herzog@dogwalk.Eng.Sun.COM (Brian Herzog, DSGG/DGDO/GSP/MRG/LMNOP) (08/03/90)
In article <4299@tahoe.unr.edu> steve@olympus.UUCP (Stephen Wheatcraft) writes: > [...] >I have been doing some benchmark comparisons of IBM >R-6000, Sun 3/80, Sun 3/280 and Sillicon Graphics >Personal Iris 4-D25TG. [...] I think benchmarks are important, valuable, complicated, and yes, dangerous. I don't wish to engage in benchmark wars, and even if I did, this would be an inappropriate forum. But I would like to point out that you've compared one company's latest and greatest RISC architecture against another's aging CISC architecture. Now, if that were all the second company had to offer, that would be one thing, but it seems to me that to compare an R-6000 against a Sun-3 is no more meaningful than to compare a SPARCsystem against an IBM RT (at least that would be RISC vs RISC ;-). What's the point? Surely one could select configurations that would better approximate an apples-to-apples comparison. Opinions expressed are my own, etc., etc. -- ------------------------------------------------------------------------------ Brian Herzog "Badges? We ain't got no badges. We don't need no herzog@Eng.Sun.COM badges. I don't have to show you any stinking badges!" ...!sun!eng!herzog - Treasure of the Sierra Madre
ddt@convex.COM (David Taylor) (08/04/90)
In article <140112@sun.Eng.Sun.COM> herzog@Eng.Sun.COM (Brian Herzog) writes: >In article <4299@tahoe.unr.edu> steve@olympus.UUCP (Stephen Wheatcraft) writes: >> [...] >>I have been doing some benchmark comparisons of IBM >>R-6000, Sun 3/80, Sun 3/280 and Sillicon Graphics >>Personal Iris 4-D25TG. [...] > >I think benchmarks are important, valuable, complicated, and yes, dangerous. > >I don't wish to engage in benchmark wars, and even if I did, this would >be an inappropriate forum. But I would like to point out that you've >compared one company's latest and greatest RISC architecture against >another's aging CISC architecture. I believe that steve compared the R-6000 to Sun computers because many of us have a good feel for the speed of a Sun machine. However, I feel that it was definitely /not/ wise to compare optimized results to non-optimized results. As for the original question asking for a comparison between the R-6000 and SGI machines ... I'd suggest you consult the SPEC Benchmark Results newsletter of Spring 1990. Each computer runs 10 benchmarks which stress all sorts of different areas of performance (refer to SPEC Newsletter of Winter '90). 4 issues of the newsletter cost $150, if you'd like to subscribe. I don't know whether we, as a subscriber, are allowed to release numbers, but if you'd like to learn them, go ahead and write: SPEC c/o Waterside Assoc. 39510 Paseo Padre Parkway, Suite 350 Fremont, CA 94538 or call: (415) 792-3334 Be sure to ask for the SPEC Newsletter of Winter '90. It's important for understanding the importance of each benchmark. =-ddt-> -- David D. Taylor, Esq, Performance Measurement Intern, Convex. (whew!) (214) 497-4860, ddt@convex.com or ddt@vondrake.cc.utexas.edu Remember, flatulation is only natural.
eugene@wilbur.nas.nasa.gov (Eugene N. Miya) (08/04/90)
In article <140112@sun.Eng.Sun.COM> herzog@Eng.Sun.COM (Brian Herzog) writes: >I think benchmarks are important, valuable, complicated, and yes, dangerous. > >I don't wish to engage in benchmark wars, and even if I did, this would >be an inappropriate forum. Er, which one is? For graphics systems maybe, maybe not. See below. >compare one company's latest and greatest RISC architecture against >another's aging CISC architecture. .... Taken to the logic extreme you can compare ENIACs. >What's the point? Surely one could select configurations that would better >approximate an apples-to-apples comparison. The problem with seeking "apples to apples" comparisons that that after a while you are debating "MacIntosh versus Gold Delicious." You end up seeking more detail to form an opinion. I think this is an appropriate forum, however, we'd have to come to some consensus on methodology. I think there are a lot of people interested in performance, but few of them willing to do any work (much of which will be error prone and there is lots of it). The basic issues are "what constitutes 'equivalence?'" "how does something compose/decompose?" When some group of people can reach a consensus on these for graphics systems (considering their diversity from the Apple II to the top of the line E&S flight simulator image generation systems). Then we can compare systems. At least two attempts and one SIGGRAPH session have tried to address some of this without success. Bogged down. There is an economic advantage for some parties to keep people confused and in the dark. Consensus is how international systems of measurement work. >Opinions expressed are my own, etc., etc. Ditto. --e. nobuo miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov {uunet,mailrus,other gateways}!ames!eugene