I soldered on a second Cx87SLC-33QP, but the results were the same. The FPU does not register in Landmark Speedtest v6. v6 shows N/A for the FPU result, while v2 shows <3.45, which is probably the CPU's FPU emulation speed. The Landmark v2 FPU score should be around 40. Refer to this previous post: Re: Evergreen 486 SuperChip - settings for DIP switch?

I am guessing that I need to play around with the missing SMD resistors, but why would the 1993 version require changing the config resistors, whereas the older 1992 version did not? According to the SMD markings, the resistor values tend to be from 33 ohms, 2K, and 20K.

Should I try shorting J1?

Does anybody have one of these upgrades with a factory FPU installed? Does the FPU work? If yes, could you provided a photo of the underside? Thanks a lot.

Here is an image I took after desoldering the CPU the first time. Maybe the visible traces could be of some help:

I figured out the probable pins for the 87SLC by looking at how an 87DLC QFP is wired on a PGA interposer and looking at the FasMath 83D87 datasheet.

I took a shot in the dark to solder 10K resistors to R22, R31, and R32. I tried all combinations except for the case in which R31 and R32 both have resistors. That was 5 iterations, but it did not result in the FPU working.

I don't know what else to test for. The situation I did not test for was running the Tx486SLC along with the 87SLC. All FPU testing was done with the Cx486SLC2-50MP installed. The SLC2 is stuck on 2x mode, whereas the Tx486SLC runs in 1x mode. I don't see why this would matter. The 1992 Evergreen interposer had a working FPU with the Tx486SLC and the Cx486SLC2 installed.

Is it possible that two 87SLC chips were bad? Yes, but not likely. I do have a 3rd NOS 87SLC, but I give it a 0.1% chance of being the issue.

I'm considering removing the Cx486SLC2 and trying a SXLC2-50 part. I was unable to get the L1 working on the 1992 Evergreen interposer with the SXLC2 installed.

Crazy ideas to get the FPU working are being accepted.

EDIT1: I forgot to point out that even with the Cyrix 87SLC removed and a DIP40 287 FPU installed w/Evergreen 1993 SuperChip, the system will hang when it tries to run floating point operations. If I have a 286 CPU and 287 FPU installed, the FPU works fine. When using the 1992 version of the Evergreen SuperChip, I was able to use external 287 FPUs when 87SLC was not installed. Something fishy with the 1993 Evergreen.

EDIT2: I probed J1, and R31/R32 with the system booting.

J1: 87SLC pin 23 (BUSY#)
this pin stays at 5 V pre-POST, post-POST, and post DOS bootup

J1: 87SLC pin 20 (ERROR#)
this pin is 0V before POST, then 5V after POST, then 0V when loading some DOS drivers, then 5V once all DOS drivers are loaded

R31 and R32 share a connection which goes to D3-10 0.008 (probably a PAL). This pin stays 0V at all times.

EDIT3:
There is a crystal oscillator on the PCB which outputs at 66.6 MHz. I suspect it is used as a means to measure the motherboard's incoming frequency. This frequency is used to determine some kind of some timing parameters on the PAL logic chips. On the old 1992 adaptor, there were DIP switches which the user set to tell the PAL's what the motherboard's frequency is running at and also to enable 2x mode.

Soldered on an SXLC2, but as with the 1992 version, enabling L1 hung the system:

Next, I soldering on a Tx486SLC, which is the same model as the original chip, but a different chip:

The Tx486SLC is in 1x mode only (Cx486SLC2 is always 2x), but the FPU still didn't function. I'm at a loss here. This upgrade adaptor is not of any interest to me if I cannot use the 87SLC FPU. As such, I will not be soldering the Cx486SLC2 back on.

I'm going to set this aside now. Looking forward to some suggestions within the next few years...

feipoa wrote on 2025-03-16, 13:37:

I'm going to set this aside now. Looking forward to some suggestions within the next few years...

🤣 I am crossing my fingers for you. In a months time though you will get a very authoritative "this is what works" answer off google or bing that is an AI rehash of your prior postings 🤣

I see the 287 socket so we should not have to worry about motherboard not having x87 support in BIOS, but do you also have to set FPU present board jumper?

Looks like the kind of board that a harris CPU gets to stretch it's legs on, but is the upgrade counting on it being a slow and nasty board, i.e. does it work if you stick a low speed crystal in it and turn on a bunch of wait states?

