This page here says the SCL3 (16bit) is a DLC (32bit) chip except with the extra address pins disabled.

"All known instances of 486SLC3 CPUs are reportedly in the 132pin PQFP packaging with the extra address lines not connected (a 486BL3 running in 16-bit mode)"
https://alchetron.com/IBM-386SLC

Not sure where the knowledge comes from or is 100%, but it makes sense.

Pinout is the same as 386dx.
link

pshipkov wrote on 2024-03-29, 07:48:

Pinout is the same as 386dx.
link

It can't be that simple. The 386DX has no pins assigned to A20, Flush, Ken, Barb, etc... without those cache coherency and control aren't possible. So, at a minimum, some of the N/C pins are assigned to those functions. Are the differences between the 386DX pinout and the SLC3 documented anywhere? If those differences are documented somewhere, that is all that is needed to complete the understanding of the SLC3 pinout. However, to the best of my knowledge, this isn't available anywhere.

i know. it sounds improbable on first read but thats ok. couple of us walked that path already.
if you flip through the same thread you will see what i mean.
there is another thread by @Feipoa that contains very good info as well.

I didn't look into the SLC3 specifically, but rather looked into the DLC3, or BL3. It was determined that a QFP 386DX could be removed from a PGA interposer and a BL3 installed. I didn't come up with this idea, but rather first saw an eBay listing (~3 years ago) for such a conversion and decided to investigate.

My rendition of the adaption is here: Re: Creating a voltage regulated 386 BL3 module from existing parts I decided to down regulate the voltage and add filtering caps. It was a straight one-to-one pin fitment, I put no thought into cache invalidation. The adaption just works as far as I can tell. I now have a whole system setup around this adaption.

Unfortunately, these DLC3 chips are hard to come buy and expensive.

feipoa wrote on 2024-03-29, 22:56:

I didn't look into the SLC3 specifically, but rather looked into the DLC3, or BL3. It was determined that a QFP 386DX could be removed from a PGA interposer and a BL3 installed. I didn't come up with this idea, but rather first saw an eBay listing (~3 years ago) for such a conversion and decided to investigate.

My rendition of the adaption is here: Re: Creating a voltage regulated 386 BL3 module from existing parts I decided to down regulate the voltage and add filtering caps. It was a straight one-to-one pin fitment, I put no thought into cache invalidation. The adaption just works as far as I can tell. I now have a whole system setup around this adaption.

Unfortunately, these DLC3 chips are hard to come buy and expensive.

Do you think your adaption of the QFP 386DX/BL3 pinout could be derived down to a non-standard 100-pin PGA? The reason I ask is that I want to adapt the IBM 486SLC3 I have to work with a 386 SX motherboard that has a Yamaichi QFP socket that allows for the PQFP 386 SX processor to be easily swapped out for anything that is pin compatible. My idea is to remove the Yamaichi QFP socket, which is attached to the motherboard using a 100-pin PGA and replace it with an adapter that will allow for the attachment of the QFP132 486 SLC3 in the pinout of the Yamaichi QFP socket.

Also, did you use the RevTo486 software to control the cache?

i should quit the habbit of responding with brief messages in passing, but do it properly when time is right.

the point i wanted to make was that 386dx and bl3/dlc3 are pin and voltage compatible, so i would assume the same applies to 386sx and slc3.

it makes sense to first confirm that by swap a 386sx chip in place.
from there on additional measures can be implemented, like what Feipoa did with his bl3 contraption.

Do you have a photo of this PGA-100 to QFP100 interposer? I don't think I've seen one. I've only seen PGA68 to PLCC68 interposers.

My best guess is that the 100-pin QFP, 16-bit, IBM SLC2/SLC3 would also be pin compatible with the Am386SX QFP100, but I have not tested this. 386SX systems have never peaked my interest and I do not own any.

Why do you want to remove your PGA100 to QFP100 interposer to test out the SLC2/SLC3? Wouldn't you merely just use hot air to remove the 386SX and replace it with the IBM SLC2/SLC3? What is your source for these IBM SLC2/3 chips? If it is on an upgrade interposer, wouldn't you just use the upgrade interposer? They usually contain a QFP100 clip-on socket.

feipoa wrote on 2024-07-30, 07:55:

Do you have a photo of this PGA-100 to QFP100 interposer? I don't think I've seen one. I've only seen PGA68 to PLCC68 interposers.

My best guess is that the 100-pin QFP, 16-bit, IBM SLC2/SLC3 would also be pin compatible with the Am386SX QFP100, but I have not tested this. 386SX systems have never peaked my interest and I do not own any.

