If NO IMAGES will be shown, use www.z80.eu/blog instead of blog.z80.eu

Please note also - THIS BLOG ENDS HERE WITH THE LAST ENTRY FROM NOVEMBER 9th 2021.

I have prepared a new blog with wordpress at https://vintagecomputing.info !!!

Thank you.

It was a kind of a horror trip, because first I was not able to screw/fasten the drives in it because of missing screw holes,then I got problems with the RS232 interface because of a problem with the 486 mainboard, then I replaced the power supply, and after this didn't fix the RS232 problem, I exchanged the whole mainboard with a Intel FX Socket 7 board.
But now I got all problems fixed so far, all components working:

Pentium 166 (no MMX)
ASUS P55TP4 Intel 430FX Mainboard
64MB RAM (more than enough for Windows 98)
ATI Xpert98XL (PCI-card)
Soundblaster AWE/32PnP (ISA-card)
CD-Burner with licensed Nero 6.0
ZIP100 Atapi Drive
8GB UDMA7 CompactFlash HDD
3.5" Floppy Drive
5.25" Floppy Drive

Even Daytona USA Deluxe (with applied Direct3D patch) is working fine (it's really playable with ca. 20-25fps). And the installed DirectX7 is enough also for Diablo I ;-)

I configured it to use optionally a Command Line only after boot, too. So I am also able to use 22Disk, Alien, Uniform and other CP/M Floppy Disk Reader/Writer/Copier Software.

The board itself even supports a Pentium 200, see related link.

I just liked to try again how it works and I tried to replaced the BASIC ROM with my Turbo PASCAL ROM, using the Turbo PASCAL version which does not generate a BAK file (this is important because the RAM Disk has only a size between 9KB and 20KB, and if 20KB, this costs may be too much CP/M TPA space).

So I tried to generate an EPSON PX-8 ROM again, and took the utility program PROMFORM.COM - using it in my MyZ80 emulator. I had to choose 256Kx1 ROM size, and created the ROM file.
But - I forgot it generates an Intel HEX file first, so I had to convert it to a binary also.
This was stopped by the HEXBIN utility, because PROMFORM generated a correct formatted HEX file, but with a heading single empty line (not understood by HEXBIN).
After I removed the first line, it was possible to convert it to a binary 32KByte ROM file, so I was also possible to replace the BASIC ROM file successfully.

I did it first in Toshiya's HC-80 emulator (still very very useful), so I just renamed BASIC.ROM to BASIC.BAK and the TPASNOBK.ROM to BASIC.ROM.
The TPASNOBK.ROM can be downloaded from >my Epson PX-8 page<.

This is the sample session I've done:

First step is set your drive you want to save the source (with "L" for "Logged Drive").
If you don't do this, you can't save the source...

Now enter something useful (this is just a demo source without any further purpose) ...

See how much (memory) space was consumed ... Turbo PASCAL is very efficient ...

It's running ;-)

Starting with installing SuSE Linux 4.2 on a real 486 PC, I thought it would be a good idea to try it in a virtual machine, too.
But guess what happens. Different errors leads in the same result - it didn't work, although console is working always, X-Server did only work with VGA 640x480 in 16 colors (not more).
First I tried to install it in VMWare Player 7, but then I tried it in Workstation 6, too.
Both installation tries resulted in a working character console, but SVGA or even accelerated X-Server aren't possible.
Also, "Soundblaster IDE" interface auto probe at boot time was a pain. This can be skipped by additional boot parameter "sbpcd=0", so started with the installation floppy disk, you had to type in "scsi1 sbpcd=0" always.

But the problem with the graphics adapter continued, also with VirtualBox and Virtual PC 2007. You will get always a scrambled screen after starting X:


You will get additionally trouble with Virtual PC, even if in Virtual PC's BIOS "ACPI" was switched off. Unexpected write errors on your virtual hard drive (in VPC) occur:

Later on, even if you managed to install something, you will be not able to boot or to use programs in general:


A good idea was to try it with QEMU. But unfortunately QEMU does only support a Cirrus Logic GD5446 graphics card, and SuSE Linux 4.2 does NOT. But MS Virtual PC 2007 supports a S3 graphics chip. So even if you can't install it in MS Virtual PC 2007, you can prepare it in QEMU.
Later on, you can convert the virtual hard drive image to a VPC hard drive, just with this command:

qemu-img.exe convert suselinux.raw -O vpc suselinux.vhd

When preparing with QEMU, choose the accelerated X-Server for S3 using xf86config, but don't try to start it in QEMU. Also, because you have to change the floppy disk images, don't forget to switch into the QEMU console to enter the "change floppy0 filename" command with Ctrl-Alt-2 and then back with Ctrl-Alt-1.

The result is a working SuSE Linux in MS Virtual PC 2007, but you can't save any file nor working with it in an usual way, because it will immediately be corrupted again.

You can click on every image to enlarge it.

related link points to version history overview of S.u.S.E Linux (not OpenSuSE nor SuSE Leap)

As already expected, a S3 805 video card seems to be faster than a Trident 9400CXI (which is interestingly displayed as 9100 LCD in some tools). This can be easily checked by using two benchmark tools, one which is well known, the other is difficult to get.
3DBENCH from Superscape in Version 1.0 (not 1.0B or 1.0C, but the first one) is suitable for older PCs below Pentium class, so my 486DX4 shows comparable values.
And Wolfenstein 3D in a "timedemo" version (not the activision version !) shows it even more precise - the S3 (the Miro Crystal 10SD) is superior compared to the Trident.

3DBench TRIDENT 9400CXI:


3DBench S3 805 (Miro Crystal 10SD):


Wolfenstein 3D Timedemo TRIDENT 9400CXI:


Wolfenstein 3D Timedemo S3 805 (Miro Crystal 10SD):


Screenshots were done with VTHIEF 0.07, viewed and again grabbed with ST201F.

Btw. meanwhile I replaced the AMD486DX4-100 CPU with an AMD 5x86 (133MHz), this results in 66.6 fps for 3DBench (with the S3 card).

Related link points to download page of 3DBench 1.0...

I found a Vesa Local Bus Card - a Miro Crystal 10SD, but it was not working.
After a deeper inspection of the card itself, I found something missing on it - a capacitor. So I added a new one, and the card worked again.

But it didn't worked in all 3 VLB slots, just in 2 of them - strange.

Also, I recognized my 486DX4-100 was slower than expected. So after some successless tries with other jumper, I was also asked for memory benchmarks, because I expected "fake cache" problems. The benchmark software SPEEDSYS 4.78 (DOS, originally from http://user.rol.ru/~dxover/speedsys, but disappeared since a few years, can be found now also >here<) helped a lot, because it was possible to see the cache had still an effect, but not much.

Last but not least I was able to blame the TURBO SWITCH for the slow function. After reseating the switch connector, problems were gone (see second benchmark picture).


My very first problem, even before the DirectX setup told me, that the video card had no usable drivers (because of using Windows 3.1 driver), was related with the harddisk, or more in detail, with the CompactFlash card, I tried to use with the IDE controller and the CF Card adapter.

I was able to create a partition and also to format it, but it wasn't bootable. But after using FDISK /MBR, it worked - I thought using FDISK for creating the partition will be enough, but no, the master boot record has to be "fixed" additionally.

The "related link" points to the wikipedia entry for VLB. Interesting facts there.