- 486 computer with Vesa Local Bus -
 |
I started to build this around 2018 with random parts that I've acquired from eBay and local offers over the years.
This machine contains the following parts:
- 486 computer with Vesa Local Bus -
|
I started to build this around 2018 with random parts that I've acquired from eBay and local offers over the years.
This machine contains the following parts:
->Asus VL/I-486SV2GX2, ->Amstrad 40046 / DFD222A01, ->STB Lightspeed ET4000W32 VLB, ->QD6580 CARD V3.0, ->The two soundcards, ->The cables sticking out from the back
 |
 |
 |
Windows 95b is currently installed on this and will be used for DOS gaming as well as testing my Win32 applications for performance on real 90s machine. I originally wanted to multiboot other Windows versions like 98FE and perhaps 3.11 but I decided that this effort wasn't worth it.
This is the hardware used to test this very website with IE3 as that was included with Win95b, browsing the web on a system like this is a real test of patience especially with the newer backported browsers that load even slower than IE3.
EZ-Drive is used to overcome 8,4GB limitation as that works so much better than XTIDE which caused weird floppy drive issues like being unable to format any disks.
In middle of 2022, this motherboard stopped POSTing and I've tried all known good 486 CPUs and RAM I had in my stash, I even put the original BIOS back. So, it must be a hardware failure as nothing made it work again.
I purchased a cheap POST card to see what the CPU was doing but to my disappointment it wasn't doing anything at all.
Now, buying an exact replacement is out of the equation as the price for vintage tech is anything but low unless it's bottom-of-line stuff that nobody wanted back then. That's when I started a long frustrating journey of repairing this board.
I started to desolder certain things, the DIP40 keyboard controller was socketed and I've tested a spare one but still nothing. I replaced the SOP20 74F245 from another broken scrap board as these got quite hot but I realized later that this was normal, I even replaced the BIOS socket with a hi-quality one.
So the CPU still didn't run properly and I looked closer at the CPU socket and noticed burnt traces.
I recall exactly how that happened.
The first time when I replaced the old 486DX2 with an 486DX4 that I scavenged from an old laptop with snapped off hinges and battery leakage.
I accidentally turned my hand while inserting new CPU with the cooler+fan attached, powered it on and waited for the familiar BEEP but powered off after 5 seconds of hearing nothing but fans. I took the DX4 out and noticed right away an burnt metallic smell.
Ignoring the fact that I pumped several amps through that poor CPU, I've reinserted that CPU back the correct way and waited for 10 minutes.
I pressed the power button and would you believe it. The CPU worked happily like nothing bad ever happened. Damn, that old tech can take some serious abuse.
I read stories about today's computers how some folks damaged their brand-new CPUs and motherboards beyond repair by having a little bit of dirt blocking some pins from making contact but this oldass shit takes several amps in the wrong way, heats up like a stove and still survives. I'm quite curious how the Am5x86 would fare in that situation.
Anyways, I DESOLDERED THE ENTIRE SOCKET 3 WITH ALL IT'S 237 PINS BY HAND WITH A FUCKING SOLDERAPULT and used a hot-air rework station for support.
The socket came out slightly deformed and banana-shaped, but it seemed OK so far.
I fixed the damaged trace which was thin and connected to the power rail, now that makes sense.
The desoldered socket went back where it was and I resoldered all 237 pins, the socket looks a bit warped but actually makes good contact with the CPU.
And guess what?
It's still broken, so few days were spent on measuring all signals with an scope but I couldn't pinpoint the exact failure. Later I got the idea putting a little load on all signals that I've measured, and that was it.
ISA D12 had no contact to the 85c471 pin 50, the trace is inside the middle layers of the motherboard.
This is the most concerning failure I ever fixed. Like, this hidden trace failed overnight on it's own for no apparent reason
. However, it's finally fixed and everything works now.
That was a wild month, I think I spent like 30-50 hours in total to make this miracle happen.
I also created a pinout for the 85c407 and some notes for the 7400 logic used on the board.
At first this wasn't working properly. The spindle motor was sometimes spinning way too slow at around 150RPM, I suspected the servo circuits for being bad and shotgun replaced all parts on the circuitboard but in the end I was able to fix that problem by flooding contact spray into the holes of the motor.
That attempt fixed the drive for good, I never had any issues since then.
There's also a weird BIOS bug with that drive, if "boot-up floopy seek" is enabled then DOS can't access the drive at all and Win95 has trouble formatting any disks. For some reason that oddness doesn't occur if that option is disabled.
- STB Lightspeed ET4000W32 VLB -
This card has a nice 135Mhz RAMDAC but the video output was way too hot, that's why I added these three resistors to get it down to the correct 0,7Vpp.
CH9294E generates the clocks but only 15 different pixelclocks as 65Mhz is assigned twice in the table.
I did some little reverse engineering on the BIOS with the intention to customize the video modes to my preferred timings but I'm far away from being done. I don't know where all pixel clock settings are stored.
The speed is very good at 6-7MB per second but only if the correct driver is loaded. The rest of the card is rather lame with an LPT port lacking bi-directional support and the UARTs don't have any FIFO buffers.
 |
 |
This version of the SB16 is known to have the hanging note bug.
A simple and effective workaround is to use a 2nd soundcard as the MIDI output, that's why the CX4235 is hooked up to the Line-in on the SB16. The bus conflict of the Adlib port at 0x388 can be avoided by setting the PnP drivers of the CX4235 to another unused address, in my case I assigned 0x588 for it.
I also set the SB16 to base address 0x240 and the IRQ to 10 as most software that is new enough to support SB16 features should be able to use these settings, the other card that also supports SB-Pro is set to the more usual base address 0x220 IRQ 5 for older hardcoded software or if something doesn't work all right with the SB16.
A Dreamblaster S2 is on the "to buy" list.
- The cables sticking out from the back -
Doing a full backup through the actual computer would take a long, long time. That's why I came up with the idea of routing the HDD cables to the outside where I'll could easily disconnect the drive from the system and hookup a external USB adapter and Molex powersupply for reading and writing without needing to take the entire thing apart and this works great as the cable going to the drive is actually a modern 80 conductor cable.
I also had that passive PS/2 to serial adapter laying around and plenty of spare D-Sub connectors, so I used that stuff to connect to the PS/2 mouse pinheader on the motherboard.
And I've added rubber protection around these cables as these early PC cases are well known for their sharp edges.
Return to main index
--EOF--
