Case Disassembly

This is..not bad! I mean, it's going to be an epic cut once I finally forget to wear gloves around this thing, and slice my hands open, but..

This case is sold as "new" and came in what appears to be the original box, but it also looks like it had a barebones system in it at one time. I don't know the history (of this horde/why the seller ended up with them, or of Techmedia itself), but for this purpose it seems like a fine chassis.

Wouldn't be a retro build without some sort of refurbishment..

What got me on this case was the size and beige-ness for sure, but its minimal design, down to the symbols used for power/HDD/keylock/reset/etc..I really like this look. Maybe not "peak 90s" for sure, but different vs Gateway, Dell, AST systems at the time. The large, rectangular LEDs are nice too. I like it.

No retrobright needed, I'll take that. Just need to get rid of whatever this is on the front panel.

Power supply:

Looks to be in good shape. I don't miss these things. Not sure if I'm going to use this supply, or a new Athena Power.

Also came with a floppy drive:

That's about all I've got for now. Time to do some more reading while I wait for the motherboard to arrive.

Luckybob did an unboxing video of his awhile ago. Nice looking case.

Oh wow, I missed that thread! Thanks! (And thanks Luckybob for doing that video - awesome dude!)

The case is super interesting and I really hope it works out. I tore down the front panel to just plastic and gave it a quick bath - hey, good as new now. The rest of the case looks, well, like it got kicked around for 25 years, but more or less it's in really fine shape.

The 6 vs 7 expansion slot thing will be interesting in my application, but it looks like I will just lose the top-most PCI slot. Not a huge deal. Funny that the barebones system comes with a similar layout - 4 PCI slots, 3 ISA. (Luckybob points this out hilariously in the video)

Plastic AT-style standoffs - In the USA I was able to get these from a couple of stores, it looks like:
https://www.bestbyte.net/plastic-motherboard- … f-at-style.html
https://www.pccables.com/STANDOFFS-MOTHERBOAR … C-100-PACK.html (went with this, and got expansion card blanks from them as well)

Storage - I have no intention of doing a "period correct" thing. Apologies and major respect to those that do, that's just not me.
feipoa suggests that the 2940U2W cards play well with these 486 boards; I'm of two minds, one is go All The Way with SCSI, the other is KISS and just stick to CompactFlash on IDE. I think the use case will be DOS or perhaps Win 3.11, so not exactly i/o heavy.

Good choice on the diamond stealth 2000, that's easily the highest quality virge card I've messed with

Very nice, I put my M919 with Am5x86-133 in that same case 😁 Went with a bit of a worse sound card though (Soundscape) and still trying to figure out what PCI video card I want to stick with.

I have one of these Techmedia cases but the front face is different. But the internals are the same.
I was on eBay looking for an AT tower case for my 486 build and I found this one.
It’s taller than I wanted but it works great. Solid steel construction. Not cheap thin tin sheet metal.
I like my front cover better. It looks really nice with the front bay door cover closed. Looks very clean.
The case is kind of difficult to work in but its worth it.
Case is very heavy due to thick steel construction.
i read techmedia was a computer builder back in 1990’s.
They designed there own cases and motherboards.
i think this case was built around Windows 95 era.
So it might have been intended for a 1st. Gen Pentium class CPU/Motherboard.

See my post:
486 Multimedia dream build ( 1993/94 )

Hey, that is pretty cool. I love the marketing and that just screams late 90s/turn of millennium in my mind.

As for this particular machine - I think it gets a 3C905 card, and the Stealth 3D in the PCI slots. So that’s that, probably no SCSI craziness. Sorry to disappoint 😁

Processor has arrived and if I'm reading correctly that is a 1996, week 48 date code. We have a theme here with end-of-1996 it seems..

Is it me, or the PCB of the mobo was clearly and intentionally made for fake cache? 🤣

PcBytes wrote on 2022-09-24, 18:37:

Is it me, or the PCB of the mobo was clearly and intentionally made for fake cache? 🤣

By the later 3.x revisions you have to figure the “secret” was out, right? Maybe the PCB space was just left there for giggles at that point, who knows. I love this though and realized (when I pulled up the manual on archive.org) that my family totally had one of these boards when I was a kid. It was a local/small business PC builder product. These things truly were everywhere.

It *may* have been this board that I misflashed with either the wrong BIOS or a badly downloaded copy, and bricked it. Definitely had no idea how to fix that situation back then.

Anyway..story time aside, I’m working on a 5.25” bay arrangement for one of the Syba CF-IDE adapters; absolutely nothing fancy but will look the part I think. Extending the power/activity LEDs out to the front panel will look nice as well in my opinion.