Why do you want to remove your PGA100 to QFP100 interposer to test out the SLC2/SLC3? Wouldn't you merely just use hot air to remove the 386SX and replace it with the IBM SLC2/SLC3? What is your source for these IBM SLC2/3 chips? If it is on an upgrade interposer, wouldn't you just use the upgrade interposer? They usually contain a QFP100 clip-on socket.

This is the QFP100 to PGA100 socket adapter:

You can just pop the processor out, and put anything else that is QFP100 in its place, assuming it uses the same or similar enough pinout as the processor it is replacing. It's how I changed the 386SX-20 processor in the brother machine to the system I am talking about to a TI 486SXLC2-40. The TI 486SXLC2 is fortunately QFP100, with just a few differences from the 386SX pinout.

That's the issue with the IBM 486SLC3-60, it's not QFP100, it's QFP132, which is why it isn't considered pin-compatible with any other 386sx design that are limited to QFP-100. It is generally believed that the IBM 486SLC3's are actually BL3 processors (explaining their QFP132 packaging) with their 16/32 bit mode selection pin pulled high permanently so it will only operate in a 16-bit bus mode, with the other 16 lines that would typically be used left N/C. If the IBM 486SLC3-60 is permanently set to a 16-bit bus mode, it should literally act identically to any other 386SX replacement processor. Since the 16 unused bus lines can be discounted, that leaves a 16 pin difference between it and the ability to function in a QFP/PGA 100-pin design. This is why I want to know the pinout of the IBM 486SLC3 for certain, once that is known I can see how many pins can be outright eliminated as N/C and what else needs to be done to make it work in a QFP/PGA 100-pin design, if possible. If it does have a simple 386DX QFP132 pinout, no need to bother trying to determine the pinout, since it is known. At that point I just need to figure out what needs to be done to bring it down to a 100 pin count, if possible. If true, I'm quite surprised since that means the IBM 486SLC3 eschews the typical pins associated with cache control and the cache must somehow be completely controlled by software, which I didn't even know was possible. The TI 486SXLC2 uses some of the N/C pins found on the typical 386SX in the QFP100 package as cache control pins (BARB, FLUSH, etc...), along with software cache management. My assumption is that the IBM 486SLC3 would do the same thing for cache control, reassigning some of the N/C pins in the QFP132 package to cache control functions. But, if it is all somehow controlled completely through software, then I it wouldn't. The fact that, to the best of my knowledge the pinout of the IBM 486SLC3 isn't documented literally anywhere on the internet, it can't be known what the pin assignments are until someone can determine such.

So, ultimately, if possible, I want to adapt the IBM 486SLC3-60 in its QFP132 package onto an interposer that matches the pinout of the PGA 100 side of the socket adapter depicted above and replace the socket adapter with it.

The source of the IBM 486SLC3-60 is a NOS upgrade for certain models of PS/2 that had an upgrade socket specifically for this upgrade. Unfortunately, the pinout of the upgrade prevents it from being used directly because it is some sort of PGA setup that was specific to the upgrade socket in the PS/2's it was intended for, it is not a match for the PGA100 pinout of the socket adapter that is currently on the motherboard of the system I want to put it in.

I remember reading an article on the SLC3 in a 1993 issue of PC magazine. There was a picture of it, and it was PQFP100.
I'm pretty sure whatever you have on on your upgrade module is a 486BL3 in 16-bit mode. I don't think the real SLC3 made it to production, as nobody has ever found one in the wild.

I have a couple of those 1oo pin PQFP to PGA adapters, but from a different company. I've only ever seen one or two motherboards that would actually accept them, but they were of such old design I couldn't be bothered. Can you post a picture of the board that you're using?

Intel actually produced a military grade version of the 386SX in ceramic PGA, but it seems to exist in VERY small numbers. This kind of establishes an "officlal" pinout though.
I've seen several 386SX boards that appears to have solder pads for test sockets, but it seems none of them really followed the MG386SX pinout and all used their own pinout.

Anonymous Coward wrote on 2024-07-31, 03:01:

I remember reading an article on the SLC3 in a 1993 issue of PC magazine. There was a picture of it, and it was PQFP100. I'm pr […]
Show full quote

I remember reading an article on the SLC3 in a 1993 issue of PC magazine. There was a picture of it, and it was PQFP100.
I'm pretty sure whatever you have on on your upgrade module is a 486BL3 in 16-bit mode. I don't think the real SLC3 made it to production, as nobody has ever found one in the wild.

I have a couple of those 1oo pin PQFP to PGA adapters, but from a different company. I've only ever seen one or two motherboards that would actually accept them, but they were of such old design I couldn't be bothered. Can you post a picture of the board that you're using?

