Image Map Image Map
Page 3 of 5 FirstFirst 12345 LastLast
Results 21 to 30 of 41

Thread: running 5x86 cpu in IBM 5150 with Inboard/386

  1. #21


    Quote Originally Posted by rmay635703 View Post
    I do think a p24t “should” help due to better cache and floating point which are what you need on a bus constrained system

    The trouble will be that the legendary Pentium overdrive was notoriously incompatible.

    But when it did work with cache enabled it did things other 486 chips (including 5x86’s) couldn’t do

    I have one in my basement
    Maybe you can get an additional speed boost by using an ST5x86 with a 4x multiplier.
    "Will the Highways on the internets become more few?"

    V'Ger XT

  2. #22
    Join Date
    Apr 2009


    Quote Originally Posted by Anonymous Coward View Post
    Maybe you can get an additional speed boost by using an ST5x86 with a 4x multiplier.
    @ rmay635703, -sorry late reply- at this ponit , i do not own Pentium Overdrive or 80486 Overdrive chips. The p24t i belive is pga235 and my the socket extension only goes as fare as pga168. Of course i will try, if i stumble upon an overdrive chip
    @ Anonymous Coward , at this time i would like to buy chips that work (out of the box), and i can not find 100% proof that ST5x86 has the x4 setting.
    speaking of buying - this chip certainly would run faster than my chip ! .. but they are rare

    (Buffalo hyper)

  3. #23


    I'd have to ask, but I was pretty certain that all ST5x86s are 3x/4x capable. I'd say about 10-20% of the Cyrix branded models are 3x/4x, and I am pretty confident none of the IBM 5x86Cs are.
    "Will the Highways on the internets become more few?"

    V'Ger XT

  4. #24
    Join Date
    Apr 2009


    actually I was investegating my own question from post "Different types of Intel Inboard 386/pc's" - do they perform different?
    But I was annoyed, not able to manage my "new" cyrix 5x86 x4, and the performance only looked like x3.


    (picture of the "new" 5x86x4. There is no "x4" labeled on the chip, but the model "G5FB607B" in my case IS a "x4" cpu.
    One remark, the CPU do get hot , and i have to cool it with a fan!

    After some hours of testing and i found a way to enable the cache on the 5x86.
    And the good and bad news is - the score from DOOM fullscreen test is now:

    8621 realticks

    equals 74690/8621 = 8,67 Fps !!

    ..the good news, the 5x86x4 were faster then the 5x86x3.
    But the bad news - my prediction that the ISA-bus were the limiting factor
    seems to bee non-void. My newes speedtest surpast the 7.x Fps "limmit" i tought were build in the ISA design.

    I did try the LSSER and the Branch Target Buffer option - but Winstone94 did NOT like it, and all the test app's crashed.
    But simple setup of 5x86 with cache enabled, works well.

    The speed600 looks like this:


    Discover 2.03 from Helix looks like this:


    PS. Puzzle.exe runs in 2.91 secunds
    the topbstub score=56.

    Winstone94 gave this result:


    My quickly assembled and perhaps too long procedure (but it works) :

    boot IBM5150
    with pcdos and Netroom3
    cx5c86.exe /CD (disable cache in 5x86)
    after boot finnished
    run: (the tool that came with the imposter)(enable cache in imposter)
    run: ET586 /SCD (disable cache)
    run: ET586 /WBE (enable write back cache)
    run: ET586 /CCD (equals CD=0 and NW=1)

    ...and the 5x86 then runs in x4 mode.

    cyrix settings looks then like this (ET586.EXE /S) :

    ET586 (c) 1995 by Evergreen Technologies, Inc. (11/28/95)
    Version 1.1 (Release Version)

    586 Configuration Register Dump

    REGISTER INDEX BITS 7 6 5 4 3 2 1 0 = Hex
    -------- ----- --- --- --- --- --- --- --- --- -----
    PCR0 20h 0 0 0 0 0 0 0 0 00h
    CCR1 C1h 0 0 0 0 0 0 0 0 00h
    CCR2 C2h 0 0 0 0 0 0 1 0 02h
    CCR3 C3h 0 0 0 1 0 0 0 0 10h
    CCR4 E8h 0 0 0 0 0 1 0 1 05h
    SMAR0 CDh 0 0 0 0 0 0 0 0 00h
    SMAR1 CEh 0 0 0 0 0 0 0 0 00h
    SMAR2 CFh 0 0 0 0 0 0 0 0 00h
    PMR F0h 0 0 0 0 0 0 1 0 02h
    DIR0 FEh 0 0 1 0 1 1 0 0 2Ch
    DIR1 FFh 0 0 0 0 0 1 0 1 05h

    MSW (CR0) = A0000011
    NW Bit = 1
    CD Bit = 0

    and thanks to forum member feipoa for answering technical questions.