For power, I like the idea of a new Athena Power unit, but I immediately voided the warranty and cracked it open - there’s nothing in there, and it weighs far less than the techmedia unit! It’s not like I need more than like 35 watts for the system but I do want to protect this hardware as best I can. Ideal preference would be AT2ATX smart converter, but I don’t think I can get one. Next best might be a well designed adapter board in between a modern PSU and the M919. The hunt continues.

Finally some progress - the board came in, and I immediately knew I had a real project on my hands when the smoke smell hit me.

NO PROBLEM - we got this. I had to remove/reinsert the video card a few times to get it to POST, but I got it there on the bench. Awesome. Swapped the FPM for a full 64MB cadre of EDO - boots right back up. 1MB L2 detected too. Absolute legend.

Time for the gold star treatment as we want those decades of tar off the board. Fresh out of the ultrasonic cleaner, that fine PCB shine is back. New set of jumper caps, a couple of massively overkill heatsinks on the chipset, and an extra bit of scrubbing on the CR2032 holder, clock gen and other metal components. I didn’t quite get the battery holder as clean as I wanted, but it seems to be working. An easy swap later of course.

Next I think is the CPU cooler - I started looking at 3D printed parts and I think there may be opportunity in getting a Socket 7 arrangement adapted for it, or just simply a bracket that grabs the existing heatsink from underneath the CPU and pulls it down.

Ordered a handful of SST39SF010A 1mbit chips - will make an attempt to get the latest BIOS burned onto it using flashrom, and my 3c509 doing double duty as an EEPROM programmer.

Separate from this, I am curious:

From http://th2chips.freeservers.com/m919/index.html ..

I had to rplace the two transistors for the onboard power plane (the ones with the heatsinks). Mine were NEC D1691. I used compatible D826. Works perfect so far.

My board seems to be stable, and very basic testing with the multimeter shows stable power coming out of the transistors..however I have read anecdotally that these components are "weak." For a novice here, what would indicate that these transistors need to be replaced? Should I also consider a recap while I potentially have the board under the iron?

Cache To Nowhere

In today's episode of I Can't Believe It Still Boots..

You can see some bits of rust starting to form on the inside of the chassis in the previous post. One gentle-ish pass with the grinder and a couple of coats of flat white primer later, and we're back in business.

I plan on doing a 3D printed 5.25" bay mounting situation for the CF-to-IDE adapter. I wanted to keep the LEDs somewhat useful, though, so I desoldered the power and activity LEDs, swapping them instead for 2x5mm units that will mount at the front of the bay. I've just got this on the bench for now while the print completes, but that was simple enough. I'll also have to pick up some beige paint and give it a quick spray before it goes in the case - can't have something not-beige in there of course.

My retro builds usually feature a Corsair RMx-series power supply, as they're pretty easy to buy, fully modular, and have fairly stout 5V rails. I'm going to shoehorn one of these supplies into our unsuspecting Techmedia relic too. A very professional looking ATX->AT adapter board just showed up in my mailbox and we'll use that for the go-between. -5V then remains in play; not that I think I have any use for it just as yet.

While I had the iron out, I went ahead and swapped that corroded CR2032 holder on the M919; seems I keep trying to blow this thing up..

Assembly in progress..

Next? Well..

• I've got a couple of potential CPU cooling solutions in the works - really appreciate all that share their 3D designs. With the lack of airflow in this case, I'd really like to get a much larger Socket 7 cooler on this 486.
  https://www.thingiverse.com/thing:2459258 - promising..
  https://www.thingiverse.com/thing:4999644 - yeah, I'd love it if I could get this to work!
• A WP32 McCake is on the way to fill another one of those 5.25 bays. Really looking forward to trying that out.

The real battle begins..

Now it's time to tame this beast, as (relatively speaking) it's a lot easier to make the hardware pretty and presentable - getting a stable rig out of hardware 20-30+ years old becomes the challenge, as this forum knows.

While I am absolutely blown away by the feats of the likes of feipoa, pshipkov and others, in the pursuit of the fastest 486..it looks like I will merely be cooling this particular 5x86 ADZ with only a slightly modified Socket 3 cooler. I bow to the incredible talents on display in this forum and simply get to stand by in awe, sometimes being able to take away nuggets of knowledge or even brand new parts or methods to use.

One such knowledge nugget is the finnicky behavior that is evidently to be expected from this UMC chipset, and its inbuilt IDE controller. I've arrived, I'm there, I confirm all of it. My setup goes like this:

-- Primary Master: Syba IDE to CF Adapter
- ATP Industrial 2GB CompactFlash card

-- Secondary Master: Sony 52x CDROM.

For convenience, I removed the "KEY" pins on the motherboard IDE connectors, which are almost always blocked off on later IDE cables.