Intel actually produced a military grade version of the 386SX in ceramic PGA, but it seems to exist in VERY small numbers. This kind of establishes an "officlal" pinout though.
I've seen several 386SX boards that appears to have solder pads for test sockets, but it seems none of them really followed the MG386SX pinout and all used their own pinout.

Here is the motherboard (identical to mine):

It dates from around Q1 1993.

Yeah, the pinout of the Yamaichi s0cket adapter likely doesn't follow any typical convention.

I also checked the original IBM upgrade card and I was correct in remembering it using a PGA that I am completely unfamiliar with, and bears to relation to any of the pinout standards used for the 386 Intel processors, it's a PGA122, with 5 pins missing (clearly by intent when it was made). I don't know of anything else that uses PGA117 of 122 pins.

Given that a QFP132 to PGA100 interposer that matches the Yamaichi PGA100 pinout is highly unlikely to exist, it will be a custom job. I planned to pay someone to design it, along with an integrated VRM, since the 386SX typically runs at 5V, and the IBM 486SLC3 runs at 3.6v

One thing that was posted on here by Furan is what leads me to believe the IBM 486SLC3 doesn't match a 386DX pinout 100%:

IBM 486SLC/2 Mega Thread (was Weird idea: 83mhz overdrive for 386 sx.)

IBM 486SLC3/486DLC2: PQFP 386DX pinout with addition of some pins for cache control, suspend mode, & CPU bus width modes. Hardware pin switchable between 24-bit Address bus/16-bit external Data bus or 32-bit Address/Data bus. 16Kb internal L1 cache. Able to run all Intel 486SX instructions. Additional Model-Specific Register from the 486SLC2. Low-power design. Internal clock tripling at 60, 75, or 100MHz. CPUID A439h in 486SLC3 mode.

On that page he further quotes from a publication the various register bits that can be altered regarding pins that don't exist for the 386DX pinout.

If they used such registers and pin assignments in the IBM 486SLC2, it seems highly unlikely they would omit such pins in the 486SLC3.

However, how true any of this is, is hard to say, since there is so little data on the IBM 486SLC3. That's why I wanted help in determining the exact pinout of the IBM 486SLC3, to find out what pinout it has with 100% certainty, does it match the 386DX pinout? Does it match the IBM 486DLC2 pinout? Does it match the IBM 486BL2 pinout? There is conflicting information with no actual pinout provided by anyone.

I still hope to find someone who is willing to analyze the IBM 486SLC3 pinout and verify 100% what it is.

Something else I've been thinking of, all the Japanese Evergreen SLC3 upgrades, which also use an IBM CPU with a QFP132 pinout, tend to have a Cyrix DLC87 FPU on their board. Because it is likely to be a BL3, just running in SLC mode, does it need a DLC type FPU given its design, or since its running in SLC mode, it would require an Cyrix SLC87 FPU? Does anyone even know the difference between the Cyrix DLC87 FPU and the Cyrix SLC87 FPU? I haven't been able to find much information on how the two differ.

Paralel wrote on 2024-07-31, 17:11:

Something else I've been thinking of, all the Japanese Evergreen SLC3 upgrades, which also use an IBM CPU with a QFP132 pinout, tend to have a Cyrix DLC87 FPU on their board. Because it is likely to be a BL3, just running in SLC mode, does it need a DLC type FPU given its design, or since its running in SLC mode, it would require an Cyrix SLC87 FPU? Does anyone even know the difference between the Cyrix DLC87 FPU and the Cyrix SLC87 FPU? I haven't been able to find much information on how the two differ.

I have also wondered this, but have no answer to you.

Turning back to your original system upgrade plans, one aspect I don't quite understand is: If your system has one of these neat PGA100-to-QFP100 clamping type CPU sockets, have you tried a regular IBM SLC2 QFP100 in it to see how this CPU performs? If they SLC2 is anything like the BL3, it should have quite the overclocking headroom. The BL3, which is often rated for 75 MHz, can routinely be taken to 112 MHz with a bit more voltage (4.2 V from my tests). Without the extra voltage, they can sometimes reach 100 MHz, or worst case can do 80 MHz.

feipoa wrote on 2024-08-01, 06:38:

Paralel wrote on 2024-07-31, 17:11:

Something else I've been thinking of, all the Japanese Evergreen SLC3 upgrades, which also use an IBM CPU with a QFP132 pinout, tend to have a Cyrix DLC87 FPU on their board. Because it is likely to be a BL3, just running in SLC mode, does it need a DLC type FPU given its design, or since its running in SLC mode, it would require an Cyrix SLC87 FPU? Does anyone even know the difference between the Cyrix DLC87 FPU and the Cyrix SLC87 FPU? I haven't been able to find much information on how the two differ.

I have also wondered this, but have no answer to you.

