• Please review our updated Terms and Rules here

Cyrix 386(SX) to 486 upgrade

N

nc_mike

Experienced Member
Joined
Sep 28, 2014
Messages
473
Got my hands on a Cyrix 386SX to 486 snap-on CPU upgrade. I installed the Cyrix CPU- its a snap on over the soldered 386sx and installed with a snap (excuse the pun). I restarted the computer and it booted fine, though I hadn't yet installed the Cyrix 486 cache SW. After boot-up I used SYSINFO7 to check my system config and SYSINFO7 correctly showed that I was running the Cyrix 486 at 33MHz. So far so good.

The moment I try to run the cyrix cache application, I get a loop of what appears to be error codes filling the screen (123F123F123F...) until I hit Ctrl-Break to break out of the loop. I don't see anything in the documentation about this kind of error - has anyone else come across something similar with the Cyrix upgrade?

Regards,
Mike
 
I have no first-hand knowledge of this CPU upgrade processor. There is info that hints that the upgrade may not always work, though.

From the 386-to-486 Upgrade Processor FAQ:

Remember we cannot upgrade 386SX-33MHz or 40MHz PCs

The original Intel 386SX-16MHz processor does not have a float pin, which the Cyrix Upgrade must have in order to work. The Cyrix Upgrade uses this pin to disable the processor and take over the processing. This is only an issue on 386SX-16MHz computers.

Older 387 math coprocessors may cause the system to stall and should be replaced with a new 387.

The Cyrix Upgrade will not work with 386 computers that only use a 287 math coprocessor. The 287 will have to be removed.

The Cyrix Upgrade will not work in Sun workstations.

The Cyrix Upgrade has been tested on many platforms. Within manufacturer models and lines, there can be slight differences that will allow the Upgrade to work well on one machine and not on another, although outwardly they may appear identical.
 
Thanks. The PC is a 386sx Intel desktop and it has no installed math co-processor. Don't know about the float pin though. Funny thing is, the system boots with the 486 upgrade and its recognized as a 486DX @ 33MHz which it what I expected using the Cyrix 25/50. Is just that the cache driver has a problem loading which is needed for optimal performance of the 486.

Mike
 
Stone said:
Ya, and they're, both way slower than a 486 on a 32-bit bus. And probably slower than a 386DX as well. :)
Click to expand...

Sure, but these are upgrades for an existing system, which was an economical way to improve performance back in the day. Saved you from buying a new motherboard, memory and whatever else you had to upgrade when you moved to a real 486DX.

Also, depending on the task, a 486 on a 16-bit bus could very well still be faster than a 386DX on a 32-bit bus.
 
I have this Cyrix upgrade. I think it is worthwhile if you can get the cache going. Although it is not officially supported, I did run mine at 2x33 with some extra active cooling and it seemed to work fine. (No, I would not recommend trying if you can't afford to lose your CPU.) My CPU passed the cyrix tests. Could the problem lie elsewhere? Maybe a bad motherboard or memory?

I think that the 16-bit data bus wasn't as big of a deal as people made it out to be. What really held the 386SX back compared to the DX was that most of the SX boards had 0kb external cache, while the 386DX boards normally came with 64kb. I have two 16-bit boards (one is an AMI 386SX-25 with 32kb, and the other is an Alaris IBM 486SLC2-66 with 128kb), they are both quite fast compared to standard SX boards.
 
Last edited:
Well, the whole point of a 386SX board is that it was cheap. But, if the memory is zero-wait-state, it'd be fine.

The 16-bit data bus would make things a hell of a lot slower than a 32-bit bus processor in many cases, unless of course the processor had cache on-die (as in the IBM 386SLC and the Cyrix and IBM 486SLCs, although the IBMs had a LOT more than the Cyrix 486SLC at 1 KiB).

And then, you had things like the PS/2 56SLC with an IBM 386SLC (8 KiB L1 cache on-die) at 20 MHz or an IBM 486SLC2 (16 KiB L1 cache on-die, more than the full Intel 486 of the time) at 40/20 or 50/25 MHz. Then, you could run RAM in pairs to get 70 ns, 64-bit wide memory, in a 16-bit system, which the on-board SCSI controller and XGA could DMA to/from. Basically, the thing probably had enough potential memory bandwidth (if I've done my math right, about 114.3 MB/s when in interleaved memory mode, I think, compared to the CPU's maximum memory bandwidth of 20 MB/s (hitting every other clock cycle)) to never get saturated if it were configured properly. Benchmarks from when that 20 MHz 386SLC machine came out, between the cache and the memory system's abilities, were showing that it was faster than everyone else's 20 MHz 386DX machines (because their buses could be saturated with less efficient ISA peripheral access, and the on-die cache helped the CPU a lot despite its narrow bus).
 
Last edited:
bhtooefr said:
Well, the whole point of a 386SX board is that it was cheap.
Click to expand...

bhtooefr is correct. A bare bones 386SX would have cost about $700 back in 1989-1990; a case, motherboard, and the chip itself. Quite an investment when you consider that you still needed to negotiate the HD, controller, floppies, video, audio, KB, and memory, and then there's still the question of the monitor. Hopefully you got a PSU with your case. The 486 anything was out of the ball park for me at that time. Very easy to see why Tandy and Commodore hung in there as they did. Gaming performance put things over the top for the home user, not spreadsheets, databases, and word processors, IMHO.
 
Last edited:
Got a little further; booted without loading anything in high memory, ran the cache driver, and no repeating digits, but it just hangs there. Still plugin' away at it.

Mike
 
Uniballer said:
I
The original Intel 386SX-16MHz processor does not have a float pin, which the Cyrix Upgrade must have in order to work. The Cyrix Upgrade uses this pin to disable the processor and take over the processing. This is only an issue on 386SX-16MHz computers.
Click to expand...

nc_mike, this seems pretty definitive... Can't use with a 386sx.

Is it even working? Have you run speed tests before and after to see if there is any speedup without the ability to run the cache program?
 
I'd guess that the cache isn't fast enough in the 486SLC parts for 33 MHz operation. So, the earliest 386SX-16s, and any 386SX-33 or 40 will not work, whereas a later 386SX-16 or a 386SX-25 will work. Potentially underclocking the system to 25 MHz may work as well.
 
It seems that your 386sx board runs at 16 MHz and the FLT pin is working because the 486SLC2 is working at 33 MHz

I have experienced a similar issue with a Texas Instruments 486SXLC2 upgrade. This chip needs software to activate cache and also the clock multiplier (in the cyrix upgrade the clock multiplier is always activated).
If I put the upgrade in a 386sx25 with headland chipset, I can activate the cache but when I try to activate the clock multiplier, the system hangs. It happens only in this motherboard, in other systems I can activate the clock multiplier.

So try the upgrade in another system, these upgrades don't always work in every machine.
 
Ive got the cyrix upgrade chip. On my computer, it stated that it was more compatible with intel maths co-processors when trying to turn on the cache.

Ive got a wanted ad for one of these (for my other computer) so if you have no luck then please contact me:D
 
You must log in or register to reply here.
Back
Top