1. The strange slot is for a proprietary 32 bit memory expansion board. It's not needed but nice to have.

2. Cover up that chip with the window ASAP if you haven't already. It is a UV erasable EPROM that might have already been corrupted or erased by light.

3. Remove the blue barrel battery near the red jumper block and keyboard port before it causes damage, or more than it already has. You can modify a 2032 socket to fit where the battery was.

I wonder if it takes 4MB density SIMMs, or if the cache can be upgraded.

I believe this board is called the Baby Gemini 386.

I'll be interested in seeing benchmarks on this one.

30pin SIMM and 32bit era was a strange one. Like you could have 1MB 4MB or maybe 16MB of ram for this motherboard, but nothing in between 😁

It's not the Baby Gemini, but rather this one:
https://th99.bl4ckb0x.de/m/M-O/30103.htm

The memory configurations are even more limited than you thought, because it does not support 4mb SIMMs. Without the memory expansion card, the only two configurations are 1MB and 4MB, Even stranger, when the memory expansion card is used, the onboard memory slots must be kept empty.

For DX33 it all seems like a strange limitation. Do you think this board is still worth owning? (Strange question, I know).

Still, I guess 4mb will generally suffice. And getting the external RAM card will increase the "cool factor" of the build.

tikoellner wrote:

I was lately looking for some 386 board. Not just any board, but something special.

It uses i386 DX-33 CPU along with i385 and i387 corpocessor.

Looks very nice! Maybe stupid questions...:
What is the i385 for?
Is a 387 meaningful together with a 386DX? I have in mind that the DX has a 387 included (in contrast to 386SX). But maybe I remember wrong?

i385 is cache memory controller.

The difference between 386DX and SX is more than just the FPU unit. Whereas 386DX was a 32-bit CPU, 386SX was made to fill the gap between higher-end 386(DX) and 286 generation. It still uses 16 bit bus. So it's much slower. Neither 386SX or 386DX has FPU built-in. And each of these processors could be coupled with 387.

386SX was 286 on steroids.

In case of 486 - it's a different story. 486SX and 486DX are both 32-bit. The difference is that 486SX had "damaged" FPU unit. So you could couple it with 487, which was actually no less no more than fully autonomous 486DX which worked by substituting (bypassing) 486SX.

The i385 was quickly replaced by cache controllers that were integrated into the motherboard chipsets, but some manufacturers kept using the i385 because it was supposedly more compatible (which is funny, because I think eventually a major bug was found in it).
Neither 386SX nor DX has an integrated FPU. As stated above, the main difference between the two is the size of the data path to main memory. It does have a negative performance impact on the 386SX, but the gap becomes noticeably larger when you don't have any cache (normally SX systems did not have any). In practice that was only a problem on 386s faster than 16-20MHz, since at that speed, good memory doesn't need extra wait states.
Another major selling point of the 386DX was that it had all 32 address lines, which in theory would allow it to access a whopping 4096MB of main memory, vesus the 386SX which only had 24 address lines and was liminited to a mere 16MB. In reality however, 75% of 386DX systems had chipsets that could only handle 16MB anyway, making this a moot point. It was only after 1992 that the DX systems could handle more, usually up to 32MB (but sometimes 64 or 128).

Also, at the time of release, the 386SX was only available as 16 and maybe 20MHz flavours. Intel purposely decided not to release faster versions of the chip to force people to buy the DX if they wanted a 33MHz chip. This was rectified a few years later when AMD started making clones.

If you had a 386SX on a good motherboard with cache, those systems were 80-85% of the speed of a DX. For example, my AMI 386SX-25 with 32kb cache benches about the same as a 386DX-20.

tikoellner wrote:

For DX33 it all seems like a strange limitation. Do you think this board is still worth owning?

You could over-clock the board to 40MHz... but I would not, not with this one.

... and yes, of course it is worth owning. It is quite a nice 386 board of that era... if you do not want it, please... let me be the first to know! 😉

Something nobody has mention yet is the cache system.... this is the same 2-way associative cache used by Compaq... which is why the 385 cache controller is in there... and will map memory very much differently than your typical 386 AT motherboard with cache. I would say the data transfer through this cache system would be quite interesting to see!

Below are both the CPU cards from Compaq's dual 386 CPU system (The Systempro: The Systempro Project; The Build) The cache chips are behind the metal grill enclosures, and the 385 next to this.

What you have in your possession is a single CPU motherboard that is built around this same caching system which is quite special... you have perked my interest a little, as I have not very often seen a typical AT motherboard with this system.

Please do post benchmark results.