ULTIMA VIII BENCHMARK RESULTS

Bernd Bruegmann, 5/13/94

Are you playing U8?
Are you unhappy with the general speed of U8 on your system?
Are you wondering how U8 performs speedwise on other systems?

Well, I am. When I started U8, I was not too happy with the speed (still am), and wanted to find out how well U8 runs on various systems. You hear many comments disqualifying the game as unplayable on a 486DX2/66, others still seem to like it on their 386DX/33.

Let's have some numbers!

That's what I asked for in comp.sys.ibm.pc.games.rpg, and that's what I got. Here are the results for a little benchmark that I made up for testing the speed of U8 on various systems and under varying conditions.

On the first line you find for each system a brief description:
CPU/frequency/cache/model
main RAM/used for disk-cache like smartdrv
harddisk/size/interface/controller disk-cache/compression
videocard/interface
For disk caching, r stands for read, w for write caching.

There follow several lines of timing information:
rating/potential rating
benchmark 1: saving/loading
benchmark 2: animation
benchmark 3: scrolling

Rating is an integer, 10 for best, 0 for worse. The first rating refers to how one rates U8 as is. For the second rating, I asked to imagine how one would rate U8 if it ran smoothly (without time delays and jumps) most of the time.

The idea of the benchmarks is to time (with a stop-watch) representative events at the very beginning of the game. Saving/loading refers to saving and loading the game position. Animation characterizes how smoothly figures are animated. The time quoted refers to the end of the execution sequence where several figures move off screen. For scrolling speed we take the time for a sprint through Tenebrae.


The Data

The benchmark times are in seconds: the smaller, the better. Standard game options are no sound, animation on, frame limiter on. The results are ordered as

  rating/pot.rating   saving/loading   animation   scrolling


Pentium/60/256kw/Comtrade 16M Quantum/540M/SCSI-2/No       DiamondViper/PCI
---------------------------------------------------------------------------
  -/-   11/ 3   16   27      
                     14 with frame limiter off 
It takes only 3 sec to save a file in Serpent Isle. From my experience, U8
does not slow down for the most part. It runs at a constant speed on my
system. With music on, there is no appreciable decrease in speed. Rating:
7(8). The parenthesis represents the game without the platform jumping.


486DX2/66/256k/?     8M/3Mr  Quantum/525M/SCSI/Adaptec1542C  ET4000W32/VESA
---------------------------------------------------------------------------
  7/8   20/ 8   15   20      times are approximate


486DX/33/256k/SBC    ?       Maxtor120/64Mpartition/IDE/-       Realtek/ISA
---------------------------------------------------------------------------
  -/9    4/ 7   24   50      with disk caching
  2/9   27/12   24   50      with boot disk (no cache)
I had trouble with the game crashing when I tried to retrieve games so
called Origin and they had me create a boot disk to bypass my caching
system. Maybe I'd have been a little happier camper if it hadn't taken a
week and a half to reload on all those jumping puzzles ...


486DX/33/256k?/DMG   8M/4Mrw/NortonSCPLUS  Connor/127M/IDE/No Tseng4000/ISA
---------------------------------------------------------------------------
  4/6    3/ 6   33   54      with SCPLUS
        39/16   33   58      without Diskcache
Comments on rating/potential rating
  4 (no show, good graphics, lousy plot)
  6 (SHOW, lousy plot, EXCELLENT GRAPHICS)


486SX/33/64k/NICE    8M      Seagate/431M/IDE/-                TVGA9000/ISA
---------------------------------------------------------------------------
  6/8   19/ 9   32   55      no sound (as defined for benchmark)
        36/24                all sounds on
        21/ 9   32   58      1M read smartdrv as recommended in docs
        30/21                dto. all sounds on 


486SX/33/64k/NICE    4M      Seagate/431M/IDE/-                TVGA9000/ISA
---------------------------------------------------------------------------
        21/ 9   32   57      no sound
        38/25   44   71      all sounds on 


486SX/33/64k/NICE    4M      Quantum/50M/IDE/-                 TVGA9000/ISA
---------------------------------------------------------------------------
        42/14   34   71      no sound
        58/33                all sounds on


486SX/33/64k/NICE    4M      Quantum/50M/IDE/dblspace          TVGA9000/ISA
---------------------------------------------------------------------------
        85/17   35   64      no sound
        87/19   45   82      music and steps off, effects on
       100/34   50   91      all sounds on


AM386DX/40/64K      20M/4.8M Seagate/120M/IDE                  TVGA8900/ISA
---------------------------------------------------------------------------
  8/10   5/ 9   31   63      benchmark
        20/23   38   70      all sound, frm skp, limiting     


386DX/25/-/CompuADD  8M      Seagate/210M/IDE/-               Tseng4000/IDE
---------------------------------------------------------------------------
  -/-   30/18   46  102      nothing loaded except mouse driver, and 
                             dblspace, but Ultima 8 was not on a compressed
                             drive (it's on c:, while d: is compressed)
        29/22   49   74      without loading doublespace drivers by 
                             pressing CTL-F8 while booting
        24/15   46   78      loaded smartdrv with a 1M read and write cache
Once, I accidentally had the music, sound fx, and steps on while saving, 
and I noticed it was twice as long as without. 

Notes on interpreting the data

This I can confirm for my 486SX/33 where even just plain walking is quite jerky and `sirupy'. One golden rule of adventuring is to go absolutely everywhere. Therefore, how fast and undetracted by system slow-downs you can get there appears crucial to me. (That's `to crawl jerkily absolutely everywhere'.) Others seem to mind the jumping exercises more.


Many thanks for data and comments to (in no particular order):
vekemans@aurora.unice.fr
erik.mattsson@p3.dolphins.ct.se
dweldy@expert.cc.purdue.edu
toma@sage.cc.purdue.edu
eilan@seas.ucla.edu
RMBailey@aol.com
---------------------------------------------------------------------------
Send me YOUR data! If you can't find the original post where the benchmarks
are defined, email me.
Thanks, Bernd                                  bruegman@iws170.mppmu.mpg.de
---------------------------------------------------------------------------
This Report is Copyright (C) 1993 by Bernd Bruegmann for Game Bytes Magazine. All rights reserved.