It maybe works, in that it worked 15 years ago, but haven't had it powered recently.

pshipkov wrote on 2024-06-23, 19:14:

The 25/33/40 labeling sounds reasonable. […]
Show full quote

The 25/33/40 labeling sounds reasonable.

I have a new in plastic bag DTK PEM-4036YB with tons of seal-labels everywhere - CPU, L2 cache, RAM, EPROM, KBC.
While the label says 40, the CPU is Intel 386DX-33.
It is possible of course the CPU was downgraded by the customer (Toshiba Corporation), but that does not sound very plausible - you get 40MHz chip from the manufacturer to swap with 33MHz one.

Toshiba used DTK boards in their systems? Weird.
It's also very odd that there would be an Intel 386-33 in the socket. By this time, not too many intel 386s were being sold because AMD was cranking them out super cheap.
My guess is that this board was originally intended for internal use, but was then repurposed for OEM use.

I think both 4036 models (Y and YB) were for industrial use.
DTK made them for some customers. They were used in their time, then scrapped en masse.
That's why these boards are so rare.
Toshiba corporation is written on the plastic bag.

I can’t remember if I mentioned this before, but the ‘Y’ in the model name was used to designate boards using the Symphony chipset.

Anonymous Coward wrote on 2024-06-25, 11:30:

I can’t remember if I mentioned this before, but the ‘Y’ in the model name was used to designate boards using the Symphony chipset.

What was the 'B' for? Budget? For all samples tested, the 'Y' models overclock much better than the 'YB' models.

It's hard to say. I was able to figure out certain things about the naming schemes just by observing a bunch of different boards. This is the first time I've seen another letter following the chipset designation. If I had to guess, the 'B' probably just means "rev B". The only difference is some surface mount chips that replace through-hole parts, right?
What I'd really like to know are what the first three letters in the model name mean. PEM seems common on the 386 boards. PKM for the 486 boards, and PAM for the Pentium boards. I figure one of them probably designates formfactor, one CPU generation, and perhaps the other the application (consumer, workstation, industrial etc).

Anonymous Coward wrote on 2024-06-25, 13:49:

It's hard to say. I was able to figure out certain things about the naming schemes just by observing a bunch of different boards. This is the first time I've seen another letter following the chipset designation. If I had to guess, the 'B' probably just means "rev B". The only difference is some surface mount chips that replace through-hole parts, right?
What I'd really like to know are what the first three letters in the model name mean. PEM seems common on the 386 boards. PKM for the 486 boards, and PAM for the Pentium boards. I figure one of them probably designates formfactor, one CPU generation, and perhaps the other the application (consumer, workstation, industrial etc).

My first socket 7 system was on a DTK "PAM" where the manual said "Pentium ATX Motherboard" but I think I've later seen PAMs not being ATX? 🤔

Think mine was 0057 or 0075. With a VX or TX chipset IIRC (at least an Intel with SD-RAM)

The first character 'P' likely stands for "peripheral", the the last one 'm' likely stands for "motherboard". The 2nd character in the middle seems to designate CPU type, at least in the motherboard category.

PIM - 8088/86
PTM- 80286
PEM -386DX/SX, though later the SX boards seem to have gotten their own PPM designation
PKM -486
PAM - Pentium
PRM - PPro/PII

The memory cards and hard disk controllers use 'I' for the last character. Possibly for "interface" card.

Spent a moment to search online about DTK abbreviations - nothing came up. But that's ok. Not important.

If my presumption that these PEM-003# were for industrial purpose is true, then they were most likely designed and developed towards some conservative specification that favored reliability over performance.
So, i wonder sometimes if the engineers back then were aware of the full range of capabilities of their products ...

---

Another topic.
From time to time i update this thread with information for VLB IDE/SCSI controllers.
Recently inspected DTC 3274 (a SCSI adapter) - it kind of sucked. Like all other VLB SCSI guys.
Made me wonder even more - why so ?
Basically VLB SCSI is not much better than good ISA SCSI.
Curious if somebody here has a theory about that.

AMI Super Voyager PCI-II based on SiS 85C496, S85C497

An interesting looking motherboard, still in excellent condition.
Made me curious what it is capable of.

