This system got some attention again.
I wanted to try another video card, because the UMC Video chip had a few quirks:
- No Windows 95 driver, So one has to use the Windows 3.1 driver in Windows 95.
- Scrolling bug in Keen.
- Issues with the image quality, though I was not sure wheter it was the card or the attached cheap LCD screen.

At first an S3 card seemed most desirable, but this article changed my mind to Cirrus Logic: Nerdly Pleasures: Meet a Better Video Card. Since I am a fan of what Cirrus Logic produced for the Sound Card business in the 90's, it would be interesting to give their graphics card a try. Fortunataly the GD542X VLB series are quite abundant, as far as VLB cards go.
This is the one obtained, a CL-GD5428 based card with 2MB. It has the latest BIOS v1.41 already installed.

The new video card was of zero use for weeks, since the existing UMC Multi-IO+VGA controller cannot turn off its VGA part. It can only turn off the IO modules. This left me with no harddisk access when the CL-GD5428 was installed. At most I could verify it posted.
Yesterday this was solved by the arrival of a new Multi-IO card. Quite a nice one: IDEVL2M with ATronics-2015PL chipset. It is brand new in the box, with cables and documentation. It matches the offering from Boca research. They are still around it seems.

Before that I also read about the Cyrix 5x86 being able to do 1x multiplier. This would be very useful, as the 486DX4s are hard to slow down sufficiently for some older games. Death Track for example.
At first I feared such a CPU would not properly run on a mid-generation 486 motherboard, but some people here suggested it should.
Anyways; eventually found a Cyrix 5x86 100GP S1R3 for a reasonable price. The modded BIOS recognizes it properly, and it works very reliable so far. It remains close to the mighty intel DX4 in the graphics benchmarks. But the 1x multiplier is enough reason to replace the previously installed iDX4.
Cx5x86 support is planned for SetMul. Actually the development version does the trick already, but the CPU detection routines must be expanded before I can release it.

Yesterday I benched all the available CPUs and the two VLB graphics Cards. Using all possible multiplier settings and without changing a single thing in the BIOS or the startup configuration.
One thing about the UMC-418 graphics: It is like 7,3% faster then the Cirrus Logic. But considering its downsides, the CL-GD5428 is likely to remain installed. The UMC-418 scored 8,3 FPS in PCPbench when combined with the intel 486DX4-100 and 7,0 with the am486DX4-100.
The last SCORE column is just a quick and dirty attempt at normalizing the results to a single value: Done this way:
Score=((3DB1+3DB2)*0,35+PCPB1*2,9+PCPB2*6,57+Doom*1,08+Speedsys*0,96)/89,5
These scores are generally fine AFAIK. Take into account that Write-Back won't work properly with this motherboard, thus Write-Through mode was selected.
I have the feeling that of all the upgrades done on this system, the two 60ns FPM sticks were the best investment.

There is one disappointment with the CL-GD5428: It does not support graphics mode 100h (640x400x8) without the Univbe TSR. Note that Univbe v5.1 has the smallest memory footprint when adding new Vesa modes: v5.1=8,2 kB; v5.3=12,9 kB; v6.5=19,8 kB.

I had the same mainboard in a "multimedia" computer sold by Neckermann (Germany). It didn't have the same Multi-I/O card though, the I/O card had a UMC chipset and there was a CL-GD5424 with 1 MB. CPU was the UMC 486 with 40 MHz, which wasn't a bad performer for an SX.

That is a sweet case. I would love to find one like that for my 486 project box but can never find one when I'm looking. They seem to be rarer than hens' teeth these days.

oerk wrote:

I had the same mainboard in a "multimedia" computer sold by Neckermann (Germany). It didn't have the same Multi-I/O card though, the I/O card had a UMC chipset and there was a CL-GD5424 with 1 MB. CPU was the UMC 486 with 40 MHz, which wasn't a bad performer for an SX.

You remembered that well. 😀 Obviously they also put a lot of UMC components in that setup.

snorg wrote:

That is a sweet case. I would love to find one like that for my 486 project box but can never find one when I'm looking. They seem to be rarer than hens' teeth these days.

Maybe it is also because computer-cases are kinda large for shipping, and therefore less suitable for ebay and such.
I am not in any hurry to put the components in their case, it would make it harder to swap cards and CPUs.

Here is a shrunk version of the Video Card test in Computer Gaming World no. 125 Dec 1994. The text colors are my doing.

What is funny is the Speedstar Pro CL-GD5428 score, it is exactly the same as my card combined with the am486DX4@66MHz: 3DBench = 47,60 FPS. The exact same benchmark from 1994 was thus recreated in 2014. How useful is that 🤣

Looking at these scores, it shows that a CL-GD5428 is pretty close to what is the maximum framerate in DOS. At least for 320x200 mode. It is in windows that other cards can give a significant boost.

gerwin wrote:

Note that Univbe v5.1 has the smallest memory footprint when adding new Vesa modes: v5.1=8,2 kB; v5.3=12,9 kB; v6.5=19,8 kB.

New favorite: Univbe v5.0=5,5 kB; This one is VESA 1.2 only. Univbe v5.1/5.3 are VESA 2.0 and v6.5 is 3.0.

Gerwin, what software did you use to create your motherboard layout?

An older version of Adobe Illustrator: v9.0, from 2000.
Starting with importing a photo of the board, which is locked in the background layer. Then use layers on top of that, to draw the shapes and add the text.
When it is done I save one version as PDF, with the background photo removed.

gerwin wrote:

Looking at these scores, it shows that a CL-GD5428 is pretty close to what is the maximum framerate in DOS. At least for 320x200 mode. It is in windows that other cards can give a significant boost.

