386_junkie wrote:

feipoa wrote:

Does it work with an Am5x86-133/160?

It does... can set clock multiplier to x2, x3, and x4... ran it on one of the Forex boards I have kickin around...

Interesting stuff. Are you sure you have that many multiplier options. AFAIK there is only one multiplier select pin (R17 CLKMUL) on the am486/586 processor, thus only two speeds..

gerwin wrote:

386_junkie wrote:

feipoa wrote:

Does it work with an Am5x86-133/160?

It does... can set clock multiplier to x2, x3, and x4... ran it on one of the Forex boards I have kickin around...

Interesting stuff. Are you sure you have that many multiplier options. AFAIK there is only one multiplier select pin (R17 CLKMUL) on the am486/586 processor, thus only two speeds..

I was wondering the same and just assumed it was a typo.

No, it's not a typo... I'm away from home with work so to look again and take another picture of the module will have to wait until return. Below is a picture I took already... unfortunately a jumper is covering two of the pins for clock select.... the x2/x4 functions.... they are essentially the same jumper setting, while in the picture below... there is x3 which is left open.

I've not spent much time in the world of 486 upgrades so this is all new ground for me. With the quick and dirty test I did previously with the Forex board.... I only found speedsys to run at either 133 or 160 (x3 and x4). I'm just wondering then why the x2 is there?

Could it be if plugging the module into a motherboard that has an FSB not of 33/40MHz but instead 66MHz?

OK, so to clarify, the upgrade did not actually run at 2x? I think this is the point gerwin was making.

To run at 2x, there would need to be a different CPU on that PCB, e.g. an AMD DX4 will run at 2x and 3x.

CLKMUL on 486 DX4/5 chips have only two possible states, high or low and, to my knowledge, it is not possible to change the PLL clock generator values as these are hardcoded into the CPU. The only socket 3 CPU which can run at more than two discrete values is the Cyrix 5x86 and this requires software to modify the register (1x, 2x, 3x, or 4x*)

Too exotic to work in an IBM compatible?

A whole bunch of 386 upgrade CPU boards have just popped up In Australia but designed for PC-98 systems, I get the impression they wont work in our retro PCs...

https://www.ebay.com.au/itm/Rare-Buffalo-Hype … 6QAAOSwPbla4sta

Some quick googling suggest the PC-9801 RA/DA are indeed a 386DX-20 released around 1988-1990, this is a Cyrix 586-120 at a 6X20mhz multiplier clock!

The seller has a few other oddballs upgrade cards too.

https://www.ebay.com.au/itm/Rare-i486DX-66MHz … ZEAAOSw7aFa4sm-
https://www.ebay.com.au/itm/Rare-Buffalo-Hype … UcAAOSwrBRa4tJo
https://www.ebay.com.au/itm/Rare-Buffalo-HDA- … QQAAOSw-YZa4sfY

Sorry for the ebay rules but these are something else...

That first link - this is one highly exotic 386 upgrade module. The PCB comes with RAM expansion slots and the proprietary RAM expansion modules. It is also neat how it can take a 16 MHz 386 all the way to 120 MHz by containing a 2x PLL on board. I don't know if it will work on a standard 386 or not. I find the more circuitry on an upgrade adapter, the less likely it is to work on non-intended systems.

The Cyrix 5x86-120 with factory heatsink isn't very common, so if it doesn't work, then whoever buys it will have that in their collection.

The first module with the Cyrix chip is interesting. It can accept its own memory module to bypass the slow memory controller on the 386 motherboard. However, I believe you need a special driver to access that memory. I am pretty sure that driver exists for Windows 95. Not sure about 3.x or DOS though.

In my experience these adapters will work to some degree in regular PCs. I have some Buffalo upgrades for my 286s that have DLC types in them. Those work in my IBM 5170 just fine...even with a DMA SCSI controller. I have had less luck with the IOdata based blue lightning upgrade though. It is pickier about which boards it will work in, and it may not run at optimal speed. Just something you might want to keep in mind before buying. None of these are terribly rare in Japan, BTW.

I will echo this sentiment. I already have a blue lightning adaptor and after finally getting around to trying in many of my 386 motherboards... I found it works in only a couple of them. I was worried that I did something silly and did wrong with the adaptor which made it blow / break... but it was ok, only it was being very picky on which board it wanted to work.

It seems these modules were designed for the NEC computers this seller has... which are all Japanese, and will explain their abundance in Japan.

Does it work with L1 cache enabled, particularly within Win 3.11 or Win95 and with sound working? I have found that checking floppy drive access in Windows (copying files) and trying to play wave or mp3 files is a good first step for checking cache coherency issues.

Despite everything a few of those adapters have bids on them.

I probably should've asked the knowledge tree of vogons before buying but Ive got one of these these adapters on the way...

http://buyee.jp/item/yahoo/auction/b306074311

Its a IOTA adapter with a cyrix 5x86-100 with a 4X multiplier.

Assuming you can unplug the CPU and drop in a IBM 5x86C-100, shouldn't this adapter have a good chance at running 4x33-133mhz?

Looking at the success of the CPUS running at 2X66mhz, would or could the adapter force operation at 4xfsb?

dirkmirk wrote:

Despite everything a few of those adapters have bids on them. […]
Show full quote

Despite everything a few of those adapters have bids on them.

