claus@sinix.UUCP (Claus Gittinger) (04/15/89)
Since many people requested them, here are the xbench results I've collected so far: Feel free to run xbench on your machine (if its not already in the list or if it's rating is better/different) and mail the .out file to claus@sinix. I'll continue to collect the data and publish the results from time to time. If there is enough interest I can also provide the individual results (xbench .out-files) and put them into comp.I.dont.know. many thanks to: john Carr at MIT for ibm, dec, hp and sun3 data some guy at univ. of dortmund for sony, sun3 and sun386i data juergen at univ. of karlsruhe for sun3/sun4 data please notice that some of the listed machines/servers are preliminary versions and still changing. (I always took the best rating I got and will continue to do so.) have fun Claus Claus Gittinger System Software Consultant currently at Siemens AG Munich claus@sinix ------------ cut ------------ .RP .TL xbench - results .AF "" .AI Siemens AG Munich Dep. D ST SP 4 .NH X Server performance overview .LP All ratings are relative to a monochrome SUN 3/50 running X11R3 communicating via unix sockets. The ratings are scaled to give this reference machine an .I xStone rating of 10000. A machine with double performance will therefore get 20000 .I xStones; a machine with half performance will get 5000 .I xStones. .LP The value in the .B line field is the .I lineStone rating for solid-line, dashed-line, wide-line and rectangle performance. .br The value in the .B fill field is the .I fillStone rating for solid-, tile- and stipple-rectangle-fill performance. .br The value in the .B blit field is the .I blitStone rating for invrects, screencopy, scroll, and bitmapcopy performance. .br The value in the .B arc field is the .I arcStone rating for arc- and filledarc performance. .br The value in the .B R field is the X server release. .br The value in the .B p field is the number of color/greyscale planes. Machines with a '1' in the .B p field are monochrome. .sp .TS .box, tab(!); l | c | c | c | c | c | c | c | c | c | c l | r | r | c | r | r | r | r | r | r | r. machine!R!p!comm!line!fill!blt!text!arc!cmplx!xstones = siemens 9733P!2!8!unix socket! 79244! 16235! 20143! 30625! 609869! 20718!26488 _ siemens 9733P!2!8!10mb ether! 71934! 15481! 18671! 30000! 611105! 18594!25026 _ siemens 9733L!2!8!unix socket! 53845! 15931! 18036! 27812! 249723! 15163!23383 _ siemens 9733L!2!8!10mb ether! 52651! 16143! 18481! 27812! 242551! 12418!22960 _ sony news 1850!1!8!unix socket! 16514! 14477! 25208! 21656! 598290! 37124!20481 _ sony news 1850(gcc)!1!8!unix socket! 16479! 14146! 25893! 21656! 600880! 37124!20424 _ IBM PC/RT!2!1!unix socket! 26964! 10828! 15934! 10937! 387039! 20257!15486 _ Sun 3/50 (FPU)!3!1!unix-socket! 13846! 11905! 14552! 13437! 43130! 10522!13448 _ SUN 3/60!3!1!unix socket! 13938! 12499! 16577! 13437! 6287! 15163!13343 _ SUN 3/110!3!1!unix socket! 11711! 10663! 14266! 11250! 5094! 12418!11229 _ IBM PC/RT Megapel!2!8!unix socket! 63908!*12240! 3162! 34718! 375060! 26405!11160 _ sun 3/160!2!1!unix socket! 11536! 10618! 9021! 11250! 233545! 10784!11035 _ Sun 3/50!3!1!unix-socket! 10000! 10000! 10000! 10000! 10000! 10000!10000 _ VS 2000!2!1!unix socket! 8727! 5788! 8820! 18160! 218701! 8082!9565 _ DEC gpx!2!8!unix socket! 4835! 7892! 5710! 30937! 390212! 5490!8250 _ VS II Parallax!3!8!unix socket! 18670! 7062! 9534! 3819! 32752! 7123!6830 _ siemens x20!2!1!unix socket! 7948! 4311! 5163! 9166! 184220! 7843!6710 _ HP 9000/320!3!1!unix socket! 3532! 5252! 10999! 10000! 11695! 8235!6424 _ VS II qvss!2!1!unix socket! 4780! 2945! 3587! 12554! 205043! 5128!5044 _ siemens ws30-430!?!1!!*15532! 9731! 2240!*?! 127340! 3856!4210 _ visual 640!2!1!10Mb ether! 5271! 3474! 1978! 7668! 25453! 3333!3917 _ Sun4/260!3!8!unix socket! 3861! 789! 847! 3966! 46667! 3267!1644 _ sun386i/250!3!8!unix socket! 2021! 452! 1334! 3191! 67722! 5032!1314 _ sun386i/260!3!8!10mb ether! 2011! 450! 1291! 3222! 70346! 4640!1301 .TE .LP notes: .br (*) incomplete test - rating computed by taking the performance of the existing tests as the .B expected performance of the missing tests. .br .LP
bob@tinman.cis.ohio-state.edu (Bob Sutterfield) (04/18/89)
First off, for readability's sake, what *roff macro package is this written for? `tbl|ptroff -ms' got the text and headers right, but `tbl|ptroff -me' got the table right. What compiler was used to gather the reference Sun-3/50 data - Sun's CC or GNU CC? What version of GNU CC? Under what version of the operating system - SunOS 3.5.1 or SunOS 4.0.1? Were the Purdue 2.1 speedups installed? What compiler was used on the other machines measured - the (usually) pcc-based one the vendor shipped, or GNU CC? In the case of the color machines, did you use Spaf's preliminary hacks to cfbmskbits.h? These are nifty numbers; now I'd like to know what I'm looking at!