Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wuarchive!hsdndev!cmcl2!uupsi!sunic!ericom!exud1.ericsson.se!exudnw From: exudnw@exud1.ericsson.se (Dave Williams) Newsgroups: comp.benchmarks Subject: Re: X benchmarks Message-ID: <1991Jun3.212807.13712@ericsson.se> Date: 3 Jun 91 21:28:07 GMT References: <1991May31.121038.27049@eagle.lerc.nasa.gov> <1991May31.151431.9127@Informatik.TU-Muenchen.DE> Sender: news@ericsson.se Organization: Ericsson Telecom, Stockholm, Sweden Lines: 39 Nntp-Posting-Host: exud1.ericsson.se In article <1991May31.151431.9127@Informatik.TU-Muenchen.DE> roell@informatik.tu-muenchen.de (Thomas Roell) writes: >>x11perf is a test of the majority of X operations (drawing, pixel, window, >>etc.). It has been organized into 4 sections by Digital Review. A script >>to run the benchmark and organize the results is available from >>uunet.uu.net in unix-today/benchmarks. Enclosed is the README from this >>directory. To see an excellent example of standards oriented benchmarking, >>get the HP 700 series benchmark report from HP. > >I think you messed something up. The x11perf is a good analyzing tool, but >for server implementators, not for users. Normally you cannot guess the speed >of a specific server implementation from this bulk of numbers. Also the >x11percompDR is COMPLETLY USELESS, since there the numbers of two test were >compared WITHOUT wightening the resuslts. It's obviousely that painting a >point is quite unimportant compared to scrolling a 500x500 area. Thus the >direct comparishon is misleading and totally wrong. > >The xbench is much better, cause not all primitives were tested, but those >which are *very* important for everydays work. And the results are much >more balanced. From the raw number of xstones you can guess how fast your >server will be at work. And don't joke with the comparing to a sun 3/50. I >think it's quite wise to select a LEVEL 0 for all other tests. Thus you can >guess what the number 20000 xstones means if you worked allready with >another x-server with a well known rating. > >- Thomas IMHO, after running both X11perf, xbench, and a "killer" application on a *wide* variety of X servers and hosts, both x11perf and xbench gave misleading results that were poor indicators of applications performance. If you have a primary application (mine was mechanical CAD/CAE), use it as your benchmark. If you are doing general office automation junk, then either of these standard benchmarks will tell you something, but neither will tell you everything you want to know. -- = exudnw@exurchn1.ericsson.se || dnw@ponder.csci.unt.edu (214)907-7928 = = David Williams = = Ericsson Network Systems = = Richardson, TX 75081 These opinions are my own. =