Still Crazy

021005-3 - Cyrix and Lacuna Part 3

Episode 3: Return of The Jedi... M.A.D?

In order to find out Cyrix 5x86 x4 100 model and ACE486 interposer in stock condition, I opened every PS/55s and PS/2s. Cyrix 5x86 100GP x4 capable model was in 9585 together with a PGA DX4. A stock ACE486 was in 8590/Type-1 25MHz complex with a Cyrix 120GP. Now, it's time to test Cyrix 5x86 100GP/4.

System Specs

9577-BTG 33MHz planar
Hard Drive: EIDE Quantum Fireball ST, 2160AT
CD-ROM: Toshiba XM3401BMA 2.1x speed off from IBM SCSI II (OEM'ed FD MCS-700)
RAM: 8MB x4
Resolution: 800x600, 65K color depth
OS: OS/2 Warp 3 Connect, NO FIX Pack applied yet.
File system: FAT

Results

Test result of SysBench 0.9.4g.
Upper five lines are took from previous page for reference.

CPU CPU
Integer
CPU
Float
DIVE PM Disk I/O CD I/O
Reference
Intel 486DX4-100 44.159 4.809 10.225 21.411 15.999 17.304
Cyrix 5x86 3x *1 45.262 6.088 10.136 14.230 9.056 17.284
Cyrix 5x86 4x ET *3 51.821 7.756 10.140 14.270 9.355 17.291
Cyrix 5x86 4x ET *5 62.986 7.471 10.111 13.154 6.456 17.270
AMD 5x86 4x PL586 *7 65.945 5.492 10.249 21.834 17.263 17.312
5x86 120GP x3 on ACE486 (x3 fixed interposer) Designed for intel DX4-100
Cyrix 5x86 3x ET *8 51.241 5.853 10.140 14.337 9.410 17.290
Cyrix 5x86 3x ET *9 45.221 6.088 10.124 14.228 9.307 17.290
Cyrix 5x86 3x ET *10 50.724 5.930 10.133 14.323 9.344 17.290
Cyrix 5x86 3x ET *11
My favorite 5x86 100GP x4 on EverGreen586 Interposer
Cyrix 5x86 4x ET *10 63.117 7.461 10.121 13.164 6.473 17.270
Cyrix 5x86 4x ET *11 51.924 7.789 10.171 14.270 9.164 17.286
Cyrix 5x86 4x ET *12 50.724 8.887 10.209 19.850 7.686 17.282
Cyrix 5x86 4x ET *13 63.214 8.602 10.222 19.828 7.690 17.282

ET means cache driver et586OS2.sys from et9603.exe is used.

EverGreen 586 Interposer is used for  *1,  *3,  *5.  CB jumper on.
PowerLeap PL586 Interposer is used for *7. WB jumper on.

  *1   Cache driver not used
  *3   Cache driver setting: default (/CCR4=1D)
  *5   Cache driver setting: /CCR2=56  /CCR4=1D  /WBE  (worked fine on 5551-R and Win98)
  *7   AMD 5x86/PGA on  PowerLeap PL586 on an 486 socket interposer (like the one for PODP)

  *8   Cache driver setting:  /CCR4=1D
  *9   Cache driver setting:  /CCR2=16  /CCR4=1D  /WBE
*10   Cache driver setting:  /CCR2=56  /CCR4=1D  /WBE
*11   Cache driver setting:  /CCR2=56  /CCR4=1D  (/CBE makes no difference)
*12   Cache driver setting:  /CCR2=16  /CCR4=38
*13   Cache driver setting:  /CCR2=16  /CCR4=38  /WBE

 *3, *5  are tested with 5x86 133MHz/4x model.

Additional information

  1. Under DOS environment, default values for Lacuna planar with Cyrix 5x86/EverGreen 586 Interposer set to 4x speed are:

    PCR0=0 CCR1=0 CCR2=0 CCR3=10 CCR4=05 SMR1 thru 2=0 PMR=2 DIR0=2C DIR1=5

  2. /CCR3 and /PMR are probably not allowed for the OS/2 driver.

