Unbelievable what happened a little while ago!!!

Lately, I've been working on some 486, including an AT Industrial UMC chipset, searching on TRW is the same as a Power Tech MB457, well to make it short, lately (years ago) I had tried to start it with an Intel 486SX25 CPU, unfortunately nothing happened, no post code on the ISA-PCI card (here obviously you can only use ISA), try various RAM both 30 and 72 PIN, an Intel 486DX33 CPU (which I was looking for but still can't find), and try and try again and nothing!

Today I finished cleaning the cache chips of the 486 PCI (Terminator), and I wanted to change CPU to temporarily install the DX33 instead of the AMD DX4-100, I managed to find all the 486 I have, except the DX33, maybe it's in a motherboard, I'll check.

So having a free space on the bench, I take the MB457 one, and I place it ready for a further startup attempt, but first I have to recover an ISA VGA and a RAM, I connect the AT keyboard and ATX-AT adapter (for the power supply), I use an old WD card for the VGA, and for the RAM I take a 70 ns 72 PIN with 8 chips, I thought it was 8 MB, in reality it is 4 MB, it doesn't matter anyway.

So I try to start it's incredible what happened: Codes appear on the post card, and you hear a beep, on the video I have the BIOS screen!!!

But why all this surprise ???

The last time it worked, was 30 years ago, then nothing!

Now I'll try to make some additions, like a controller card for disks, but maybe also the 30 PIN RAM, to have a total of 8 MB, then if it works maybe I'll try with a 72 PIN RAM of 8 or 16 MB, replacing the 4 MB one.

One last problem I'm having is that I'm almost without VGA ISA, unfortunately I've been looking for a Trident (8900) for many years, but the requests are too high, I was looking for this model because it's the same as the one I had in the early 90s, then of course it won't be identical but for me it would be fine in a PC like that, even if originally I had a Cyrix DX2-80 that was quite fast, this SX-25 I don't know but I think it's at the same level as a slightly faster 386DX, even if it has an 8 KB L1 cache, which at least does something, since the Motherboard does not have L2 cache on board.

I started to do a first upgrade, to the MB457 motherboard, I added a controller, unfortunately among my stocks I only have another one the same, they are 16 bit ISA controllers for HDD and FDD, no serial and parallel ports, and anything else 😨 there may be, an old and simple controller for drive A: B: C: D:, so two FDDs and two HDDs, but also in the BIOS there are two FDDs and two HDDs, so if we want to connect other drives, I think it is a must to add a second card (probably?).
For the moment I recovered an old 540 MB HD (approximately) from IBM, years ago I had removed it because after loading DOS and Windows 3. X, the next day it didn't start anymore, it lost the boot files, even if the surface of the disk was ok, without any defective sector, I'll try again and see what happens, I hope 🤞 that the problem was not unsolvable, and that maybe it can be recovered ❤️‍🩹
Since I have the ISA controller, I connect the 1.44 MB floppy player 💾, set the PC from the BIOS because by default there is the 1.2 MB one, I insert a bootable disk with MS-DOS 4.01, and the PC boots, at this point not having other disks at hand, I execute some commands, like Ver or Mem, just to do something.
From what I remember, this controller does not recognize the IDE CF adapter, but maybe I could try to see if the IDE SD one works, in case I would make a bootable disk in an SD memory card.
Finally, I wanted to send the images of the manual to TRW, actually it is a sheet folded in two, very simple, but anyway the card is also very simple, there are not many jumpers, and you can select 4 FSB, 25 33 40 and 50 MHz.
At this point I'll add them here, then if maybe someone can send them, that's fine with me Thanks

