<H2>Scores list for Comptest v2.59 or v2.60</h2>
<i>Updated 3/16/95</i><p>
<hr>

Some important news:<p>

No Pentium list yet.  Comptest apparently is unable to correctly identify
the level 2 cache on most Pentium systems.<p> <hr> Thanks to everyone who
has submitted results so far.  This list uses Comptest 2.59 or 2.60 to
compare the memory throughput of different 486 chip set/motherboard
combinations.  Comptest 2.59 is available by anonymous ftp at
oak.oakland.edu in the file /pub/msdos/sysinfo/ctest259.zip.  Comptest
2.60 is available via www at http://www.dfw.net/~sdw/ctest260.zip.<p>

Comments and suggestions are appreciated.  A brief description of the
result headings are at the end of this document.<p>

Procedure: Boot off a clean boot floppy to run tests.  Ideally, system files
should be set up as follows.<p>
<pre><b>        
        Config.sys              Autoexec.bat
        ==========              ============
        files=20                prompt $p$g
                                set path=c:\dos
</b></pre>
Run comptest and save results to a file, (e.g. "comptest results.txt")
Send your results to mintz@biology.ucsc.edu<p>

If no cache size is listed then the result is from before I started noting
them, not that there is no 2nd level cache.<p>

<pre><b>
Brand/Model      chip set   CPU       L.1   L.2   Main  Dhry  Cache    Notes
=============================================================================
Intel and AMD DX4/100 CPU, and AMD DX2's clocked at 100 MHz
-----------------
ASUS GX4VLB      SIS85C471  I.DX4/100 64063 39649 30267 46674 
unknown          SIS85C471  A.DX2/100 97403 60082 22285 50831 256
Generic          FuguTech   I.DX4/100 65157 40065 19224 49824 256
ASUS PVISP3      SIS85C496  A.DX2/100 97403 60212 19072 50516 256
ASUS PCISP3      SIS85C496  A.DX2/100 96419 60212 20036 50516 256
????             Saturn?    I.DX4/100 64935 40054 17011 50764  
ASUS  PCISP3G    SaturnII   I.DX4/100 64935 40054 17011 50764 
Unknown          Saturn?    I.DX4/100 42424 28926 21163 40000             1
ASUS PVISP3      SIS85C496  A.DX4/100 64935 40149 14112 48690 256
Gigabyte GA486-VS SIS471    A.DX4/100 64496 40065 12086 48035 256        18

Intel DX2, AMD DX2, and AMD SX2 chips clocked at 80 MHz
-------------------------------------------------------
Acer Vi15G       SIS471     A.DX2/80  75758 47160 29124 39380 256         1
Microstar VLB 486SIS471     A.DX2/80  77605 48058 18641 40414 128        11
Generic          UMC        A.DX2/80  77605 48073 18134 39921 256
Generic          SIS471     A.DX2/80  76979 47922 17861 40800 256
Silicon Star AN4R2 SIS471   A.DX2/80  78241 48065 17848 40547 256
Asus SV2GX4      SIS471     A.SX2/80  77605 48065 17386 40547 256      11,2
Soyo             SIS471     I.DX2/80  77605 48080 17344 40800 256        11
Silicon Star AG4 SIS471     A.DX2/80  76364 38887 16275 40640 256        11
GA-486 VF        SIS471     A.DX2/80  78241 48073 14878 40414 256        11
????      Contaq 82C596     A.DX2/80  37287 23942  6988 34505            11

Comments:  See note about write-through performance.  Avoid the Contaq.


Intel DX2, AMD DX2, Cyrix DX2, and AMD SX2 chips clocked at 66 MHz
-------------------------------------------------------
ASUS 486SV2GX4   SIS471     A.DX2/66  64063 39659 30305 32958 256        1
ASUS 486SV2GX4   SIS471     A.DX2/66  63636 39546 30227 33000 128        1
ASUS 486SV2GX4   SIS471     A.DX2/66  64063 39526 30186 32879 256        1
Soyo Green VLB   SIS471     ?.DX2/66  61189 37902 28946 33166 256        1
Tyan S1437       SIS411     A.DX2/66  63215 39106 33797 33025 256        1

Comment: Most boards seem to perform better on real world applications
with write-back enabled.  The above scores show the SIS chipset in 
through-mode, and they are so much different than the other boards on the 
list I have a feeling that the high main memory scores are an artifact.

Micronics        SeeComment I.DX2/66  61983 39021 24141 32951            8
Micronics        MIC471     I.DX2/66  63636 39136 24165 32695            8
Micronics JX-30  ???        I.DX2/66  63215 39131 24164 33024

Comment: The Micronics board used in the Gateway 2000 DX2/66V.  Looks like
a nice performer, although again I suspect the results.  The dhrystone
numbers look low.

