A BIOS option "Slow refresh" will usually bring the DRAM refresh rate to 1/4 of the original value - some 64ms vs 16ms. Note that the performance improvement is single % at best (depends on the CPU, I suppose it would matter more for slower CPUs with no cache) while the risk of glitches raises considerably. Espcially at higher temperatures. It does save some power though so it's mostly used in laptops equipped with low-power DRAM chips that are less susceptible to charge drift.

Anyway, you can't use utils meant fo Cyrix on a "real" 486. A full-featured 486 can both invalidate cache entries one by one with bus snooping or flush everything, a DLC/SLC CPU can only flush. Both can use KEN signal but Cyrix has additional internal registers to define cache exclusion zones CPU-side. There's no burst transfer on Cyrix since there are no signals for chipset to control these, even if the chipset was a 386/486 combo one that actually could do it.
All that extra glue logic on the adapter PCB might allow certain "hacks", like maybe if the chipset does drive CPU address lines properly during DMA cycles you could use that for cache invalidation. If not you can only flush, and detecting when to do it might not be possible so you have to flush on each bus hold. There are software workarounds like trapping BIOS disk interrupts and doing INVD/WBINVD in the TSR, but that works only in DOS and any app doing low-level disk/floppy access, or using protected-mode 32-bit OS-native drivers, will fail.

In general I'd suggest using utils that came with the CPU adapter and only that. Anything else is like trying to drive a nail in by throwing a hammer at it. Might just work if you hit it right but mostly will not. There might be some hidden features or parameters not explained in the manual so disassembling the app might help you learn more - but note that most TSRs use various coding tricks and obscure DOS functions and structures so analysis is usually difficult.

Hi Guys,

Thanks for all the input - I've checked multiple websites and they all say that the Kingston 486 Now SX version *should* be able to accommodate a DX, but it's just not true, I've tried all types of DX CPUs and they all hang during POST. Perhaps it's just my old compaq not being able to handle the upgrade card, but I won't know for sure as I don't have another 386 computer lying around. This ancient 386 (one of the first 386 boards from 1987) doesn't have a BIOS, it just has a basic configuration utility that runs off floppy, and jumper switches (the only CPU related ones are to run the cpu at a lower speed for backwards compatibility with software that would run too fast on a 386).

So that leaves us with just using SX based CPUs, and the fastest would be an SX2-66 by AMD or the SX2-50 by Intel.

I found that there is software for this card (486NOW.exe) so I will download that and try it out and see if it will help with the L1 cache issues when using the SX2. I'll also download SETMUL and play around with that to see what comes of it. I'll post my results back here this weekend, hopefully at the very least I can get one of the SX2's to work.

Thanks again guys!.

DX and SX cpus have a slightly different pinout. Compare the Kingston 486 NOW SX version to the DX version. It's possible that a few jumper or resistor packs have been moved between the models. I believe some of the 3.3V upgrade CPUs might also be designed to plug into 486SX sockets.

Where did you find 486NOW.exe ? I'd like to try that on my Transcomputer module. For 386 to 486 upgrade kits from Kingston, I only have the software for an IBM Blue Lightning.

There is a 3rd possibility

Unless the upgrade has support circuitry
386 boards required a special chip in the fpu to allow an onchip coprocessor

So Perhaps it’s the same issue as the rapidcad

I've just tried shorting A15 and B15 with a wire and it works, but the result is a bit inconclusive.

The problem is my mobo booted with 486DX set as SX even without the wire, which I was not expecting. I suspect the NMI pin floated with charge high enough to register as logic H (there is no pull-up), so the system didn't randomly freeze. In both cases I was able to run some DOS programs, including those that make use of co-processor - that is actually expected, what will hang the system is unmasked numerical exception. Most software avoids exceptions or ignores them. NSSI would freeze on FPU detection unless run with /SAFE switch.

Well, this doesn't guarantee that a DX with a jumper will work in place of SX in all cases, as I've said the BIOS could be doing something weird that would trip this setup. But it gives me some hope that mobos with 486 sockets wired permanently for SX processors could be made to work with DX, mostly. It's way easier and cheaper to find Intel DX2-66 than AMD SX2-66. I have one now but if this works, then perhaps a DX4 overdrive could be made to work as well.

Ok So I finally had some time to test out this board, both in the machine it was designed for (IBM portable P70) and my Compaq portable 386.

I can confirm that it only accepts SX type CPUs - I also managed to get my hand on a Green CPU 40mhz variety - which I can confirm is faster than a regular SX. I also can confirm that this upgrade module won't accept any SX2 CPUs.

I haven't yet had time to upgrade the crystal on the board to overclock the cpu.

On the software front - the 486NOW.exe just contains generic diagnostic software published by Landmark (the same landmark from the CPU Benchmarking software) and does not have any utilities to affect the operation of the upgrade board.

SETMUL could not communicate with the CPU as it threw errors saying that it could not identify the CPU.

I haven't tried shorting pins A15 and B15, might give it a try and see what happens.