I noticed that when starting the AT Industrial motherboard, with the 486SX25, it reads PT-2068.1, this instead of MB-457, this means that the motherboard is actually a PINE PT-2068.1, I find out now after over 30 years!
I also have the original 📦 box, unfortunately it doesn't say anything about the brand or the model, maybe in that period (mid-90s), there were many manufacturers and some preferred to remain anonymous (or generic), to avoid being labeled as producers of poor quality cards.
Seeing the board in operation, I got an idea 💡 about what could have happened, on the Cyrix DX2-80 5V, they had put a heat sink plus fan, but without thermal paste 🍝, this could have damaged the CPU, which initially worked badly, and then it broke 😑, maybe the PC in the laboratory 🥼 was left on (?).
Currently it seems to work very well, even if I haven't connected a card adapter or a HD yet, I haven't upgraded, especially the memory and CPU, also because I want to use it a little like that, for the memory I could add the 4 SIMM 30 PINs of 1 MB each, at the moment it's not needed, but maybe it would be the best thing, go directly to 16 MB with a 72 PIN, currently I use a 4 MB single face, with 8 chips TI TMS444000DJ -70, so I would like to put something equivalent maybe from -60, I have no idea 💡 if it's possible to use EDO RAM.
The BIOS of the PINE, seems simplified compared to 🫡 to the MB-457, maybe as soon as I have a disk connected, I make a copy of this, for a possible comparison with the other, which is on TRW, I saw it on YouTube and it seemed to have several more lines, and unlike the manual, it seems that the FSB can be set from 20 up to 80 MHz, in total there are 8 possible combinations.
I finish with the ISA sound card, I selected among those I have this one with Diamond DT-019H chip, you can see it on TRW, but in the image where the connectors are marked, there is an error, the 40 PIN one that could be used to connect a CD-ROM player, it is marked as JP1, while in reality it is JP2 as you can see from the photo.

Just finished installing the 527 MB IBM DSAA-3540 disk, on the ISA controller of the Pine PT-2068.1, via bootable floppy, I loaded MS-DOS 5.0 ITA, initially I encountered a problem, via FDISK I activated the partition (Menu 2), then I formatted the disk, and I made it bootable, I copied all the files present in the floppy, and I added a program called PCTOOLS, which includes some useful things, for disk management, there is also a bench in System Information Service, which gets the following score:

Relative speed (orig PC=100%): 3290%

As an absolute score, it's not great, a DX2-66 makes about 7000-7500, if I remember correctly the DX2-80 is about 8500, let's say we are at the level of a 386DX-33, I don't know if more or less.

Now we have to wait, tomorrow or in the next few days, to see if the disk continues to work, I had put it away because after a while it was inactive, it didn't start anymore, maybe the active partition is memorized (?), we'll see if something happens, for the moment it's fine, it's also true that currently I haven't loaded Windows 3.1, it remains to be seen if the problem occurs occasionally or punctually.
Only the ISA card with serial and parallel ports is missing, maybe I could use a parallel cable, to connect another PC, through the Laplink program, and in this way I could transfer the Bench files, such as those of Phil's HWINFO ASTRA NSSI QTPRO etc...

When the transfer is finished, I should start the benchmarks and see with the diagnostics, if the hardware doesn't show problems, if everything goes well, I would say that the PC is almost ready, except for the ISA video card, which I will replace, the rest could stay like this, but if it had more RAM and a faster 486, it would be better for me, unfortunately I don't like the 486SX very much because it's a lame CPU, it's not possible to add the Coprocessor, even if maybe it's almost useless in DOS, except using spreadsheets or processing and CAD programs.

Unfortunately I removed the 486DX33, because I put it in another PC, the problem is that being a card without a linear regulator, it does not support CPU with a voltage of less than 5V, and the maximum is 80 MHz for the Cyrix DX2, or 100 MHz for the iDX4 Overdrive, here both have a certain charm, but being an ISA card, probably the best of the two is the DX2-80, because the DX4-100 could be used in a 486 VLB, which would have L2 cache on board.

I keep transferring files, from the QDI Advance 10T (P3S-1266), to the PINE PT-2068.1 (i486SX25), unfortunately the second 1.44 MB floppy reader does not read, so I do the transfer, removing the working one, and moving it to the other PC, after all I don't have to copy much, three attachments and re-attaches were enough, to copy some bench and diagnostic programs, and these showed the actual power (scarce!) Of the i486SX25, one sees it equivalent to a 286, and all in all we may not be so far away.

I think this card can be considered an NOS, because in fact it is, except that it was purchased in 1995!, and it was kept in its box until the day before yesterday.

Looking at the info provided by the diagnostic programs, there is very little, that I remember in other 486, there were more details, maybe there is almost nothing here, because the MB is very essential, but despite this, even if slowly, until now it has worked without showing any sign of overheating, the CPU is strangely fresh, maybe it takes hours to heat it up.

