VOGONS

387 FPU identification

Topic actions

First post, by RandomBlankUsername

Posted on 
User metadata
Rank Newbie
Rank
Newbie

Starting a new thread, as I didn't want to hijack this thread:
A brief comparison of 386 FPUs

Long story short, I recently bought a CX83S87-33 from .cn from a source I previously had good experiences with, what arrived was some badly lasermarked chip (I mean look as the "F" and "sM" using different fonts than the rest).
After checking out the GND/VCC pins I decided to give it a try in a system I don't care too much about ... turns out it works.
So what did I get? Without checking the die I'm currently unsure whether it's a legit chip.

What I tested so far:
LandMark 6.0 -> 83D87
Sysinfo -> Non-Intel 387
NSSI -> C&T 38700, 33MHz
IEEE-Test -> Fine (taken from: https://theretroweb.com/uploads/drivers/cyrix … 06816664168.img)
IIT 4x4 Test -> Fail (taken from: ftp://ftp.oldskool.org/pub/drivers/IIT/IIT Math Coprocessor/IIT.ZIP)
Intel FPU Diagnostics -> Fail (fails transcendental test ... taken from: https://winworldpc.com/product/386sx-math-coprocess/10)

The main reason I'm going rather indepth is because the final target system would be a PS/2 Model 60 with an "IBM Cached Processor Upgrade Card" (IBM 486SLC2-66) that can enable FPU caching for Cyrix FPUs, where it would run at 33MHz. While I've used that FPU in that system and the FPU cache can be enabled and provides a nice 50% boost in Checkit3, I noticed the machine getting rather unstable (hangup on warmstart, Speedsys CPU test hanging). Quake Timedemo works.
The machine I did the initial tests on has a VLSI TOPCAT board running at 25MHz, where it runs stable.

Currently I'm unsure whether it's a legit Cyrix FPU (vs. C&T 38700), if it's rated at 33MHz (or less - but I'd assume Quake would be an issue then) or if the IBM CPU Upgrade is the source of the instability (runs stable without FPU).

Any leads would be highly appreciated.

Reply 1 of 13, by Anonymous Coward

Posted on 
User metadata
Rank l33t++
Rank
l33t++

It definitely looks weird. It’s probably a 20 or 25mhz c&t that’s been remarked as a 33mhz Cyrix. What’s funny is the c&t chips are far more valuable to collectors in the present day.

See if you can download the manufacturers’ diagnostic from c£t, ulsi and cyrix to see if it can pass the tests.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 4 of 13, by MikeSG

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

There were some "REVEAL by Cyrix" 387sx FPUs on ebay not long ago. I tested one up to 45MHz. They are hardware implemented like the Cyrix and Ulsi's...

Reply 5 of 13, by RandomBlankUsername

Posted on 
User metadata
Rank Newbie
Rank
Newbie

After going down a rabbit hole last night, I guess my system might be affected by this? This would explain why a considerable number of 286->486 CPU upgrades were downgraded to 50MHz and/or 66MHz expansions being rather shortlived.

Reply 6 of 13, by RandomBlankUsername

Posted on 
User metadata
Rank Newbie
Rank
Newbie

Oh ... and in case anyone ever finds the datasheets for the 87DLC and 87SLC (QFP package) FPUs ... please let me know.

Reply 7 of 13, by Anonymous Coward

Posted on 
User metadata
Rank l33t++
Rank
l33t++
RandomBlankUsername wrote on 2026-02-14, 17:55:

After going down a rabbit hole last night, I guess my system might be affected by this? This would explain why a considerable number of 286->486 CPU upgrades were downgraded to 50MHz and/or 66MHz expansions being rather shortlived.

Only the early versions of the 83D87 and 83S87 had this problem. The old 83D87s are easily identified because they had grey tops and smaller lettering. The older 83S87s were ceramic PLCC, switched to plastic in the later models. The vast majority of Cyrix FPUs you encounter will be the later models, as they were produced is far greater quantities.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 8 of 13, by RandomBlankUsername

Posted on 
User metadata
Rank Newbie
Rank
Newbie
Anonymous Coward wrote on 2026-02-15, 01:33:
RandomBlankUsername wrote on 2026-02-14, 17:55:

After going down a rabbit hole last night, I guess my system might be affected by this? This would explain why a considerable number of 286->486 CPU upgrades were downgraded to 50MHz and/or 66MHz expansions being rather shortlived.

Only the early versions of the 83D87 and 83S87 had this problem. The old 83D87s are easily identified because they had grey tops and smaller lettering. The older 83S87s were ceramic PLCC, switched to plastic in the later models. The vast majority of Cyrix FPUs you encounter will be the later models, as they were produced is far greater quantities.

I'd definitely consider mine to be plastic instead of ceramic ... is there a way to derive the production date from the code in the bottom right on the CX83S87, so I don't pull yet another dud? I noticed there are variants with a small space between CX and 83S87 and those which don't have it.

Reply 9 of 13, by Anonymous Coward

Posted on 
User metadata
Rank l33t++
Rank
l33t++

The silkscreen on these chips is supposed to be bright white.
The printing in the bottom right corner should contain a datecode, but I've forgotten how to decipher it. CPU-World is a good place to check.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 10 of 13, by lti

Posted on 
User metadata
Rank Member
Rank
Member
Anonymous Coward wrote on 2026-02-15, 13:07:

The printing in the bottom right corner should contain a datecode, but I've forgotten how to decipher it. CPU-World is a good place to check.

These relabeled chips usually have a somewhat recent fake date code laser-marked on them (along with the fake markings), so seeing a 2015 date code on a Cyrix chip will be another sign that it's fake. It doesn't really help much since the rest of the markings are so obviously wrong, but it could identify some of the better-quality fakes.

Reply 11 of 13, by RandomBlankUsername

Posted on 
User metadata
Rank Newbie
Rank
Newbie

First one I got was a badly relabeled Intel (possibly even a 80186/286 as the pinout didn't comply? - marking on backside, placement of ejector pins on top, pin material showing through [see the white spot on the F]), that went straight to the trash after getting my money back. Problem is on most major sites they'll show originals (which happened to me twice, the later even had videos) and send you fakes/relabels.

Reply 13 of 13, by RandomBlankUsername

Posted on 
User metadata
Rank Newbie
Rank
Newbie

Comptest seems great - already tried that and it detected my .cn Cyrix as a 83S87
PC Config is another tool that can differentiate between old and new Cyrix FPUs.
http://www.dcee.net/Files/Utils/conf802e.zip

Go to top of page Go to top of page
Back to General Old Hardware
Close window
Close window