Generic          FuguTech   I.DX2/66  64935 40065 19223 33675 256
Genoa 486PCI-46  Saturn II  I.DX2/66  64935 40028 17600 34093 256
ASUS SP3G        Saturn II  A.DX2/66  65480 40049 17010 34012 256        2
Tyan S1437       SIS411     A.DX2/66  64496 40007 15757 33802 256        2
????             SIS461     I.DX2/66  64063 40673 15106 33881 256
???  486G Rev.E  SIS471     ?.DX2/66  64935 40149 14911 34174 256 
ASUS 486SV2GX4   SIS471     A.DX2/66  64935 40044 14822 33883 256        2
ASUS 486SV2GX4   SIS471     A.DX2/66  64496 40044 14780 34012 256        2
MB-1425/33/40/50UCV Bioteq  A.DX2/66  64935 40007 14869 33188 256
Soyo Green VLB   SIS471     ?.DX2/66  64496 40002 14501 33109 256        2
Silicon Star AG4 SIS471     A.DX2/66  64496 32506 13403 33214      
????             SIS??????  I.DX2/66  63636 39872 13038 33067      
Soyo VESA 486    SIS85C460/1I.DX2/66  64496 40002 12951 33188      
GigabyteGA486S-VS SIS471    CxDX2/66  64496 37183 12412 32133 256
Silicon Star AN4 SIS461     I.DX2/66  64496 32506 12377 33431 256
GA-486 VF        SIS471     A.DX2/66  64935 40049 12103 33267 256
Silicon Star AG4 SIS471     I.DX2/66  64496 32506 12083 33143 256
Gigabyte         SIS471     I.DX2/66  62799 38675 11967 33131 256 

Comments:  How did that Bioteq sneak in there?  :)  Find a board with the 
SIS chip set, then make your decision based on price and features.
Boards built around the SIS chip set are solid performers with a good 
reputation for compatibility.  Saturn II boards are also excellent performers.

???? NAT48PV     OPTi495SLC I.DX2/66  64496 40023 32638 32988            14
????             OPTi495SLC I.DX2/66  62799 38626 30061 32648 256        14
none 3/486WB     OPTi495SLC A.DX2/66  42805 28781  9659 32059 256        13
none             OPTi495SLC ?.DX2/66  64935 40060  9161 33188      
AMI              OPTi       I.DX2/66  62389 35826  8871 33083            

Comments: The third board on this list is my own system.  When you increase
wait states on this board, eventually you get to a point where the comptest
results look very high, as in results #1 and #2, but in reality real world
performance goes down.  This chip set is slow, but boards with it tend to 
be very inexpensive.

F.I.C 486-GVT    VT 486     A.DX2/66  64496 32547 10357 33360 256
Micromax Climax  Cntq 596   A.DX2/66  63215 39649  9181 33109 256 
AMI Enterprise III          I.DX2/66  43388 29000  9339 32940 256
????             Cntq 596   A.DX2/66  31093 19945  6937 29330      

Comments: The VT486 board falls just under the SIS boards in performance.  I 
don't know what to think about the Contaq.  The AMI board is rather aged at 
this point, I believe.

Dell 466/l       Shasta?    I.DX2/66  63215 ----- 24826 34237   0         
Dell Dimension 466/66       I.DX2/66  61189 ----- 24814 34104   0        
IBM ValuePoint 466 DX2/d    I.DX2/66  40793 24858 16105 30474 128
Compaq Deskpro XE 466       I.DX2/66  40619 ----- 21283 29830   0

Comments: Some Brand name systems for comparison.  The Dell and Compaq
systems were not equipped with Level 2 caches.


Intel DX and AMD DX chips clocked at 50 MHz
-------------------------------------------
Generic          ALI        I.DX/50   37142 25794 20199 24387 256
???? VIP-A486    ALI        I.DX/50   38490 25890 17973 23958            1
A.I.R. 486SH     SIS85C461  I.DX/50   64935 39141 16572 25586      
Soyo SY-025I2    SIS471B    A.DX/50   64935 32620 13876 25690            6
????             OPTI495SX  I.DX/50   64496 34042 11910 25650 256
????             SIS471     I.DX/50   64496 32554 11653 25448      
Microstar v1.0g  SIS471     I.DX/50   64063 32551 11393 25520 128
???? VIP-A486    ALI        I.DX/50   48701 30112  9897 25451            2

Intel DX2, AMD DX2, and Cyrix DX2 chips clocked at 50 MHz
---------------------------------------------------------
VEGA VL486F-3VL  Symph.Hydn A.DX2/50  47022 29134 15120 25217      
????             SIS350     I.DX2/50  48454 30044 11144 25359      
????                        C.DX2/50  47967 21694  7438 24033  64
Compaq Prolinea 4/50        I.DX2/50  30693 ----- 17659 21917   0

Intel DX and AMD DX chips clocked at 40 MHz
-------------------------------------------
A.I.R. 486SH     SIS85C461  I.DX/40   51877 31313 15018 20551            10
Silicon Star AN4 SIS461     I.SX/40   51597 31249 13465 20514 256        10
Soyo SY-025I2    SIS471B    A.DX/40   51597 26040 11053 20384      
???              SIS411     I.DX/40   49976 30032 14479 ?????            10

Intel DX and SX chips clocked at 33 MHz
---------------------------------------
????             DTK Symph. I.DX/33   41866 25668 19580 16783            5
????             Contaq     I.DX/33   42050 25444 13411 16920 ???
A.I.R. 486SH     SIS461     I.DX/33   43192 26094 12429 17037      
Amptron 6800"G"  SIS461     I.SX/33   42424 25897 12425 16959 256
????  SIS486-471 SIS471     I.SX/33   42424 25983 12424 17093 256
Firenze 486VL    UMC491F    I.DX/33   43192 23662 12256 17057 256
????  UC4913     UMC 491F   I.DX/33   32468 20825 10357 16906 256
GA486VT          UMC?       I.DX/33   42805 21697  8282 17029 256
????             OPTi495SLC I.SX/33   42805 21697  7767 16971       
Empec            OPTi495SX  I.DX/33   31297 19323  6465 16572 128        16
Escom            SIS?       I.SX/33   40793 ----- 20710 16577   0        15
HP Vectra 486/33N           I.DX/33   40447 ----- 20079 16532   0

Cyrix 486DLC/40 chips
---------------------
????             Forex      CxDLC/40  37287 25895 17088 16516 ???
????             ALIM1419A  CxDLC/40  38490 26016 15604 16465            3
Amptron          SARC       CxDLC/40  38490 26103 12634 16512            4,7
Amptron?     Bioteq 82C3491 CxDLC/40  38490 26098 12429 16496            9
????             Bioteq???  CxDLC/40  37287 26098 10345 16496            4
M-321            CHIPs      CxDLC/40  37287 25970  8965 16367 128        
M-321            CHIPs      CxDLC/40  39141 ----- 10141 14686 128        

Intel DX and SX chips clocked at 25 MHz
---------------------------------------
PQ4VIP           Cntq 596   I.DX/25   32140 17791 15588 12799 256
Micromax Climax  Cntq 596   I.SX/25   31818 18642 11654 12674 256        17
????             SIS350     I.SX/25   32248 19525 10010 12768      
AMT Cache        ????       I.DX/25   31400 17483  5502 12741  64

1: Cache set to Write-through
2: Cache set to Write-back
3: Cyrix Math Coprocessor
4: ULSI Math Coprocessor
5: Clearly not clean booted when tested
6: AMD 486DX/40 clocked at 50 MHz
7: Full chip set is: SARC RC4018A4
8: Motherboard from Gateway 2000 466V
9: IIT Math Coprocessor
10: clocked at 40 MHz
11: DX2/66 clocked at 80 MHz
12: No Level 2 Cache installed
13: Yes, folks, this board *is* set to minimum wait states
14: These memory throughput figures are bogus.  When wait states are
    increased past a certain point on these boards, they show a huge
    jump in main memory throughput without a corresponding increase in
    real world performance.  Since I don't know why Comptest reports it
    this way, I don't feel comfortable just throwing out the results.
15: SX/25 clocked at 33 MHz
16: This board not set to minimum wait states (known!).
17: User reports system unstable with external cache enabled.
18: AMD DX2/80 clocked at 100 MHz and switched to clock tripling mode.

Brand/Model: If Known, the brand name of the motherboard.
Chip set: The brand, and, if known, the model of the chip set.
CPU: Brand, type, and clock speed of the CPU on the board.
L.1: Throughput of the internal memory cache on the CPU, in kilobytes/second.
L.2: Throughput of the external memory cache on the board.
Main: Throughput of the main system RAM.
Dhry: Performance on the dhrystone benchmark.
Cache: If known, amount of cache installed on the board.

More information can be found in the documentation for Comptest.
-- 
Eric Mintz       Dept. of Biology, UC Santa Cruz, mintz@biology.ucsc.edu
  "If it's green and slimy, it's biology"
  "If it smells, it's chemistry"
  "If it doesn't work, it's physics"
</b></pre>
<hr>
<address>
Send Comments and Suggestions to --- <a 
HREF="mailto:sdw@dfw.net">sdw@dfw.net</a>
</address>