Let's move on to the Bench, I've only run a few, of Phil's there is little that works, the main reason is the CPU without a coprocessor, some show it clearly with a message, in short, forget about 3D and similar stuff, in the end I only used 4, let's see them:

Landmark Speed 6.00 (119 MHz AT)

CPU: 118.67 MHz - FPU: <N/A>

Video: 2002.12 chr/s

NSSI 060

CPU Performance: Intel i486SX, 25 MHz

14536 Dhrystones/s (* 14602, 16 bit code)

SpeedSYS 4.70

Processor Benchmark 9.17

QTPRO

CPU benchmark

10013 Dhrystones - 8232k Whetstones

Comparative System Benchmark

CPU Speed: 26 MHz

xBench Speed: 70.09 MHz (486SX25 = 69)

Fortunately, the 527 MB IBM disk did not lose memory, in particular it was the active partition that memorized, which then made all the contents of the disk illegible.

If there were benches for Windows 3.1, maybe it would be installed, but from what I remember, it would take at least Windows 95, now with just under 4 MB available, trying to install it is not the case.

I finished transferring files via floppy, some bench and diagnostic programs, they are too big to put them on a floppy, I had to give up copying them.

Let's say that maybe something is still missing, I don't know exactly which one, I'll do a check, eventually I can transfer it, always via floppy.

Even with these other programs, the results are probably adequate for an i486SX, moreover I believe that the frequency, relatively low (25 MHz), also helps to keep the scores low.

Maybe even the disk, it's not really fast, from what I saw on a version 3.05 of HWINFO, it has an RPM of 1090, an AVRGseek of 11.4ms, and a transfer rate of 12.576 KB/s, but the values should not be compared with the latest generation disks, but with those of a mid-90s HDD, and maybe they wouldn't be so bad.

Since I have concluded, at least for the moment, the SW tests, I made a small upgrade, yes because it is true that in its time 4 MB of RAM, were enough for most of the DOS programs, but having more, you can run even the most demanding ones, or make it possible to install and use Windows 95 in a dignified way.

Here there is the possibility to use only the 30 PIN RAM or the 72 PIN ones, or use both (at the same time), the problem with the 30 PIN ones is the difficulty of finding the 4 MB ones, the cost can be considerably higher than the 72 PIN ones, so I decide to change the 4 MB 72 PIN one with a 16 MB one.

Upgrade successfully completed!

The configuration of the PINE PT-2068.1 with i486SX25, is approaching the final one, in fact I would say that we are almost there, because there is very little left to do, exactly what ???

The increase in RAM, could be done using only 30 PIN RAM, I would have four SIMMS of 1 MB each, but I'm pretty sure that having 20 MB or 16, wouldn't make much difference, if instead I had 32 MB, the difference would be such as to make Windows 95 work better, however if the current 16 MB weren't enough, I could temporarily bring them to 20 MB (16+4), but I don't see the utility, rather those 4 MB could be useful, if in the future I recover some old motherboard 386-486 with 8 RAM sockets 30 PIN, in short I don't think it can happen soon, but You never know in the future...

I added the ISA sound card, I didn't try to make it work, also because I would have to track down the drivers, and considering that I don't have sound cards with the same chip, I don't even know if it's possible to use a special diagnostic program, to check the functionality and or change the settings, I think you have to load some drivers, when starting DOS, to be able to use it with the various Games.

Previously, I added an 8-bit ISA card, with 2 serial ports, 1 parallel, 1 game, to one of the serials I connected a mouse 🖱 (serial), and it works (!), the parallel could be used for the transfer of files through the Laplink program, and not only that, the Game port instead is a duplicate, since it is already present in the sound card, I hope it doesn't create problems to have two, but theoretically I don't need them, if necessary I would only need one.

I have two free 16-bit ISA slots left, despite the number of cards, three 16-bit and one 8-bit, I could only recover one 8-bit ISA slot, changing the disk controller, with one that has included serial and parallel ports, for the moment I don't need it, in case in the future, if I add more cards, it would be an ideal solution, so as not to occupy all the free slots.

