.35 6.53 long 3.12 6.82 3.52 4.90 1.32 6.53 float 11.92 13.02 2.43 11.27 1.60 14.47 double 13.22 15.63 2.02 10.85 1.50 14.97 sysoh 1.10 2.03 0.63 0.90 1.00 1.28 pipe 0.92 2.72 0.93 1.70 0.80 2.02 context 0.63 0.93 0.40 1.22 0.50 1.10 create 1.23 2.22 0.52 0.98 0.70 1.15 execl 3.43 3.03 1.27 2.12 1.80 2.53 shell1 4.06 8.33 4.30 6.40 2.11 3.98 shell2 5.80 11.57 4.47 8.52 3.67 7.19 shell4 9.60 18.93 7.33 14.57 7.40 13.62 shell8 17.30 33.60 13.00 26.58 13.96 27.98 cc 2.08 5.68 3.61 5.03 2.41 4.51 dc 0.63 0.65 0.48 0.53 0.76 tower 0.56 1.63 0.42 0.62 0.30 0.92 A couple of points about the above benchmarks 1) the Everex machine is an Everex 33 MHz 386 running SCO Xenix 366 V2.3.1 with 4 meg and a 80387 math co-processor 2) The dhrystone test for the RS 320 is based on a HZ of 100 3) The dhrystone test for the HP 360 is based on a HZ of 60 4) The dc test(which test the calculator) didn't work on version 8943Q of AIX 3.0. I don't have my results with me, but I recall the results of the dc test with the 9005 version being approx. 0.50 secs. This number is very likely wrong since I'm very hazy on the exact number. In addition BYTE rates the machines based on the results of a few of the benchmarks when comparing the computers against the Everex 386. The benchmarks are dhry2, float, shell8, cc, dc and tower. The results for each of the computers are: Everex 6.00 NeXT 3.54 DEC 3100 11.12 HP 370 5.37 HP 360 4.10 RS 320 14.59 (note this is without the dc test times) In addition there are other tests that measure file-system throughput and other items that a potential buyer might want to know. If there is enough interest I can post the entire results. PS: I have a letter stored off where someone offered to put the benchmark source up for FTP if I sent it to him. I should be sending it out sometime in the next couple of days so it will be available this week for anyone else who wants to test the IBM machine or any others. Shawn Hayes
nghiem@ut-emx.UUCP (Alex Nghiem) (03/12/90)
In article <1990Mar12.045655.12943@jdyx.UUCP> shawn@jdyx.UUCP (Shawn Hayes) writes: >PS: I have a letter stored off where someone offered to put the benchmark >source up for FTP if I sent it to him. I should be sending it out sometime >in the next couple of days so it will be available this week for anyone else >who wants to test the IBM machine or any others. I tried to send mail, but apparently it got lost somewhere in the ether... I tried to download the latest source from ByteNet and Fiver BBS, but my modems would not sync, it seems, because of extremely poor line quality. Is there any way I can get the source from you before Wednesday, Mar. 14? Thanks for the comprehensive posting. __________________________________________________________________________ This article is posted for factual information only. Any misrepresentation, if any, is purely unintentional. Any opinion expressed or implicit in these remarks are solely my own. nghiem@emx.utexas.edu !cs.utexas.edu!ut-emx!nghiem nghiem@walt.cc.utexas.edu !cs.utexas.edu!ut-emx!walt!nghiem
nghiem@ut-emx.UUCP (Alex Nghiem) (03/13/90)
In article <1990Mar12.045655.12943@jdyx.UUCP> shawn@jdyx.UUCP (Shawn Hayes) writes: >In addition there are other tests that measure file-system throughput and other >items that a potential buyer might want to know. If there is enough interest >I can post the entire results. Yes, please post the results. High CPU benchmarks are good, but high filesystem and ethernet benchmarks are better. __________________________________________________________________________ This article is posted for factual information only. Any misrepresentation, if any, is purely unintentional. Any opinion expressed or implicit in these remarks are solely my own. nghiem@emx.utexas.edu !cs.utexas.edu!ut-emx!nghiem nghiem@walt.cc.utexas.edu !cs.utexas.edu!ut-emx!walt!nghiem