• Please review our updated Terms and Rules here

386DX-33 got slower after I added a math coprocessor

Zap!

Experienced Member
Joined
Jun 20, 2021
Messages
115
Location
Staten Island, New York
So I've had this 386DX-33 for a year and a half, and last night got and installed this coprocessor last night. It is correctly installed, and shows up in programs like MSD (and during boot). This is my motherboard.

Anyway, things are much slower. Windows 3.11 takes much longer to load, and when playing .mod files it no longer plays them smoothly. Games that ran are now slow, and games that were slow before (One Must Fall: 2097) now play at a crawl. What gives? Shouldn't my computer be faster? Do I have to change any BIOS settings because of the chip I added? I know that I don't have to change any jumpers. Please help.
 
So I've had this 386DX-33 for a year and a half, and last night got and installed this coprocessor last night. It is correctly installed, and shows up in programs like MSD (and during boot). This is my motherboard.

Anyway, things are much slower. Windows 3.11 takes much longer to load, and when playing .mod files it no longer plays them smoothly. Games that ran are now slow, and games that were slow before (One Must Fall: 2097) now play at a crawl. What gives? Shouldn't my computer be faster? Do I have to change any BIOS settings because of the chip I added? I know that I don't have to change any jumpers. Please help.
Wouldn't hurt to got through your BIOS to see if there is an iem to be toggled. Question: Are you running anything like a spreadsheet or cad/cam app that uses the coprocessor?
 
Shouldn't my computer be faster?
NO! The FPU does not make the computer faster. Only software that uses the FPU will gain speed (or then run at all, as what needs an FPU won't run without it).

Only XTs and very early 286s require telling the system that an FPU has been installed. I would guess the chip is either bad or fake and causes bus errors or something. I would not recommend buying such chips from China. Most are fake. If anything, the speed rating has been raised.
 
Stupid possibility: Does this motherboard have "Turbo Switch" functionality you may have inadvertently toggled either in the BIOS or by knocking a wire loose? As Timo says an FPU shouldn't matter at all to software that doesn't use it. I would suggest running a program like TOPBENCH and seeing if it thinks your computer is running at 8mhz (common de-turbo-ed speed), and if so try to work out why that might be going on.

I'm kind of skeptical that a "fake" FPU could slow down the machine with bus errors, etc, without crashing it entirely, but I suppose weirder things have happened. (A "fake" chip that allows the machine to boot at all would probably consist of a 387DX that's been overclocked; such a chip might crash the machine or return bad answers when the FPU functionality is used, but just slow it down? I dunno.) I'd also suggest installing a program like Check-It and running a general diagnostic; the "System Board" test includes FPU tests.
 
Windows 3.1 has a x87 emulator built-in. For most short floating point operations, it takes longer to send the code to the 387 than it does for the 386 to run it with the emulator.

I would suspect the addition of wait states or cache getting cleared excessively. What motherboard is being used? It's possible that there is documentation on how best to add a 387.

Running validation tests on the 387 won't hurt.
 
Agreeing with Timo...you wouldn't notice any improvement in speed with a CoPro. Only heavy math-involved software (i.e.-CAD, large spreadsheets, etc.) would take advantage of it. As with others, I'd suspect a fake and advise removing it immediately if it's slowing the system down
 
Wouldn't hurt to got through your BIOS to see if there is an iem to be toggled. Question: Are you running anything like a spreadsheet or cad/cam app that uses the coprocessor?

Didn't see anything that mentioned coprocessor, but yes, I wanted to run my old Renegade BBS that I had saved. It requires a coprocessor, and didn't run without it.

NO! The FPU does not make the computer faster. Only software that uses the FPU will gain speed (or then run at all, as what needs an FPU won't run without it).

Thanks, I didn't know that. For someone who isn't a noob, I sure act like one sometimes lol.

Stupid possibility: Does this motherboard have "Turbo Switch" functionality you may have inadvertently toggled either in the BIOS or by knocking a wire loose? As Timo says an FPU shouldn't matter at all to software that doesn't use it. I would suggest running a program like TOPBENCH and seeing if it thinks your computer is running at 8mhz (common de-turbo-ed speed), and if so try to work out why that might be going on.

I'm kind of skeptical that a "fake" FPU could slow down the machine with bus errors, etc, without crashing it entirely, but I suppose weirder things have happened. (A "fake" chip that allows the machine to boot at all would probably consist of a 387DX that's been overclocked; such a chip might crash the machine or return bad answers when the FPU functionality is used, but just slow it down? I dunno.) I'd also suggest installing a program like Check-It and running a general diagnostic; the "System Board" test includes FPU tests.

Thanks so much, you guessed it. While I made sure that I re-attached everything (took pics to make sure), somehow I must have pressed the turbo button, as turbo was off. Using Top Bench, it saw my machine as a 286. Now it correctly sees it as a 386. Man, I sure feel stupid now. :)

Windows 3.1 has a x87 emulator built-in. For most short floating point operations, it takes longer to send the code to the 387 than it does for the 386 to run it with the emulator.

I would suspect the addition of wait states or cache getting cleared excessively. What motherboard is being used? It's possible that there is documentation on how best to add a 387.

Running validation tests on the 387 won't hurt.

That makes sense, since my .mod player in Windows is still skipping, although not nearly as badly as when the turbo button was off. However, before I put in the coprocessor, it never skipped. Here is my mobo:

https://stason.org/TULARC/pc/motherb...386-MB-35.html
 
Thanks so much, you guessed it. While I made sure that I re-attached everything (took pics to make sure), somehow I must have pressed the turbo button, as turbo was off. Using Top Bench, it saw my machine as a 286. Now it correctly sees it as a 386. Man, I sure feel stupid now. :)

Yah! The old aphorism about ""When you hear hoofbeats behind you, don't expect to see a zebra" wins again. ;)
 
That's always good news. And your BBS software runs too? Wonder why it requires a co-processor?
 
That's always good news. And your BBS software runs too? Wonder why it requires a co-processor?

I'm not sure, but Renegade can be quite finicky. It won't run on my K6-2 for some reason, and said "math coprocessor required" before I upgraded the 386. Now it runs great, it's like stepping back in time seeing all of those posts from 1997.
 
I ran a BBS back then, actually before then. From about 1993-1997 or so. Don't have any disks of any of that now. Those were the days. Then the Internet happened and pretty much killed BBSing. I remember when I upgraded from a 386 to a 486 and people were amazed at how much faster it seemed. :)
 
I ran a BBS back then, actually before then. From about 1993-1997 or so. Don't have any disks of any of that now. Those were the days. Then the Internet happened and pretty much killed BBSing. I remember when I upgraded from a 386 to a 486 and people were amazed at how much faster it seemed. :)

I ran mine from 1994-1997 in NY. Great times! I still logon to other boards via Telnet. Maybe one day I'll bring back my own.
 
Back
Top