Going back to the CPU, I don't like this i486SX25, having a 486SX that goes like a 386DX, is not so pleasant in my opinion, maybe the difference between a 386DX and a 486DX is not much, but I saw from the bench facts on my 386DX40 (years ago), that we are quite close there, and therefore changing CPU, could be the ideal solution, if you don't want such a slow PC, I think the minimum would be to switch to DX2, and maybe increase the FSB from 25 to 33 or 40, in this way you would get a gain of about twice the Scores, and would make the use of Windows 95 less annoying.

I used the NSSI program to save a copy of the PC BIOS. Making a copy is very simple; you just enter the file name, and once saved, it's found among the files in the program folder. In addition to the MB BIOS, you can also save a copy of the video card BIOS, and I saved that as well. The card is an older ISA card with a WD chip. I can provide all the necessary information if needed.
In short, the motherboard is the PINE PT-2068.1, and the VGA is an ISA card with a WD chip.

I renamed the files as .txt, but they're actually .bin files.

I overclocked the PC i486SX25, changing the jumpers I set the FSB to 33 MHz, the PC seems to have no problem, and the temperature seems to be slightly higher, but nothing hot.
The reason why this is fundamental, about 30 years ago, I bought this motherboard with a 486DX2-80 to replace my 386DX33, a few days ago I managed to make it work after 30 years of inactivity, with a different CPU, Intel i486SX25, because the Dx2-80 was broken.
But with this 25 MHz SX, I don't see any substantial difference, between this and a 386DX, the performance is quite similar, and considering that the motherboard has no L2 cache, it also seems quite slow.
After overclocking, I ran various benches, to see what difference there was, let's say that it improved a little, but we are at the levels of a 486DLC40 more or less, for me it is not enough, I would say that even if it were an i486DX33, it is not the best example of 486, here is perhaps a DX2 or a DX4 (Overdrive), in short, 5V CPU, it would be an ideal replacement, but according to what I read in the manual, the motherboard was made for UMC U5 CPUs, which should be about 30-40% faster than those of other brands.
Without thinking too much, I start looking among the various CPUs I have, I know I have various 486DX2-66, in any case to differentiate this PC a bit, and since it does not have L2 cache, I could use a 486DX2-80 5V from IT'S, very similar to the Cyrix 486Dx2-80 that was originally mounted.
In this case I would have to reset the FSB jumpers, opening J6 and leaving J5 and J4 closed, in this way I would have the FSB at 40 MHz, it would remain to check the various dividers in the BIOS, so as not to run out of specification, the ISA slots and anything else.

For the moment I'm trying to save a copy of the BIOS, of the PINE PT-2068.1 motherboard, it's an AWARD v4.50G but the chip is a DIP28 with the initials 27XXXX, these chips should have a window for deletion, but I don't need to reprogram it, it would be useful to save the content, I tried with AWDFLASH and nothing, I don't know if you need a specific version of this program, maybe in some 486 motherboards this utility, was provided on a floppy .

As far as performance is concerned, it is to be carefully considered, the CPU change, not all 486 are the same, and the newer MB, had a myriad of jumpers and relative combinations, the easiest way would be to use a 486 Overdrive, but especially the DX4-100, in this motherboard it would be a bit limited, since there is no L2 cache, so the ideal would be a 486DX2, since the motherboard does not have the regulator for VCORE voltages below 5V, as we will see the best solution would be an interposer, and CPU iDX4 or AMD X5-133, because these last two CPUs have A double cache (16 KB).

Let's consider now, the 486DX2 from Intel and AMD, in the classic 66 MHz version, apparently it would be the same CPU, and doing some while the values may differ slightly, but at the time there was a price difference between the two, so it was chosen based on the available budget.

There was an alternative, which was the Cyrix 486DX2-66, this seems to be different from the other two, and overall it was slower, but it had an integrated cache L1, also of 8KB, of the "Write Back" type instead of the "Write Through" one, here for me this is something that you don't worry about, and in the absence of special jumpers, you can make the performance not optimal, important and not to ignore the BIOS setting.

So I would say that in the absence of jumper, for the type of cache of the 486, and of linear regulator for the VCORE, and L2 cache, it would be better to use a special interposer to install CPU, Intel or AMD or Cyrix, which depends on what is available, but from 100 MHz up, the one we prefer is fine.