I dunno just crossing "i"s and dotting "t"s here I guess.

Edit: Thought I was being funny about AI but it's so fricking overconfident, did an image search on a case and it told me it was SPECIFICALLY an Inwin model, not even close, the buttons were in a different place and the front ports different... it was vaguely the same format... dudebro, it's a Chenbro.... an alternative image search and eyeball scan found it way down the listing with the actual identical photo to the one I was searching... c'mon, you're not even trying now, old image search would have put that at the top.

Wondering what J1 is for, sitting there glaringly with nothing there

BitWrangler wrote on 2025-03-16, 16:17:

lol I am crossing my fingers for you. In a months time though you will get a very authoritative "this is what works" answer off […]
Show full quote

lol I am crossing my fingers for you. In a months time though you will get a very authoritative "this is what works" answer off google or bing that is an AI rehash of your prior postings lol

I see the 287 socket so we should not have to worry about motherboard not having x87 support in BIOS, but do you also have to set FPU present board jumper?

Looks like the kind of board that a harris CPU gets to stretch it's legs on, but is the upgrade counting on it being a slow and nasty board, i.e. does it work if you stick a low speed crystal in it and turn on a bunch of wait states?

I dunno just crossing "i"s and dotting "t"s here I guess.

This is the motherboard I'm using for testing:
https://theretroweb.com/motherboards/s/leadma … ectron-lm-103fs
https://www.petershipkov.com/temp/retro_pc_im … _ht12p-16_a.jpg

There is only one jumper on the motherboard and I don't know what it does. With the jumper in either position, it does not help the FPU situation. I tried the Evergreen 1993 module in another board and the FPU still doesn't work. I tried the 1992 Evergreen module in both computers, and FPU works fine. Thus, the issue is with the 1993 Evergreen module itself.

There are no options to set wait states. In terms of slowing it down, I have tried a 20 MHz crystal oscillator, which is 10 MHz FSB. ALU works, but not FPU. I tried a 12 MHz crystal oscillator, but no lights. I tried a 33 MHz crystal oscillator, but the system will hang up with the Evergreen is installed. This is a bit strange since the module is designed to work with a 16.0 MHz FSB. The fastest crystal oscillator I can use with the Evergreen upgade installed is 30 MHz, for a FSB of 15 MHz. On the other hand, I can use a 50 MHz crystal oscillator with a Harris-25 installed.

The limit on the crystal oscillator speed is due to the combination of the Evergreen upgrade and this particular motherboard. On another motherboard, I can run with 16.67 MHz FSB.

Matth79 wrote on 2025-03-16, 17:00:

Wondering what J1 is for, sitting there glaringly with nothing there

I'm with you. The spacing of the J1 pins doesn't follow the standard header spacing though (2.54 mm, I think). The silkscreen around it suggests it is for a connector. The left-most pin traces to 87SLC pin 23, or BUSY#. The right-most pin traces to 87SLC pin 20, or ERROR#.

Jp1 on the motherboard controls sync/async fpu.

Either the 1993 of the module is damaged in some way, something you mentioned in DM that you cannot test easily because you dont have other 286 mobos readily available for that, or there is some specific incompatibility between exactly this mobo and exactly this cpu module.

I would guess that the jumper position 2-4 would be ASYN, but pin 4 isn't soldered on, and neither is the ASYNC crystal oscillator. This leaves pin position 1-2 and 2-3. I don't know what these two positions do. Having the pin in either position (without ASYNC crystal), doesn't affect the FPU results when a 287 FPU is installed with a Harris 286.

The issue is likely not an incompatibility between just this Headland MB and the Evergreen 1993. I say this because I tested the Evergreen 1993 module in a VLSI MB and the FPU also doesn't work there. When using the Evergreen 1992 module, the FPU works in, both, the Headland MB and the VLSI MB. These are my only 286 motherboards. The VLSI is in a case.

It is possible that some logic on the PALs are bad, if they are in fact PALs. I don't want to remove the stickers to find out. I feel that the most likely issue is related to some mis-configuration of the onboard resistors. There are a lot of 0603 resistors on this module, more than the 1992 version. There's also a bunch more resistors under the PLCC socket. Maybe someone with a working 1993 module will share an image of the bottom of their module.