  5. #25


    Very nice! ...from the galley of fellow FrankenXT lovers.


  6. #26


    Great work, love it! Setting now new targets for my XT 5160 - still to be fully assembled

  7. #27


    My results with 50Mhz oscillator on the inboard (=25Mhz bus-speed) and about 8Mhz bus-speed on the 5160 Mainboard. The 4Mb custom expansion was used for these tests. I had to solder the pin B13 on the AMD CPU to VCC to get Write-Back cache instead of Write-Through. It gives a bit more performance (5-10%). Sadly Doom or more advanced benchmarks will not work correctly with the 50Mhz oscillator. I had to switch to a 40/46Mhz oscillator to get stable results without freezing. Doom (from dosbench in fullscreen) runs with that configuration at around 14fps ('timed 2134 gametics in 5274 realtics' for 80Mhz and 5020 realtics at 94Mhz). The Cyrix CPU may be a bit faster at 100Mhz based on the Landmark results.

    50633382436_4b3a852034_o.jpg 50633470087_e60d027b99_o.jpg
    50632632228_2ae83d399c_o.jpg 50633469412_02924b210b_o.jpg

    Last edited by stynx; November 22nd, 2020 at 07:22 AM.

  8. #28
    Join Date
    Sep 2008
    SE MI
    Blog Entries


    Good job! I love it when you can 'hot rod' an old girl like that and it work out.
    Surely not everyone was Kung-fu fighting

  9. #29
    Join Date
    Apr 2009


    late responce due IRL activity

    Normaly i define my hobby to components from pre 2k and no soldering , but this modification of the Inboard386 is so easy i couldn't help trying it out.

    out with the 32 Mhz crystal , in comes the IC-socket-legs and the 40Mhz oscillator.

    And - yes, this new clock frequency makes new speed-results:

    The winstone94 says result 20,0 !

    and Helix-discover2.03 calculates performance to 56Mhz pentium

    And "Puzzle" starts-and-stops in 2,36 sec !

    BUT landmark 6.00 gets lower results then 64Mhz ?
    Dont know exactly why this is lower.
    Perhaps some sort of aligment/non-aligment motherboard(frequency) <> Inboard 16/20Mhz ??

    but sets new topscorer in the original 5150:
    Doom Demo (a) small screen 74690/2093 = 35,6 fps
    Doom Demo (b) full screen 74690/8103 = 9,2 Fps

    @stynx your setup clearly is more pimped then my setup - the motherboard frequency most likely boosts your results, to the impressive score you get.
    @stynx congrats with the 4Mb ram card and the PAL breakthrough - sounds like a fun enhancement of the original inboard, to get around 12 Mb ram in total
    I am intrigued by this RAM expansion project and going to try it out my self. I will return with result/findings a later state (dont hold your breath).
    And great if large RAM expansion becomes available to the masses - great fun with this expansion.


  10. #30


    Man I love this these FrankenXT threads. When I built mine I took a LOT of crit and heat and felt like a parish as a result. Now it's acceptable and I find myself cheering all of you on and mightily impressed!

    My o' my, 8+ MB of extended mem + 586's - just WOW. I don't know how ya'll are managing to manage the power supply and cooling though, did you ditch the case? I could get Win3.1 working well in 5MB and a 486 plug-and play without ditching the case but I couldn't get Win95 or OS/2 to load so I settled for a Win95 UI shell. Might running Win95 become possible with your rigs?

    As an aside I do run all of the old OS's on much faster Pentiums (check out my Multi legacy OS Franken Pentium4 here: ( but I love to see how far the old 5160's can be pushed just for the fun of it. At some point the 8-bit MOB bus has to hit a wall, no?

    Last edited by nc_mike; November 29th, 2020 at 07:59 AM.

Tags for this Thread


Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts