Conference hydra::amiga_v1

4780.0. "Cheap?, It might be !." by VIVIAN::D_VISTUER () Wed May 29 1991 18:56

    A few notes back I asked if the 'New14Accel' project was feasible.
    Well I built it and it DID'NT WORK !!!.
    
    I think that whoever wrote the wiring list must of been drunk at the
    time. Nevertheless I attempted it.
    
    Unpeturbed, I decided to give the 'CheapAccel' project a bash. To my
    surprise it worked first time !!!.
    
    A characteristic of this hardware hack is that the Machine (A500, 3MB +
    A590, 20Mb) sometimes failed to boot. A cold boot solved that !.
    Also, when it did boot (Which was about 75% of the time) it did
    cleanly.
    Some of the software I tried to run on it did'nt work at all, I presume
    this was due to some sort of bus/speed problem - which is obvious I
    suppose as the hack is rather crude.
    
    Anyway, I ran some performance tests and here are some results. Some
    are strange and unexpected.
    
    1. AibbV2 - all test are setup with, 'be selfish', stack 12000 and 
    		'alt A2000 Base'. Note: A2000=100% . A bootable disk was
     		made which ran 'Aibbv2' from the startup-sequence. This 
    		disk was used to obtain these results.
    
    
     a. 68000 @ 14Mhz.
    
    	Writepixel- 125.1%
    	Dhrystone - 118.0%
    	Sort      - 121.3%
        Sieve     - 118.7%
    	Savage    - 135.4%
    	Matrix    - 125.0%
    
     b. 68010 @ 7Mhz (The processor normally resident in my machine)
    
	Writepixel- 121.3%
    	Dhrystone - 114.2%
    	Sort      - 113.2%
    	Sieve     - 114.6%
    	Savage    - 115.7%
    	Matrix    - 107.9%
    
    
    2. SI reports a 68000 running at 13.22Mhz (?) and gave the rating over
       an A1000 as 1.8. My 68010 gives 1.3 (best) Note: Again as in 1. this
       test was run from a bootable disk.
    
    3. Sysinfo gave a rating of 1.51 over a normal A500 and a clock speed
       of 13.32Mhz. My 68010 is rated at 1.42. Note: this test was run
       from my harddisk folowing a normal boot from it.
    
    4. Mips gives a rating of 1.00. My 68010 gives 0.89. My old 68000 @
       7mhz gave 0.81.
    
    
    I think the 'CheapAccel' hack needs work doing to it in the bus
    compatibility department. I will be looking into it myself and see
    exactly what can be done to improve on this.
    
    As a side bit, does anyone know where I can get hold of a copy of...
    
    VBR
    &
    MoveSSP
    
    Aibbv2's doc file makes a mention of these as ".. special performance
    enchancing software ...". I believe they move certain type of data from
    chip to fast ram to avoid bus contention in chip ram. Also MoveSSP,
    moves the contents of the stack to fast mem to speed it up.
    Any help would be gratefuly recieved.
    
    Finally, could someone run the program 'Cpusped' which you will find on
    tape::amiga:[upload] (I believe) and post the results so I can compare
    results. I am not bothered about the Cpu type but please make sure you 
    run it in 16 bit memory, Thanks.
    
    
    See ya around !!!
    
    Dave Viz.
    4. Mips