Graphic cards based on the CL-GD5428 are indeed fast, but their BIOS (optimization?), ram speed and manufacture quality also seems to be important too.
I own a CL-GD5424 based card that is slightly faster than a CL-GD5428 based one in DooM... Both using 1024 KB 80 NS.
Something funny is that a S3 805 is faster in pcp bench, 3dbench2 and in DooM (more than 3 fps gain on a dx2 66 😳 ) than a GD5428, but seems to run Hexen slightly slower. 🤣

I got some extra RAM, but made a mistake.
Being 1MB 30-Pin SIMMs, they can only be combined with 4MB 72-Pin SIMMs. They will not combine with the current 8MB modules, like I intented to do.
Half the 30-pin modules are 60ns, the rest are the usual 70ns.

Recently I got a motherboard I was not able to find much information about. Its model number is "MB455". After some research I figured out that this board is almost identical to the TK8498F / F4DXL-UC4 or PT-430. It is also pretty similar to the Biostar MB-1433 UIV. Assisted by some very informative images provided in this thread, I was able to add the originally missing voltage regulator components. With LT1584, R54=270 ohms and R55 = 150 ohms I get 3,5 V output voltage. R56 seems to be only important for an additional regulated voltage e.g. for some Cyrix models. I'm not able to clearly identify this resistor value in the given pictures though. gerwin, could you please post the value here?

Bios-wise I am now using an AMI Winbios designed for the PC Chips M912. A big advantage is that you can use EDO memory.

I just checked it; R56 is 1500 Ohms.

Thank you for your quick help. Using a 1500 ohms resistor with JP23 bridged gives a somewhat low output voltage around 3.15 V. Maybe the used components have disadvantageous tolerances. Replacing it with a 2200 ohm part gives a sufficient 3.35 V output.

Sorry to up the thread but after reading interestingly I think I made a bad mod on the very same mainboard. As I bought it I immediately changed the battery (I discover now it was rechargable) and installed a new three cell 3.6v Ni-Mh 80Mah BM704-33. I have not boot the mainboard yet missing some components, but will I destroy the battery as power going to it will charge it? I don't think this battery is rechargable...

I also bought a "three cell 3.6v NiMh" for this board, A bigger 600mAh pack. The original barrel battery marked 'GP60BNKX3' was a NiCad.
I found this in an ebay add:

A motherboard uses only a very, very low and steady trickle charge current which is just high enough to maintain the battery in a fully charged state, but not high enough to overcharge. Trickle chargers are very simple designs and do not detect anything. NiCd and NiMH batteries are interchangeable for use with trickle chargers such as used in motherboards.

Actually the 486 worked fine for months on 3 AA Alkaline batteries connected to JP1.
http://www.batterymart.com/c-battery-facts-faqs.html:

In most cases, replacement of the CMOS battery is an easy task. It is simply a matter of locating the battery on the computer's motherboard, removing it and plugging in a new one. As a rule, internal batteries should be replaced by the same type of battery which was originally used in the machine or according to the manufacturers specifications. The major exception to this rule are older PCs which were manufactured with a NiCad battery soldered onto the motherboard. These computers usually have a three or four pin male plug, with two of the pins connected via a jumper (this is generally found in the same area of the motherboard as the original battery). This plug gives you the option of leaving the soldered battery in place and replacing it with a plug-in lithium or alkaline battery. Removing the jumper tells the computer to ignore the soldered battery and to look to the pins for its power source. If the motherboard has this provision, you can install a standard PC plug-in battery instead of removing the soldered battery and re-soldering a new one (the standard PC plug-in battery is Electro Battery part number BAT 2005 (lithium) or BAT B40 (alkaline). These batteries are interchangeable).

Well I read this too late... I pointed now the positive side of the ni-mh battery on the first pin of JP1 as seen in this thread leaving the jumper on 2-3. So the positive hole on the original position battery is now empty. Is it right, isn't it?

What I understand from the quoted text: The original barrel battery connector is trickle charged, suitable for NiCd or NiMH, whilst the use of JP1 will not charge a connected battery. So one can connect any battery to JP1; it will work, but the battery will run empty.

IIRC it does not matter if you use the negative terminal from the barrel battery or 'JP1 pin 4', as they are the same line.

(Am unable to verify, this system is in storage.)

Past week I rearranged the storage attic. I took a better look at the yellowed keyboard belonging to this 486, and surprisingly it is actually a quality item: A Cherry G80-1000 HDU /05:

deskthority: "G80-1000 HDU (US layout, black switches, N-key rollover, beige |\ key)"

The black switches are probably of the 'Vintage' type. De keycaps are 'doubleshot'. Must try this myself, as I only used rubber dome keyboards up to this point.

Also had a look at the monitor. It is a Hansol 710A; nothing special I suppose.

This board make me crazy. It has no regulator so 5v only cpu on it. I tried my Overdrive DX4-100 16WT and set up everything as P24C on 33Mhz bus in it and booting up it hang after ram count also reporting the cpu as DX4-S 120Mhz !
33Mhz x 3, how can it feel a 120Mhz cpu that never existed. Tried different cpu but don't seems to be stable at boot sometime it run sometimes it hang. Motheboard is perfect, no traces broken or anthing.

I will soon get one of these boards so I will see what kind of CPUs I can get running.

I have an AMD 5x86-133p75 I could try as its an Evergreen 5V model. 😀

386SX wrote:

Tried different cpu but don't seems to be stable at boot sometime it run sometimes it hang.

Are you saying that you haven't found a single CPU to run reliable so far? In that case I would try a simple intel 486DX-33 first.

My board worked fine with an Am486DX4-100SV8B. Even before updating the BIOS.

Well, it seems that probably is related to my VLB controller ide/floppy card. I am not sure cause I don't have any spare working card to replace and other two I had can't even see anything on the ide cable.
But I begin to suspect it that card.