Without interposer, instead I would say that the best solution is a 486DX2-66, to choose between Intel or AMD, I would avoid slower versions like the 50, but it could be used in those motherboards, without jumpers and with 50 MHz quartz, without having to look for a faster quartz.

Ultimately I will replace the current i486SX25@33, with an Intel i486DX2-66 CPU, I have to open another PC to replace the DX2-66 (which I will recover), with a CPU probably DX4 of about 3V.

I spent a lot of time ⏱, looking for the jumper settings for the SMT 486 PCI (Terminator) motherboard, for the record it is the only MB 486 PCI I have, the current jumper settings are for a 486DX2-66 5V, but from that I will switch quite quickly to an AMD 486DX4-100 or an Intel 486DX4-100, later I don't know if I will make a further update, going to replace the CPU, with the fastest one I have, an AMD X5-133 P75, which eventually you could also try at 160 MHz.

The card has a large number of jumpers, moreover many are attached to those next to them, it is difficult to understand the numbers, and therefore without looking at the manual, you can easily go wrong with something, which can create operating problems, or can in the worst case, prevent starting.

After taking note of the position of the jumpers, necessary for an AMD 486DX4-100 (normal), since there is also a later version with Write Back cache, I see that a couple of jumpers need to be removed, and moved someone, nothing complicated now, luckily the current settings are quite similar, otherwise I would have had to change the position to almost all the jumpers.

In total for the correct setting of a CPU, there are more than 10 jumpers, and each of them has at least two positions, four have seven or more (positions), two instead up to three, in short even with the manual you have to be very careful.

Luckily for the VCORE, the voltage selection is very simple, you can do it without looking at the manual, just read the screen prints on the motherboard.

I hope by this week or next, to try the card and get a video BIOS screen, and then complete the final configuration.

The recent successes with the 486, have encouraged me to review the ones I have, they are all except some of the DX2-66, this in light of the tests carried out with the iSX25, they are quite faster to detach a little a 386DX, which in turn is quite faster than a 286, in short between these the difference in speed can be seen.

I have among the 486 boards in repair a ZIDA Z4DVP VLB (on TRW ZIDA 4DVP), years ago I had added another 4 cache chips and a second TAG chip, the cache on board had gone from 128 KB to 256 KB, I had borrowed the chips from another 486 motherboard of the Chicony VLB, but later I don't remember if at the same time as the power supply failure, the PC did not start, or rather it starts but gets stuck in the second BIOS screen, to make it work, I had to deactivate the L2 cache.

At the time I did some tests, moving the chips, and returning to only 128 KB (removing the ones I had added), but despite doing various tests, it was not possible to get a working L2 cache.

But if I find the time ⏱ I would like to review this card, unfortunately I don't have the possibility to check the individual chips, and I don't have other new replacements, but I can do some tests, and above all I could check the general state of the welds and the tracks, in case I missed something.

If the end result is the L2 cache works, the performance in some benches, should improve a little, but being a first generation card, the only thing you can do to update is an Intel DX4 Overdrive CPU, the rest of the upgrades would be a VLB controller with cache, and SIMM 30 PINs of 4 MB each, unfortunately for cost reasons, it was not easy to find something like that in the past, today maybe by resorting to DIY, you could do something at least for the RAM, even if I saw relatively cheap interposers, That could expand the choice of CPU for the upgrade.

I restarted one of my 486DX2-66, this is a complete PC, typical of the mid-90s, a case 🏘 full Tower and display for MHz, a 5.25" floppy player and a 3.5" floppy player, I only added a 52X CD player and probably an ISA SB sound card.

The most interesting thing could be the MB, it is not a well-known brand for quality, in fact as we will see it is a very cheap card, exactly I do not know the brand and model, but it is an equivalent to the PCCHIPS M912, it has ISA and VLB slot, Socket 3 so possible upgrade with the latest 486 CPU, 1 MB VLB CL GD5428 video card, VLB controller, and a 30 PIN RAM bank, plus two 72 PIN, the installed CPU is a Cyrix brand IT'S 486DX2-66, the installed RAM is 20 (16+4) MB 72 PIN.