I probably should've asked the knowledge tree of vogons before buying but Ive got one of these these adapters on the way...

http://buyee.jp/item/yahoo/auction/b306074311

Its a IOTA adapter with a cyrix 5x86-100 with a 4X multiplier.

Assuming you can unplug the CPU and drop in a IBM 5x86C-100, shouldn't this adapter have a good chance at running 4x33-133mhz?

Looking at the success of the CPUS running at 2X66mhz, would or could the adapter force operation at 4xfsb?

I bought one of these a few years ago with a similar plan. Unfortunately, it's not that simple. The PLL on that adapter can only handle 100MHz CPUs, and unfortunately a part with the same pinout with a higher frequency rating does not exist. Perhaps it would be possible to rig up an adapter for the IC socket in question to permit the use of another chip.
Assuming you are able to solve that problem, at least as far as I am concerned, that adapter is tailored for specific PC98 systems, and getting it to work on your board may be a crapshoot. On my target motherboard, I was faced with no POST. It worked in another board of mine, but only if both levels of cache were disabled, rendering it somewhat lame. I think Feipoa had better luck than I did in this regard, but it was not reliable at 133 with CPUs known to work reliably at 133MHz (at the correct voltage).
Even if you had such an adapter that worked flawlessly, you're still going to have to find a 5x86 that can handle 133MHz.

I didn't run into the no-POST problem that AC did with this adapter. However, I can confirm that it doesn't run at 133 MHz, even at 3.7 V. The adapter can only seem to manage 25x2x2 = 100 MHz. or 33x3 = 100 MHz. The external 2x multiplier option cannot seem to manage FSB's higher than 25 MHz.

Similarly, when I tried to add a custom 2x multiplier between the motherboard and the CPU, I too could only run up to 100 MHz.

In my experience, Cyrix or IBM 5x86c-100 CPUs which will run well at 133 MHz are rare.

Did any Vogon'ers nab any of the modules?

I would have thought the Cyrix 120MHz with the green heat-sink would have more interest than it did.

Whoever paid $250AUD is going to be a rather sad man unless he has a PC98 to put it in.

A minor correction. That first upgrade adapter is actually for 486 systems. Is there a 386 version of that upgrade?

Anonymous Coward wrote:

Whoever paid $250AUD is going to be a rather sad man unless he has a PC98 to put it in.

A minor correction. That first upgrade adapter is actually for 486 systems. Is there a 386 version of that upgrade?

From my experience with these Transcomputer modules, I would say they can work in some motherboards unintended for them... but it is a case of trial and error to find one that will.

I have two IBM related modules: 486now, and a rather large and cumbersome Blue Lightning (DLC2-66). Both these modules were designed for upgrading IBM systems only and both also have GAL's on them, yet they both only work on only a few certain systems... I expect these Buffalo modules will all be the same.

Come to think of it... was your Blue Lightning upgrade module not a Buffalo?

EDIT: -
I saw there were floppy disks that came with these modules... it made me wonder if there were programs included which could manipulate the CPU registers and functionality.

You need to play around with the register settings on all BL3 upgrade modules. The PAL's and GAL's on the Buffalo and IOData modules don't setup the registers - the drivers do. The diskettes which come in the boxes aren't readable on our western systems as is. They seem to have some strange language formatting or type setting or something such that the drivers won't run normally on our systems. Maybe if you setup your MS-DOS for the Japanese language you might have a better chance.

Did you buy the Buffalo for $250? I didn't buy any of these. I have enough. They are frustrating to get running. I've spent weeks fooling around with the registers to get them running on non-BL3-aware motherboards.

feipoa wrote:

They are frustrating to get running. I've spent weeks fooling around with the registers to get them running on non-BL3-aware motherboards.

This sounds like you had success?

On only a handful of motherboards like me... or any / all motherboards?

With the Buffalo, I was able to run all my motherboards with all DOS benchmark I threw at it after determining the right combination of start sequences and registers, however, I was having trouble getting Windows 3.11 started with L1 cache enabled. I tried so many things that I forgot exactly which combination of things allowed the Buffalo to work in DOS. The IODATA module has a DIP switch which allows the cache to flush on I/O operations; when set, Windows 3.11 loads just fine. I ran my IODATA up to 110 MHz reliably (120 was too much for it). I'm not finished with testing and optimisation. I ended up putting the stuff away many months back to work on less stressful things.

How come I missed this thread altogether??

What do you think of this adapter? It seems to have a 168 pin 486sxl2-50 CPU and 132 pin underside - drop in replacement for 386DX. Or am I missing something?

I discuss that adapter in this thread, Re: Custom interposer module for TI486SXL2-66 PGA168 to PGA132 - HELP!

Did you get one? I hope you didn't pay the stupid amount this UK eBay seller was asking. He has 30 of them and isn't a very pleasant fellow. He asked for assistance and I provided many paragraphs of explanation of what these are, what they are good for, probable value, and he simply ignored me. There was one other unit for sale in the US, if I recall right.

This adapter is from Improve It Technologies and doesn't have the voltage regulator, so you are stuck with a 5 V SXL2 PGA168. If you view the link above, you'll see that I added a voltage regulator module to mine and was able to get a 3.6 V SXL2-66 working, but it wouldn't clock to 80 MHz like my QFP SXL2-66 upgrade unit would. I also had to modify my voltage regulator module for use with the SXL2 chips, so, not the most ideal solution for a SXL2-66. We need to get our own PCB designed to get an SXL2-66 3.6 V chip working at 80 MHz on a PGA132.