Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.cs.su.oz.au!inferno.mpx.com.au!goliath.apana.org.au!news.syd.connect.com.au!news.mel.connect.com.au!yarrina.connect.com.au!news.mira.net.au!news.vbc.net!alpha.sky.net!winternet.com!mr.net!umn.edu!zib-berlin.de!unlisys!uriela.in-berlin.de!bolzen.in-berlin.de!hahn1.informatik.hu-berlin.de!suncom.rz.hu-berlin.de!zrz.TU-Berlin.DE!cs.tu-berlin.de!informatik.uni-bremen.de!nordwest.pop.de!uniol!news.uni-stuttgart.de!uni-regensburg.de! lrz-muenchen.de!isar.de!augsburg.isar.net!194.45.233.6!roell From: roell@xinside.com (Thomas Roell) Newsgroups: comp.os.linux.development.apps,comp.os.linux.development.system,comp.os.linux.x,comp.os.linux.hardware,comp.os.linux.setup,comp.unix.bsd.386bsd.misc,comp.unix.bsd.bsdi.misc,comp.unix.bsd.netbsd.misc,comp.unix.bsd.freebsd.misc Subject: Re: Sometimes you need X server source (Was: Why to not buy Matrox Millennium) Date: 31 Mar 1996 14:44:24 GMT Organization: X Inside Inc. Lines: 71 Message-ID: <ROELL.96Mar31164424@blah.xinside.com> References: <4ja28c$9da@vixen.cso.uiuc.edu> <4jb67e$eil@solaris.cc.vt.edu> <ROELL.96Mar28184328@blah.xinside.com> <4jhl5p$59t@nntp.ucs.ubc.ca> NNTP-Posting-Host: blah.a.isar.de In-reply-to: napier@theory.chem.ubc.ca's message of 29 Mar 1996 21:36:57 GMT Xref: euryale.cc.adfa.oz.au comp.os.linux.development.apps:14119 comp.os.linux.development.system:20642 comp.os.linux.x:28515 comp.os.linux.hardware:35381 comp.os.linux.setup:48939 comp.unix.bsd.386bsd.misc:440 comp.unix.bsd.bsdi.misc:2976 comp.unix.bsd.netbsd.misc:2741 comp.unix.bsd.freebsd.misc:16612 > This would constitute an extrordinary level of customer support, the likes of > which I have never personally encountered ... I'll give Xinside the benefit of > the doubt here. Just try it out. > >[snip] > >software. They spend huge amounts of money for their graphics card, > >just to get half of the possible performance because they have to run > >XF86. If one is smart instead buy a decently fast graphics card that > >is optimally supported by Accelertaed-X and you will save money and > >get a faster overally solution. I'm just starting to get frustrated > >that even reasonable bright people seem to have a mental barrier to > >realize this. > > Can you qualify this with some benchmark results (or point us the way to some)? > I'm not trying to be the peanut gallery here, but I would be interested to > see some evidence for this. I tried many times before, but I usually get flamed down whenever I post any benchmarks. Either because they didn't count, or because they were simple not allowed to be faster than XF86. For starters, the Millenium is now at over 1Mio Xstones on a P166. In direct compares with XF86 one has to be benchmarking very carefully. I made the observation that people simply do not understand how do benchmark reasonable and stumble over details which render the results useless. Something like using DRAM cards and running them at high resolutions and high color depths so that DRAM bandwidth is saturated by the refresh allone. A simple example might be that XF86 on a S3 968, which is currently their fastest driver, tops out on my P90 at 280k Xstones (1280x1024, 8bpp, 135MHz PixelClock). Xaccel on a Cirrus Logic 5436 same resolution and depth gives me about 290k Xstones. The price difference between those cards is quite remarkable. Another thing which I tend to critice is the use of Xstones in itself. This benchmark tends to measure raw graphics engine performance, rather than a good mix of CPU and graphics engine which is more typical for applications. To get a real indication of how good the X-Server/GraphicsCard combo will be in every days work are the Xmarks. They are not used to often for two reasons. The first one is that it really takes more than 10 minutes to generate the numbers (basically two complete x11perf -all runs). The other reason is that many X-Servers look very bad on this benchmark, since it covers many more areas than xStones. At last one comment about pure speed. I often hear that 150k xStones for example feel fast enough, and everything bejond this will not be noticeable. This is only true if your machine is not doing anything else. Then, right 1000k xStones will not buy you more visible performance. But what if you have a compile-run in the background, or another time-intensive task ? Then a 1000k xStones combination needs only 1/7th of the time to do the GUI than a 150k xStones combination. I'm one of those users who see X mainly as a means of opening multiple xterms and then use emacs/gcc/gdb or whatever to do the real work. And believe me it is then quite noticable when you got a slower (or more CPU intensive) graphics board. - Thomas -- Denver Office THOMAS ROELL /\ Das Reh springt hoch, +1(303)298-7478 X INSIDE INC / \/\ das Reh springt weit, 1801 Broadway, Suite 1710 / \ \/\ was soll es tun, Denver, CO 80202 roell@xinside.com / Oelch! \ \ es hat ja Zeit. -- Denver Office THOMAS ROELL /\ Das Reh springt hoch, +1(303)298-7478 X INSIDE INC / \/\ das Reh springt weit, 1801 Broadway, Suite 1710 / \ \/\ was soll es tun, Denver, CO 80202 roell@xinside.com / Oelch! \ \ es hat ja Zeit.