bakemono wrote on 2025-12-29, 19:35:

Following in the footsteps of llama98, someone has released an AI doohicky for CP/M. https://github.com/HarryR/z80ai

I downloaded GUESS.COM to my Z280 board and ran it there. I asked some questions, but ultimately lost interest before guessing the object. It takes a while for each response.

I wonder if I can get this running on my homebrew Z80-MBC2 build.... it meets the criteria. CP/M 3, 128K, and I have it running at blazing fast 8MHz. 🤣

I'm always looking for an excuse to play with my Z80. 😀

I heard my name. Who summoned me?

myne wrote on 2026-01-04, 12:15:

Special server chipsets have existed more or less since day 1.

Iirc there were dual 486s too. Possibly more. NT multiprocessor x86 had to be developed on something after all.
Probably managed that similarly to the way someone like serverworks got more than 4 ppros running.

Micronics did a dual 486 as far as I found, so likely those were used for creating the NT multiprocessor kernel for x86.

Windows NT 3.0/3.1 must have felt like a time warp compared to what was contemporary at the time. The only thing comparable would have been high-end UNIX workstations and mini computers, I believe.

Imagine using Windows 3.1 and the coming home to your budget Compaq DOS PC.

Kahenraz wrote on 2026-01-04, 18:27:

Windows NT 3.0/3.1 must have felt like a time warp compared to what was contemporary at the time. The only thing comparable would have been high-end UNIX workstations and mini computers, I believe.

Imagine using Windows 3.1 and the coming home to your budget Compaq DOS PC.

There was no Windows NT 3.0
It started at 3.1, followed by 3.5, 3.51, 4.0, 5.o aka 2000, 5.1 aka XP, 5.2 aka 2003 (Server) and XP 64 bit, 6.0 aka Vista, 7, 8, 8.1, 10 and 11.

But, yes, there were some HALs for early multiprocessing systems.

PcBytes wrote on 2026-01-04, 17:24:

Micronics did a dual 486 as far as I found, so likely those were used for creating the NT multiprocessor kernel for x86. […]
Show full quote

myne wrote on 2026-01-04, 12:15:

Special server chipsets have existed more or less since day 1.

Iirc there were dual 486s too. Possibly more. NT multiprocessor x86 had to be developed on something after all.
Probably managed that similarly to the way someone like serverworks got more than 4 ppros running.

Micronics did a dual 486 as far as I found, so likely those were used for creating the NT multiprocessor kernel for x86.

Well, I've looked at that board and I got doubts but was uncertain whether it was a fake.
I've edited that post.
I've edited it again. It IS a FAKE.

It is NOT a dual 486 motherboard.
It is this board:
https://theretroweb.com/motherboards/s/micron … 86v-09-00169-xx
https://theretroweb.com/motherboards/s/micron … bus-09-00144-xx
It has a 486 (PGA-168) and a 487SX (PGA-169) socket.

Joseph_Joestar wrote on 2026-01-04, 11:38:

I just watched this episode of The Computer Chronicles from 1993 where they talk about the Pentium processor in detail.

About 13 minutes in, they show a server which had 64MB RAM, and could host up to 1GB at maximum. Didn't realize that was even a thing back then. When asked who is buying these systems, the salesman answers: people who are working on neural networks and AI. I had to do a double take on that.

The idea goes way-way back and was discredited by the boldness of the claims its founders made. They called a neutron, a perceptron. The original perceptron network was proposed in the late 1950s and consisted of a single layer of perceptrons.

The inventors claimed their perceptron networks could identify objects in photographs. Militaries tested this with photos of tanks and trucks, which the perceptron networks correctly classified.

But the computers of their time could not consider all pixels, and the hype stretched far beyond what the machines could read. It later emerged the perception networks succeeded by assuming that blue sky meant one type of object was in the photo, and a green background meant another object was in the photo.

For the next decade or two AI research focussed on alternative approaches such as Logic Programming or methods of finding ways around maps.

There's a book called Perceptrons, 1969. I found it very helpful in setting new questions on the limits of neural networks, and later convolutional neural networks.

1. tested a few video cards with Necroware's video memory test: it works wonders.
2. soldered a new connector on a wifi card, IPX4 size
3. soldered new ram to a Rage 128 Pro; that was boring 😀

Nice evening here 😀

Did another Beige box conversion, Sprayed a typical XP/vista era black box to make it retro Beige.

https://youtu.be/FDqIr6zbd94

The recent RetroBus ATI Rage 120 Pro revisit got me thinking about a problem I keep running into when rebuilding retro PCs.

Compatibility doesn’t always reproduce the intended experience. Some games technically run but still feel off: input timing, frame pacing, or general “feel” never quite lines up, even on period-correct hardware.

I wrote up some field notes on how hardware, OS/API stacks, and performance constraints interacted to shape how those games were originally played.

If it’s useful to anyone working through similar rebuilds, the first set is here: https://computing-culture.github.io/essays/

Yesterday I replaced the SIO on my 939 A8N5X.
From 7.5 ohms between +5V and Ground, now I'm at 200 ohms but should be on the thousands.

Before it would go instant ON when PSU switched ON; now it doesn't but a mild current is going through anyway, no +12 no +5... :
- chipset fan won't move but if pushed with a finger it would spin (not enough juice to turn but enough to sustain movement)
- if I try a lot of times, it'll turn on fully: PSU would start issuing +12 +5...

Probably it wasn't the SIO but a component in between. I have to replace a LM324.
It took me a to solder the SIO as the leads were probably oxidized and I didn't check before soldering...... (shame on me).

---

Tried to revive 2 2TB HDDs, no go. One spins but not recognized - adapter senses something but not showing "no media" inserted, the other doesn't spin correctly and adapter shows "no media".

---

Needed to reball a DDR2 chip before soldering as the balls it came with, pre applied, are lead free.

---

Trying to solve a DDR1 1GB issue: gives errors in Memtest @ 512,6 -768,6 MB (3rd 256MB block) but not sure of the order on the stick.
From some diagrams it's U1 to U9 (U5 is skipped) from pin 1 to 93; but IDK if on the back U9 is followed by U10 on pin 94 or not.
I hae another faulty 1GB stick, maybe I can swap chips to get one working. I want to prove a point.

Long post 😀