There is something about maxing out a certain motherboard platform, in this case PGA-132 (386) motherboards, that many people, including myself, find fascinating to attempt. I try to take each setup I have to the very limit of their capabilities. For the PGA-132 386 platform, I'm trying for a system with SXL-40 which does not have limited RAM read/write times, like most of the later 386 boards do. Alternately, I'd also like to get my SXL2 running at 2x33, as a possible alternative to the SXL-40. I

have two variations of PGA-168 (486) to PGA-132 (386) adapter boards that I'd like to test out, however, it is unlikely that I would use these in a final setup; I have 486 boards for that.

feipoa wrote:

There is something about maxing out a certain motherboard platform, in this case PGA-132 (386) motherboards, that many people, including myself, find fascinating to attempt. I try to take each setup I have to the very limit of their capabilities. For the PGA-132 386 platform, I'm trying for a system with SXL-40 which does not have limited RAM read/write times, like most of the later 386 boards do. Alternately, I'd also like to get my SXL2 running at 2x33, as a possible alternative to the SXL-40. I

have two variations of PGA-168 (486) to PGA-132 (386) adapter boards that I'd like to test out, however, it is unlikely that I would use these in a final setup; I have 486 boards for that.

Yeah that's it for me too. I am going to build the most powerful system for each 386 and 486 I can have. But on the 486DLC or SXL I am confused cause they are not exactly 386s... so... I would prefer to use the original 386 40mhz for now. But if I find the SXL2 66Mhz I'm going to test it for sure.

386SX wrote:

But if I find the SXL2 66Mhz I'm going to test it for sure.

There is no SXL2 66MHz part for PGA132. The fastest part is 50MHz.
But there are IBM BlueLightning parts for the 386 platform, up to 100MHz. These chips also have 16KB of internal cache.

But as mentioned here before: There will not be 100% stability due to different issues. But in practice, for DOS gaming, I have very good first hand experience with all these chips.

What about this?
http://www.cpu-cn.com/en/showpic.asp?id=1169&no=1

kixs wrote:

What about this?
http://www.cpu-cn.com/en/showpic.asp?id=1169&no=1

This is a 144pin QFP 3.6 Volt part. The 132pin socket for 386DX cpu needs 5V. So you need either a special motherboard and / or an interposer socket.
The fastest standard drop in PGA132 part of the SXL range is the SXL2 50MHz.

FGB wrote:

kixs wrote:

What about this?
http://www.cpu-cn.com/en/showpic.asp?id=1169&no=1

This is a 144pin QFP 3.6 Volt part. The 132pin socket for 386DX cpu needs 5V. So you need either a special motherboard and / or an interposer socket.
The fastest standard drop in PGA132 part of the SXL range is the SXL2 50MHz.

I believe there is an upgrade unit with that SXL2-66 residing on it such that it fits right into the PGA-132 without the voltage trouble.

Why no Pentium to 486 upgrade adapter board?

Anonymous Coward wrote:

Having the FPU on chip absolutely was not an issue for getting the 486 onto the 386 bus. The Rapidcad was essentially the i486 core with the 8kb cache removed.

The external IRQ13 error handling has to be dealt with: http://web.archive.org/web/20130303195857/htt … handling_on_x87

Scali wrote:

That is the issue I was talking about.

Also, according to this page: http://www.cpushack.com/RAPIDCAD.html

Haha, that picture of the Rapid-Cad set is mine, took it ages ago with a Creative webcam II 🤣

It seems that there is one here which has something to plug into the FPU socket: http://www.vintage-computer.com/vcforum/showt … board-386/page3

That extra plug is actually supposed to go under the i385 cache controller chip (for L1 cache coherency).

I'm not sure what the deal is with the RapidCAD "set". Numerous other upgrades have proven that the FPU can either be integrated into the CPU die or present on a "dingus" that plugs into the 386 socket without any need for the "FERR" signal coming from the 387 socket. My best guess would be that intel just used the dummy FPU for marketing reasons (maybe some people would question the empty socket?). Normally I'd cut them some slack, but after the i487 fiasco I'm not willing to overlook the possibility that they might have done something silly.

I used to have the RapidCAD set, and I kind of wish I had held onto it a little longer for testing. It was a rather unimpressive upgrade, since I ran very little software that needed a 486 class FPU. Why do you think they removed the i486 instructions? That just struck me as rather odd.

Anonymous Coward wrote:

It seems that there is one here which has something to plug into the FPU socket: http://www.vintage-computer.com/vcforum/showt … board-386/page3

That extra plug is actually supposed to go under the i385 cache controller chip (for L1 cache coherency).

I'm not sure what the deal is with the RapidCAD "set". Numerous other upgrades have proven that the FPU can either be integrated into the CPU die or present on a "dingus" that plugs into the 386 socket without any need for the "FERR" signal coming from the 387 socket. My best guess would be that intel just used the dummy FPU for marketing reasons (maybe some people would question the empty socket?). Normally I'd cut them some slack, but after the i487 fiasco I'm not willing to overlook the possibility that they might have done something silly.

Or that software just didn't use x87 exceptions very often.

Anonymous Coward wrote:

That extra plug is actually supposed to go under the i385 cache controller chip (for L1 cache coherency). […]
Show full quote

It seems that there is one here which has something to plug into the FPU socket: http://www.vintage-computer.com/vcforum/showt … board-386/page3

That extra plug is actually supposed to go under the i385 cache controller chip (for L1 cache coherency).