As for the BIOS, it is AMI type Windows, Relase 12/01/1994D, at the bottom you read this:

40-P101-001437-0010111-072594-GREEN-H

L2 cache speech, fake chips are present on the MB, the BIOS misleads the unsuspecting user, because it reads:

256KB Cache Memory

An upgrade should be performed, with real chips, but at the moment I don't have the possibility to perform this update, rather I can do shortly, the CPU change with another, I could for example put a DX2-80 (3-4V), but maybe I could use a DX4-100, I haven't decided which of the two to put.

Finally I did some bench, some strangely don't work, I think it depends on DOS, however in short these are the results:

Speedsys 4.70 : 26.65

Sysinfo 6.01: CPU Speed 78.1

3DBench2 : 38.1

Chrisb : 42.9 is 25.7 FPs

SI8 : CPU Speed 113.4

Main Processor Cyrix 486 67 MHz

One thing that is not good, on this MB PCCHIPS M912 (alias xxx), is that fake L2 cache, the chips have the writing Write Back, and are soldered directly on the card, without the use of special sockets, moreover the BIOS deceives the user by showing 256KB Cache at startup.

It can be remedied, if we have real chips available, replacing the fake ones, using hooves, for pressure insertion, the most difficult job is the removal, then once the holes are freed, and making sure that the orientation is right, you can proceed to weld the hooches.

All this work, serves to have a working L2 cache, and that can improve at least a little performance, but there are a couple of jobs to be done, in addition to the chips, you need to insert jumpers for the selection of the cache capacity, it is not essential in case we have the same capacity, but it is needed in case we have less than 256 KB, or if we have more, currently there is a welded bridge that sets the quantity at 256 KB, even if in reality the KB are 0.

Another job that is needed, is the reprogramming of the BIOS chip, so that the real amount of KB appears (L2 cache), or you could leave it like that, but with any size of the L2 cache, 256 KB always appears.

From the Speedsys 4.70 program, I saw a linear trend of the cache, after 8 KB, usually when it is present, it drops in correspondence with the installed quantity, generally on 486, we find 128 or 256 KB, perhaps in some of the first 486 we can find 64 KB, mind 512 KB are not very common, they are probably upgrades made later, and recently I saw upgrades to 1 MB!

I noticed that on the screen prints, there are beyond the usual multi 2X and 3X, which are those of the DX2 and DX4, there is a multi intermediate of 2.5X, this was used on the Pentium Overdrive for 486, but from what I remember there is no need to change multi from the motherboard, it should do it automatically, so it could be possible, for a DX4-100 select 2.5X 40 instead of 3X 33.

If it works it could be used for other 486 CPUs, probably some DX2, with multi 2.5X will work well, in case there are stability problems, with the multi 3X.

On the motherboard PC Chips (alias XXX) M912, as I had pointed out there is the multi 2.5X, the same multi is for example on my Chicony CH-471B, I saw it on the pdf manual (on TRW), but can it work? Here it seems that Intel had planned a 486DX3 CPU with this multi, these CPUs had frequencies of 63 and 83, which are then the same as those of the 486 Pentium Overdrive, but they were never sold.

But is it possible that it works? And which CPUs could work?

Since trying doesn't cost anything, you can give it a try, but maybe it's better not to try CPUs at random, the only one on which I think it's worth trying, is Intel DX4-100, with FSB 40 and multi 2.5X (instead of 33 and 3X), the frequency remains the same, but there should be a gain in the scores, I don't know how much maybe not much.

Currently on the Chicony I have an Intel 486DX2-66 with blue heatsink (without fan), I think it is a CPU of an IBM PC, originally there was an AMD 486DX2-66 with low voltage, which I keep in the collection because it broke, I will see to check the status of the 3.6V battery, I hope there are no leaks, however I will change the CPU, hoping that the ones I have work, in case the DX2-66 I recover it, to replace the 486 SX25 on the PINE PT-2068.1 motherboard.

Of course, before any upgrade, I will do the usual bench, I think that at the end as soon as possible, I will collect all these scores, and I will insert them in a spreadsheet, so that I know which boards go best with the same CPU.

Restarted the PC with Chicony CH-471B Rev.1.0 motherboard, the CPU is Intel 80486DX2-66 with blue heatsink, SiS 85C471 and 85C407 chipset, two VLB slots plus another seven 16-bit ISA, which are reduced to five, if we install two VLB boards.