  3. /WBE improves CPU Integer and Floating point scores dramatically but at the same time it decreases video and disk access performance.

  4. Not suitable switches are:

    CCR2=22, CCR4=56,

  5. Don't know how "/CBE" works. /WBE means Write-back cache Enabled.

Impression

Tired and exhausted!

I may say:

  1. Cyrix 5x86 series is not good at DISK I/O on SysBench.exe, I don't know if Cyrix is really so bad at Disk Access on every bench program and on usual usage for daily job.
  2. Video performance is one of weak points of Cyrix CPU.
  3. It's difficult to determine *good* sw parameter for register setting.

By the way 5x86 100GP x4 model marked 85.300 for Dhrystone in CPU Integer test. Don't you think it's astonishing?

As far as OS/2 and Sysbench concerns, it seems that 4x models rated 133MHz and 100MHz are almost same.

And in the end, I don't want to say but if I trust the result then AMD 5x86-133/ADZ in PGA package may be a one to take for OS/2 on Lacuna. No need to hustle to get appropriate cache driver setting.

But how do you consider following scores? Here is Today's Final Result for usual Cyrix 5x86 100GP 3x speed model. I think I probably succeeded.

CPU: Cyrix 5x86 100GP 3x speed model
Interposer: ACE486 designed for intel DX4-100, no jumpers no switches. Fixed to 3x multiplying.
Cache Enabler statement: Device=C:\ETI\ET586OS2.SYS  /CCR2=16  /CCR4=38  /WBE
   The statement is exactly same with *13 above used for 5x86 133GP in 4x multiplication, but the result for 3x model is very different from that of 4x model as shown below.

ET586OS2 .sys should be added at the bottom of config.sys (for OS/2 only. Not for DOS/Win9x)

CPU CPU
Integer
CPU
Float
DIVE PM Disk I/O CD I/O
Intel 486DX4-100 3x 44.159 4.809 10.225 21.411 15.999 17.304
Cyrix 5x86 100GP 3x 51.014 6.794 10.195 21.462 16.756 17.314
2nd execution *1 50.766 6.779 10.205 21.448 15.707 15.707
Cyrix 5x86 w/o driver 45.262 6.088 10.136 14.230 9.056 17.284
IBM 5x86 100 3x *2 50.703 6.482 10.194 21.447 16.679 17.302
   *1 Executed the program twice to confirm above result.
   *2 Added just for confirmation. Same statement was used. Tested on 2002.10.26

Update 2002.10.14

George Albus finally got following result with his Lacuna 25 MHz planar using the same driver arguments as mine (/CCR2=16  /CCR4=38  /WBE). Integer and FP are improved a lot. DIVE is a bit behind the DX4-100. PM and Disk I/O are both almost the same with DX4-100.The et586OS2.sys driver really brought out some hidden abilities of the Cyrix CPU. Don't you think?

CPU CPU
Integer
CPU
Float
DIVE PM Disk I/O CD I/O
Intel 486DX4-100 21221 3348 16174 20464 15119 31076
Cyrix 5x86 No Driver 20517 3967 12825 12367 6530 26904
Cyrix 5x86 et5x86OS2 35400 4687 13597 19534 15269 31083

But let's see his comment about a disadvantage of the driver:

"Only thing is, I lost the Ctrl-Alt-Del function :-("

I have no clue for this symptom. Don't know if another HEX value could solve it or not.

Content created and/or collected by:
Louis F. Ohland, Peter H. Wendt, David L. Beem, William R. Walsh, Tatsuo Sunagawa, Tomáš Slavotínek, Jim Shorney, Tim N. Clarke, Kevin Bowling, and many others.

Ardent Tool of Capitalism is maintained by Tomáš Slavotínek.
Last update: 24 Mar 2024 - Changelog | About | Legal & Contact