I'm not sure what the deal is with the RapidCAD "set". Numerous other upgrades have proven that the FPU can either be integrated into the CPU die or present on a "dingus" that plugs into the 386 socket without any need for the "FERR" signal coming from the 387 socket. My best guess would be that intel just used the dummy FPU for marketing reasons (maybe some people would question the empty socket?). Normally I'd cut them some slack, but after the i487 fiasco I'm not willing to overlook the possibility that they might have done something silly.

I used to have the RapidCAD set, and I kind of wish I had held onto it a little longer for testing. It was a rather unimpressive upgrade, since I ran very little software that needed a 486 class FPU. Why do you think they removed the i486 instructions? That just struck me as rather odd.

All at 40Mhz.

At floats the rapidcad shines.

but only at floats.

other results just for reference.

100Mhz pentium class.

some of the cpus I used.

Scali wrote:

Yes, but a 486SX is really just a 486DX chip, so it has the same pinout. It cannot communicate with an external FPU. Instead, if […]
Show full quote

kixs wrote:

Maybe 486SX has integrated FPU, but it's disabled or non-working - i.e. not present.

Yes, but a 486SX is really just a 486DX chip, so it has the same pinout.
It cannot communicate with an external FPU. Instead, if you put a '487DX' in your system, the entire 486SX is bypassed, and the 487DX is really just a 486DX.
So it's not the same type of bus as a 386/387 system.

FPU wasn't really important yet at that time unless you were using a computer for scientific or mathematical calculations. The average computer user still didn't need one. That didn't change until the Pentium came out and Quake and having an FPU became essential if you wanted to play games.

FGB wrote:

1. There are sockets that carry a normal 486 168-pin cpu to put on a 132-pin 386 socket. Compaq made them, maybe in cooperation […]
Show full quote

1. There are sockets that carry a normal 486 168-pin cpu to put on a 132-pin 386 socket. Compaq made them, maybe in cooperation with Kingston, but I'm not sure.
2. The SXL has full 8KB L1 cache opposed to just 1KB L1 cache in the DLC CPU's.
3. There are clockdoubler chips for the 132-pin 386 socket. The 1KB DLC has the DRx2 (Cx486DRx2) variation and the faster 8KB SXL has the SXL2 (Ti486SXL2) variation.

I have all these CPUs and adapters and I suggest you to use a real 486 platform if you want to use a common 168-pin 486 CPU.
If you like more exotic solution like I do, get an old 386 board and pimp it with some DRx2 or better SXL2 chip.

Has anyone done some benchmarks between the various Cyrix and Texas Instruments 386 to 486 CPU's that were available?
I only have a Cyrix 486 DLC-40 (with the math co-processor) and I would be very interested to know how this CPU's performance compared to the Cyrix and Texas Instruments clock doublers.

vmunix, nice work. What program did you use to benchmark those processors? Curious that the integer add/sub. on the Am486-100 is faster than on the Cyrix 5x86-100. Which register settings are you using for the Cyrix and which Am486 did you use? My results show that only the Am486-100 with WB cache has improved add/sub results over the Cyrix 5x86-100, but the Am486-100 with WT cache has worse add/sub. performance.

jesolo, I too would like to see a comparison of all the "486" PGA-132 chips. One day I will get to it and add the results to the Ultimate 486 Benchmark Comparison. Finding motherboards which work properly with the L1 cache is the difficulty. So far, this is all I have,

feipoa wrote:

What program did you use to benchmark those processors? Curious that the integer add/sub. on the Am486-100 is faster than on the Cyrix 5x86-100.

Passmark Performance Test if I recall it correctly, this was more than a decade ago, the mobo for the 486 was not very good in contrast to what I used for the 386 and Pentiums, that's all I remember, and possibly the Am486 had the magic 8B or 16B at the end.

I recently ran across this old thread while looking for my own thread. Here is an example of a 486-to-386 socket adapter running an AMD Am5x86-160 on a 386 motherboard - www.vogons.org/viewtopic.php?f=46&t=53096

FGB wrote:

386SX wrote:

But if I find the SXL2 66Mhz I'm going to test it for sure.

There is no SXL2 66MHz part for PGA132. The fastest part is 50MHz.

FGB wrote:

kixs wrote:

What about this?
http://www.cpu-cn.com/en/showpic.asp?id=1169&no=1

This is a 144pin QFP 3.6 Volt part. The fastest standard drop in PGA132 part of the SXL range is the SXL2 50MHz.

WRONG

Amoretro, can you not count... !?

This is a QFP 3.6 Volt part with a PGA socket that has... ... ... 132 pins! 😳

386_junkie wrote:

WRONG WRONG WRONG

Amoretro, can you not count... !?

I never denied of the existence of this CPU, but it's obviously you who cannot count: As said earlier in this thread, this is still a 144pin CPU. Sitting on a PGA132 adaptor doesn't make it a native PGA132 CPU. The fastest native PGA132 part is the SXL2-50.

FGB wrote:

The fastest standard drop in PGA132 part of the SXL range is the SXL2 50MHz.

1, SXL2-66 is in the SXL range

2, It is a drop in part, seems to drop into any PGA-132 I have

3, It's the fastest in the SXL range, faster that the SXL2-50

Yep, it seems to be a the fastest SXL2 drop in replacement for a PGA-132 socket.

If not a PGA-132 socket, how else can it be used!?

What about this one ?

also qfp on interposer 😁