The additional power connector suggest some special purpose - maybe server model or something like that.
Very nice jumpers layout. Logically ordered left to right and top to bottom, with additional notes printed in the silkscreen next to most of them.
PS/2 mouse support with associated jumper in the lower left corner.
Unfortunately the good part ends here.
AMI BIOS, obviously. Absolutely rudimentary. Only the very basics are exposed in the UI.
Cannot remember touching another motherboard based on SiS 496/497 that struggles with 4x40MHz.
Maybe that was the reason why the manual lists only 25/33 MHz base frequencies, despite the 3 jumpers and a clockgen than can do more.
No EDO RAM support.
Couldn't locate jumpers that force L1 cache in WB mode and 4x multiplier for 486DX4 / DX5 CPUs. Had to use interposer to force these modes.
System gets highly unstable at 4x40 - completely unusable. Was not able to test anything.

Story is slightly better with Intel Pentium Overdrive P24T (POD100) 2.5x40Mhz.
DOS prompt is stable. File operations are stable. But most apps/games tests fail hard.
Quake, Windows, C compilers, offline computation loads all fail hard.
What worked produced ok results.
For example Wolf3D came at 105.8 fps, PC Player Benchmark 22.1 fps, Doom 47MHz.

Didn't bother capturing SpeedSys screenshots.

---

This motherboard is no different than the few other AMI products i was able to touch so far.
They all feel like overoptimized assemblies towards specific use case.
Work well within the established specifications and not a millimeter beyond that.

This is probably the most visibly appealing PCI socket 3 motherboard I have ever seen - worthy of being a 4:3 desktop wallpaper. Too bad it's a lemon.

Was AMI still making motherboards by the time of the PCI socket 3? Or is this an Intel design with AMI stickers?

Did you try using other BIOSes to get around the numerous drawbacks?

It looks gorgeous indeed. That's why decided to give it a try.

Added a note about the default BIOS to my post above.
The issues are clearly on a signal level, but i was indeed thinking to try the Asus PVI microcodes on it. Will know more tomorrow.

The no-mod-required PS/2 mouse support is certainly a plus. Saying this because i know you have a thing about that interface on 486 mobos.

Tried the latest Asus PVI3. Gets stuck at PnP init during POST. Too bad.

When I've run into this issue, I adjusted the IRQ routing table on the donor BIOS to match that of the target system. Also adjusted the Auto Table to agree with the target system's BIOS. Then it worked.

Use Awdbedit v1 to jot down IRQ routing table entries, also the auto table settings. Then open the PVI's BIOS with Awdbedit and adjust the PVI BIOS settings to match the settings you jotted down. Save the PVI BIOS and try it on the American Megatrends motherboard.

That motherboard left the house and has a new owner already, so the chapter is closed. : )

pshipkov wrote on 2024-07-01, 17:57:

That motherboard left the house and has a new owner already, so the chapter is closed. : )

Wow, you had just posted the photo 2 days ago. Someone bought the lemon? Looks over functionality!

There's a good chance the board could have turned out a winner with the correction of the IRQ routing table. I recall offering this solution to another member here and it also worked for him.

Yes. As you know, I am more and more aggressively clearing the junk.
The issues of this board are on signal level. That's why my BIOS effort was kind of weak. Somebody already got the mobo, so i had few moments only to try the PVI BIOS.

The 486-voltage-interposer for Intel- and AMD-486-CPUs is ready for use - see pictures.

Not all GND-pins are connected to the GND-Vout of the buck-generator (only few pins at one site), this seems to be enough.

About the buck-generator from this link shown above (https://www.conrad.de/de/p/az-delivery-lm2596 … ductDescription): POOR quality, I must say:

- 2 of 5 generators ordered with NO functional trimers (maybe they are defective, maybe bad soldered)
- the Vout-display of the working ones shows up to +/- 0,1 V than measured by myself (so better you don't trust the display)
- the solder-quality itself seems to be not good

Anyway, using one working buck-genereator, on the HOT-433 Rev. 4 shown above, it was possible to run STABLE an Intel 486DX4-100WB @ 133 MHz (FSB 66) and 4,25 V in Windows! Not bad 😀 The CPU was able to boot in DOS @ 150 MHz/4,8 V, but was completely unstable for Windows (BSOD at booting).

The rest of the hardware:
- 256 KB L2 @ 3-2-3
- one module of 32 MB EDO-RAM (50 ns)
- GeForce 2 MX-400

The results for 133 MHz:
- GLQuake (800x600): 24,5 FPS
- Quake II: 14,0 FPS

Until know I was not able to run some of my Am486DX5-133, which already runs @ 4 or 5 V at 180 MHz, at 200 MHz (not all testet yet). On the other hand, I found another such CPU, which can do 180 MHz at only 3,70 V (but nothing faster like this).

So in conclusion it's very useful to have such an interposer to find out the highest stable frequency of a 486-CPU.

next pics

and the last pics