The video card is a 1 MB VLB of the Trident, there is no controller card, because they are integrated in the MB, there is also an ISA sound card, SB Vibra16XV CT4170, the BIOS is AWARD v4.50G, with HIGHSCREEN customization visible before the main BIOS screen, the version is a VBS1.04 CH-471B of 10/11/94.

The 72 PIN RAMs currently installed are two of 8 MB each, in total there are 16 MB, I think I will add two more RAMs of 8 MB each, so as to have a total of 32 MB.

If you want the card could easily have even ️ more RAM, using those of 16 MB or even 32 or 64 MB, but for now having 32 MB is far beyond what is necessary, then if later I were to install Windows 95 or similar, maybe it might be necessary to have 64 MB or more.

Before changing the CPU, with an Intel DX4-100, I will do a series of benches, some I have already run, others instead I have to copy them to the disk, I think I use an IDE CF 40PIN adapter, put as a Slave together with the CD player, but now I notice that it is not connected, the 40-wire cable is missing, maybe I put that player on to fill the empty space, and to prevent dust from entering, it would therefore be to check if it works (?).

In short, no problem, I will use an IDE CF 40PIN adapter, connected directly to the MB, and I hope there are no detection or reading problems, sometimes it happens in fairly dated PCs.

I'm having some difficulty, to make the IDE CF 40PIN adapter work, if connected to some old PCs, in particular I'm currently with 4 different PCs (two of which are on the counter), to insert the files in the CF no problem, but connecting the adapter to a PC with or without HD, it fails, or the drive is not detected, or if it detects it, the PC does not start loading the OS.

Years ago, I had solved it by trying some controllers, and finding the ones working, the others I used, for normal HD, I could very well do without these CF memory cards, if I had an ISA USB card, or I saw recently, a memory card reader that connects to the parallel port, this would also be a good solution, but since I don't have anything like that, I will do as in the old days, Laplink cable and program.

In this way, I will replenish files (and programs), without wasting time ⏱ with floppy disks, which then to be honest, have such a limited capacity, that it complicates transfers a lot, when the files are large, you would have to break them and then put them together, in short, nothing impossible but not even simple enough.

In a motherboard, instead the controller is on board the board, it is one of the last S.3, a Chicony CH-471B of a Highscreen PC, currently an Intel 486DX2-66, but after the bench that CPU will go on the PINE PT-2068.1, and will be replaced with an Intel DX4-100 of about 3V (I don't remember the exact value), later I will see to add two more RAM 72 PIN of 8 MB each, to get a total of 32 MB.

At this point, the PC could be ready to install even Windows 95, although you could put Windows 98SE, but the 32 MB for me would be a minimum amount, more would be needed to make it work a little faster.

Anyway, I'll try to update some of the DX2-66 that I have in my collection, because I would like to differentiate the CPUs, so as to have some intermediate frequency, 100 for the DX4 and 80 for the DX2, I don't have DX2-50 CPUs because years ago, I didn't consider them at all, but with the experience of the 486SX25, I must say that at 33 MHz it works better, so maybe if I had one, I wouldn't use it I would keep it in the collection.

So I think that from five DX2-66, one I update to DX2-80, and another two to DX4-100, only two will remain, plus what I recover to put it on the PINE PT-2068.1, so in total three DX2-66 a DX2-80 and two DX4-100, to these then I would have to add the PC with X5-133 P75, which is a 486 with multi 4X.

At the moment I can't connect the PC with MB PCCHIPS M912, to the one with MB Chicony CH-471B, maybe I changed the serial ports years ago, with the ones I have available, it doesn't connect via cable and Laplink, while between PCCHIPS M912 and PINE PT-2068.1 it works, and I've already dechipped some directories.

So I performed the CPU update in the Chicony, the 486DX2-66 was removed, and replaced with an Intel DX4-100, this CPU is low voltage, it works quite well, for now I have not done long tests, but only a test with some bench, which actually show an increase in performance, although I have no references, so I guess they are normal scores for this CPU.

Before I spent a lot of time ⏱ straightening the PINs, some were visually crooked, others were slightly, however they are quite robust, so in short the work was successful, without any of them coming off, during the insertion test in socket 3, I had to correct some pins that were a little misaligned.