Turning back to your original system upgrade plans, one aspect I don't quite understand is: If your system has one of these neat PGA100-to-QFP100 clamping type CPU sockets, have you tried a regular IBM SLC2 QFP100 in it to see how this CPU performs? If they SLC2 is anything like the BL3, it should have quite the overclocking headroom. The BL3, which is often rated for 75 MHz, can routinely be taken to 112 MHz with a bit more voltage (4.2 V from my tests). Without the extra voltage, they can sometimes reach 100 MHz, or worst case can do 80 MHz.

I already have a brother system to this one that has been upgraded from a 386SX to a TI 486SXLC2-40, and the only real difference between the TI 486SXLC2 & the IBM 486SLC2 is the additional 8k of cache. I figure the difference between the two is likely to be small. Also, the IBM 486SLC2 doesn't run at 5V, does it? If not, it wouldn't be compatible with the socket in question since it only operates at 5V. However, based on benchmarks people have put up for the BL3, even blunting that for operation in 16-bit bus mode, the difference between the TI486SXLC2 and the IBM 486SLC3 is likely to be substantial, especially given its overclocking potential since it will run quite a bit cooler at ~3.3v instead of 5V.

I now know for sure that the IBMSLC3 is indeed a BL3 in 16-bit bus mode. I was able to match the ID on the chip to the picture of a known IBM BL3.

It also appears that the cache, etc... is indeed all controlled through software, which I find very interesting. It appears that the IBM 486SLC3 shouldn't work any different than an Evergreen Revto486 upgrade.

So, I just need to create a 386DX QFP to 100 PGA interposer with voltage regulation to make this all work. I doubt any exist currently because I honestly don't know of anything that would use a BGFP-132 to PGA-100 interposter.

I'm guessing the BS16 pin of the 386DX pinout is what allows for the selection of the 16 bit bus so it will work in an SX system.

Have you tried to run the IBM SLC2 at 3x via software? How fast have you been able to take the IBM SCL2 up to? What is your target frequency for this system?

feipoa wrote on 2024-08-06, 19:50:

Have you tried to run the IBM SLC2 at 3x via software? How fast have you been able to take the IBM SCL2 up to? What is your target frequency for this system?

Unfortunately the chipset really sucks, and I can only get the CPU up to 40 MHz. My target for the SLC3 is 60, which is probably underclocking the chip, since I believe it is natively a 25/75 chip. Plus, the system is entirely passively cooled, and the location of the processor doesn't leave much overhead, so a fan isn't really possible. The best I can do for heat is a low profile heatsink.

I think might be able to get it to 75 if I replaced the chipset with its successor that is pin compatible, but that might be going too far.

In many ways it's a waste of a chip, only using it up to 60, since, according to a post by pshipkov, it can apparently clock up to 100-120.

At least I now also know it will work with a Cyrix 387 SLC FPU, I was able to find a motherboard that is natively SLC3 that IBM created that has a socket for an SX compatible FPU.

If the spot UNDER the cpu has a clear shot to sheet metal or there is empty board there (or only ground plane). over sheet metal you could drill hole(s) avoiding traces and place a bit of copper wedged between the bottom of the chip and the sheet metal through your drilled hole(s) make a spreader if you want or even drill holes in the sheet metal to put a fan there underneath where there is more clearance.

I’ve only seen this done once on some sort of industrial machine, a copper “standoff” was bolted to the sheet metal and was a precise length to touch the bottom of the cpu, the sheet metal had holes in it like you use for a speaker and a low profile fan
but it definitely moved heat, guessing it works best on old not flip chips.

Other option is a thin but very wide heat sync on top, would work as a heat pipe to get around stuff in your way like a laptop

If you mainly want to test high speeds, I've had a TI486SXLC2-50 at 90Mhz before, but as multiplier is 2x, you have to take the system clock high.

Memory speed is still the bottleneck... While L1 scales with CPU clock, RAM stays with the system clock and wait states.

3DBench results (C&T 8x2836B-25, One CAS Extend WS, 16-18Mhz ISA, CL-GD5429):
45Mhz system - 25.6
45Mhz system/90Mhz CPU - 27.0

50Mhz system - 28.5
50Mhz system, no fpu - 29.4

The best coolers are these raspberry pi heatsinks & fans. You can get mini-tower coolers as well, but this fan keeps the heatsink cool just the same... Size is 25mm square. 5v.

Has it been determined if the QFP100 IBM SLC2 chips can run at 3x via software? Or did IBM only unlock the 3x multiplier with the QFP132 series of IBM Blue Lightning chips?

I once bought a Raspberry Pi fan, I think from digikey, and it was so weak that I couldn't feel any air flow from it. I think it was a 30 mm fan.