The motherboard's IDE controller is configured as follows:
- Used the "Auto IDE" function in the BIOS to configure the CF card geometry; it is correctly detected when compared to a known-good SS7 system.
- LBA is detected and enabled for the Primary Master channel as part of this process.

The problems start when we try to ratchet up the performance:
- PIO Mode 3 or 4 seems to guarantee failure on any kind of write operation.
- PIO Mode 2 vastly reduces, but doesn't eliminate, the write failures; they still however happened often enough to cause Windows 95 installation to go completely sideways.
- PIO Mode 0 appears to be safe but obviously slow.

Other experimentation and observations:
- The 32-bit transfers option immediately breaks MS-DOS compatibility, however it works fine with FreeDOS. (I have not yet tested with Windows 95)
- block transfers may serve no benefit here for a CF card, so I have left that setting at the auto-detected single sector per interrupt.
- Enabling DMA in Windows 95 caused "Windows protection error" on boot.

I'm still at this point not considering a move to SCSI, and I would really prefer to use CF or SD cards for the convenience of being able to move files easily using a modern machine.

What is on my mind is that last PCI slot. I could put a more capable Promise or similar IDE controller in there, and stop fooling with the UMC controller..

Separately and somewhat related - I'm going to attempt to burn a new BIOS EEPROM, using flashrom and the DIP-32 socket on the 3com card. More to come on that.

So, here is your m919 build log. Good stuff.
One thing to note - the stability issues listed in your last post sound unusual.
At up to 160mhz you should be able to max-out all bios settings without any side effects.
I can point you to two posts - one from me and another by @chadti99 talking about that - for reference, if you want.

Thanks so much for the input! I am always, always interested to learn more. So certainly, would love to see the posts if you have them at the ready.

I left memtest86+ running on the machine for the last 24 hours or so. No issues to report there at least. It is simply storage I am having trouble with.

Since 86Box can emulate the M919 board, I've taken to doing a bit of software configuration R&D via a modern machine, while I want for some parts to arrive. I can nearly exactly mirror the hardware configuration of my real rig to the virtual machine. Some ..entertaining speedsys results, but the emulated processor speed is bang on, and I suppose that's mostly what we're about here anyway. What's ultimately nice here is being able to prove out the software in a basic way before potentially fighting with it on the real thing.

The two links:
Re: 3 (+3 more) retro battle stations
Re: 3 (+3 more) retro battle stations (there are 2-3 more pages after the post with chatter about M919 and its peculiarities)

Thanks very much pshipkov, more reading material for the evening and more experimentation once the machine is back on the bench. soon. (tm)

Socket 3 cooling, for the modern computing enthusiast

There's some fascinating stuff available from either Mouser or Digi-Key, if you can successfully navigate the maze that their sites can be. I picked up both industrial grade CF and SD cards from Mouser, which I hope will last a bit longer than the plain 'ol consumer grade stuff.

- 8GB SanDisk SD Industrial XI
- ATP 2GB Industrial CompactFlash

You can see right away that the SD cards are significantly less expensive than the CF.

Anyway - back to cooling. Given that the AMD 5x86 is a 45mm square chip, it stands to reason that a more modern 45mm square cooling solution can/should/will work just fine.

Here's the Mouser search pattern to help show my thinking.

The "FanSINK" (part # ATS-61450R-C1-R0) sounds right, but given a bit of study on the datasheets, I wasn't certain that the "MaxiGRIP" retention mechanism would fit. Advanced Thermal Solutions also make a smaller clip kit called "SuperGRIP" that is meant to fit in more crowded PCB spaces. Part # ATS-SG425-R0 if it helps someone later. I did not end up needing this.

Here's a comparison of the mounting solutions:

In this application, both the square clips grab onto the sides of the processor, and then the retention mechanism goes over the heatsink and clips down, increasing the tension on the square clip. Simple, effective.

This assembly didn't come with a fan, but on the datasheet one of the recommendations was Sunon ME45101V1-000U-A99 (45mm x 10mm, 12vdc), so I went for that.
All said, about $30 USD for what should be a great cooling setup.

Here's once again a comparison shot of what I had come up with, a very old EverCool socket 3 cooler, with the super noisy fan thrown out and a 40mm Noctua bolted on instead. The retention mechanism was 3D printed in PETG.

And the final product. The heatsink comes with screws of various lengths to accommodate a fan. The Sunon fan however just comes with bare leads - no problem, I crimped Dupont pins on and connected it to the M919's fan power header.

So the tl;dr summary, here's your BOM:

- Advanced Thermal Solutions FanSINK assembly: ATS-61450R-C1-R0
- Sunon fan: ME45101V1-000U-A99