The most complex work was the change of position of the jumpers, there are several to move or close, initially I followed the TRW manual, unfortunately especially the Voltage selection part, it is not precise and not reliable, I followed the indications on:
https://www.elhvb.com/mboards/chicony/ch471b/ch-471b.html
which are simpler and more precise.

I chose this DX4-100, because it has double L1 cache compared to the 486 and DX2, now the cache has gone to 16 KB, I don't know what type it is, but from SpeedSYS 4.70 I saw that it is recognized, like the 256 KB L2 one, this means that the chips are real.

The Intel 486DX2-66 CPU with blue heatsink, will go to the PINE PT-2068.1 instead of the i486SX25@33, for now I haven't removed it, I'm a little suspicious, I don't know if after the replacement there may be problems, in short it has been stopped for 30 years, I just hope that the problem is not in the CPU Socket, unfortunately I can't say I have to try and hope that everything goes well.

I only ran a few benches, in this PC with MB Chicony CH-471B Rev 1.0, replacing the CPU was easy, configuring all the jumpers was a bit difficult 😩, and currently it works I would say well 😌, I still have to understand if the jumpers for the L1 cache are correct, I still don't ⚓️ understand if the cache is Write Through or Write Back ?, according to what I read from Jan's program, the L1 cache is Write Through, so I think the jumpers are fine like this.
Looking at the BIOS, in Chipset Features Setup, instead I find External Cache WB/WT, here I imagine it refers to L2, it is not very clear if they should both be of the same type, or they can be of one type for L1 and another for L2.
A further check concerns the serial ports, with this PC I can't connect the 486SX (PINE PT-2068.1), and not even the DX2-66 (PCCHIPS M912), I wanted to check the jumpers on the MB, because the controller is integrated, the one I would use would be COM2, which would use the large port (25 PIN), if it has the right connections, assuming I know what the PINOUT is on the motherboard, it should be simple to transfer the files, without having to open the PC, and connect the flying HW.
Originally I think this PC had an ISA LAN card, but for several years I used the CASE 🏘, with another motherboard, I think a dozen years ago, I fixed the motherboard, changing the CPU with another DX2-66, its was broken maybe they had played with the frequencies in overclock, but all in all after so many years, the motherboard is still ⚓️ in excellent condition, and also the 3.6V battery 🪫, seems to hold well 😌 the charge, shows no signs of leakage, that's why I haven't removed it yet, I'll check its condition from time to time, and in case I'll remove it.
It would remain to be seen if the CD player 📀 works, it would be a 50x, a bit too recent for this PC, but it must be reconnected via IDE 40PIN cable, at the moment I don't have any available, moreover I should be out of stock, especially if I continue to recover ❤️‍🩹 very old cards, I don't think the cables with 80 wires are usable, in memory I remember that there was some problem using them.
If I can transfer the files via Laplink cable, I can easily close this PC, reassembling the cover, I don't think I need a fan for the CPU, even if the heatsink is not very high, but to the touch the temperature, it seems acceptable to me, nothing that burns.

I modified the connections of the DB25 serial port, and it won't be a big surprise, but it works!!!

I would therefore say that at least in these old PCs, which are mostly ISA or VLB, it is better to have a DB25 Serial port and a DB9, respectively COM2 and COM1, in this way you can easily transfer files and folders, via Laplink serial cable and its DOS program, even if it is slow it does not matter, to do the same work via floppy, it could take as much or even more, with the risk of some error due to defective sectors, since the floppy disks are now very old, and with use and time ⏱, the surface tends ️ to deteriorate.

Obviously it could also be used with more modern PCs, but in that case it is not a good idea, except that you want to connect the modern PC to the old one, but the problem of the COM2 arises, which usually in the MB ATX, is in DB9 format (nothing to do with the Aston Martin!), so in that case you should try a DB9 F - DB25 M adapter, or if the MB has a predisposition for the COM2, and there is not in the holes of the shield, you can install a COM2 DB25, or if already present the DB9 should be changed.

The last unknown remains, which would be what kind of connection do you need? That is, a port with standard or non-standard connections ("this is the problem"), making sure you have a working serial mouse, you can connect it and see if it works, in case it works, you need a DB25 port with the same type of connection, so you can do a quick test, trying PIN2 and wire n.2, if wire 2 is connected to PIN2 it is standard, if it is connected to PIN6 it is non-standard.