Part 1: the motherboard

Immediately after opening the case and performing a visual inspection of the motherboard, I found what is most likely the reason of the "non-working" condition:

A couple of the legs on the main chip were crushed, and also a few others were bent and had been desoldered, probably due to some sort of impact. I straightened and resoldered some of them, but when I tried to fix the ones marked in red in the picture they just fell off, and worst of all it was at the very base, so it's not possible to solder some resistor legs there unless I expose more of what remains of the broken legs. Ouch, things didn't start out in a promising way.

So, I did some research and I found the datasheet for the UM82C088, as expected it's described as a "PC/XT Integration Chip". Additionally, I noticed that the label TACT80101 was silkscreened next to the chip, so doing a search on that I was able to find the datasheet for the Texas Instruments TACT80101, of which the UMC chip seems to be a clone of given the identical pinout descriptions. I was unable to source UM82C088 chips in the usual places (UTSource, eBay and so on), but I did find a guy from Germany that had new old stock of the original TI chip, for about $15 + shipping. I ordered one and waited.

Fast forward until I got the chip, soldered it and I got a POST! The motherboard seems to be this TURBO MAINBOARD MB-101 in TH99, which appears to be a clone of this one. It features a NEC V20 8MHz CPU (overclocked to 10MHz as you can tell by the 30MHz crystal) and is fully decked with a whooping 640KB of 150ns DRAM:

It has a JUKO ST v2.32 BIOS, pretty common in these types of motherboards, as well as sockets for an FPU and an option ROM. Also, I replaced the original power connector because it had some corrosion on the pins (probably from the I/O card battery, more on that later), I used a typical AT/XT power connector with flat pins which seems to fit better with the PSU P8/P9 connectors than the original one which had square pins.

With the motherboard now working I was able to start looking into the other components, more to come later!

Very cool.
Will be watching the thread.
Let's see what comes out of this XT guy.

TheMobRules wrote on 2023-02-21, 21:20:

so doing a search on that I was able to find the datasheet for the Texas Instruments TACT80101

Where exactly?
I can't find that datasheet...

Nice one on the repair job! I have a soft spot for XT class machines, so I'm eager to see what you'll make of this 😀

Nice! That front panel is *identical* to the one from the first PC I grew up with (other than the color of the turbo button, which was dark blue). That was a no-name Taiwanese XT as well. The rest of the case is different, but those things were all mix-and-match...

That's an interesting version of the Juko ST. Almost all of the ones I've seen have way fewer DRAM sockets.

Part 2: graphics card(s)

This is the graphics card that came installed in the computer:

It is a basic CGA/Monochrome card by Mitsubishi, some details about it are discussed in this thread, such as the function of the DIP switches and also VileR provided some info about the character set that resides in the ROM of the card.

At the time I did not have any other 8-bit graphics card (or 16-bit VGA that works on an XT-class machine), so I had to perform my initial tests with this card. For this purpose I built an RGBtoHDMI converter since I don't have an RGB monitor either. The card works fine in both CGA and Mono modes, all CheckIt video tests passed without problems.

However, for the final build I REALLY wanted to use a small CRT that would be a better fit for an XT instead of a flat panel with HDMI 😛. Now, ideally I would get an EGA card with an RGB monitor as I want 16 colors, but acquiring the monitor doesn't seem viable for now, due to cost/lack of availability around these parts. So I settled on this:

A Paradise 8-bit VGA card based on the PVGA1A-JK chip with 256KB of RAM. I know XT and VGA is not a match made in heaven, but this card offers good compatibility with EGA/CGA modes and is from around the same era when this PC was built (the Paradise card seems to be from around 1989 and all the parts of the XT clone have 1990-91 date codes, so it's a good fit in that sense). Most importantly, I can use it with a small 14'' AOC CRT I bought back in the early '00s that goes really well with the "cheap clone" look and feel of the machine. I will add some pics of the monitor when I get to the final part of the build.

Next up, I/O and then "mass" storage!

Grzyb wrote on 2023-02-22, 09:17:

TheMobRules wrote on 2023-02-21, 21:20:

so doing a search on that I was able to find the datasheet for the Texas Instruments TACT80101

Where exactly?
I can't find that datasheet...

Ugh, it seems I didn't save it! But I am 100% certain I had it at some point because I did a pin-by-pin comparison with the UMC chip before I ordered the replacement. Here's the datasheet for the UMC part, in case it's useful, as it's a clone of the TI chip:

TheMobRules wrote on 2023-02-22, 17:44:

I know XT and VGA is not a match made in heaven, but this card offers good compatibility with EGA/CGA modes and is from around the same era when this PC was built

I hear this quote a lot, but I don't tend to agree. While you definitely can't expect full motion video at 640x480, 16 colors from an XT, even a Turbo XT, I consider a VGA to be generally superior over a comparable EGA card: You lose register level compatibility to the CGA card as soon as you go to EGA. There is register level emulation provided by a lot of third-party EGA and VGA cards, and as far as I know, the CGA register level emulation on third-party EGA cards is not superior to the CGA register level emulation on third-party VGA cards. As VGA cards use the same basic memory access scheme as EGA cards, but EGA cards run at a system/memory clock of 14/16 MHz, whereas VGA cards run at 25/28 MHz, video memory access on VGA cards is considerably faster than on EGA cards. And that's likely notable in a lot of EGA games.

Furthermore, the VGA cards provide the 320x200 256color mode, which has a very simple programming model, and due its low resolution, only moderate bandwith requirements if you only animate some sprites. That mode is also available on the PS/2 model 30 MCGA graphics, which is comparable to a Turbo XT regarding CPU power. 256-color puzzle games like Brix worked perfectly on a PS/2 model 30, and are a good fit for your Turbo XT, too.

The only unique selling point of EGA over VGA I commonly hear which I cannot refute on a technical basis is that colors look more vibrant / beautiful / appealing on an IBM 5154 Enhanced Color Display compared to typical analog VGA monitors with VGA cards set to the default palette. I have no first-hand experience with the 5154.

Finally, some people consider the 8-pixel character box of the EGA an advantage, because it allows tiling the screen with the 25%, 50% and 75% dithering characters without vertical bars (as you would get it on VGA cards with their 9-pixel character box). The VGA can easily be switched to an "EGA compatiblity mode" using the EGA 8x14 character box instead of the VGA 9x16 character box, even without enabling a register-level EGA emulation mode. There is a BIOS call to choose between "CGA 8x8 character box", "EGA 8x14 character box" and "VGA 9x16 character box" for text modes. This call is persistent until reboot because the requested character box size is stored in the BIOS data area RAM, and not cleared by video mode switches.

So, please just enjoy your VGA card, which is, in my oppinion, an overall upgrade to EGA even in low-end computers!

TheMobRules wrote on 2023-02-22, 17:57:

Here's the datasheet for the UMC part, in case it's useful, as it's a clone of the TI chip:

Thanks...
Looks like those JUKO ST boards can be based on two different chipsets:
- TD3300A + discrete chips: 8253, 8259, 8237
- UM82C088 = TACT80101, with everything integrated

The first variant is well-known to support 1024 KB of RAM: 640 KB conventional + 384 KB for RAM drive/print spool/EMS.

The second... probably not, there's only room for 640 KB on that board, and there's nothing about accessing any extra RAM in the datasheet.

Part 3: I/O card and floppy drive

The I/O card that came with the machine has a floppy controller (double density only of course), a battery-backed real-time clock, 1 parallel port, 2 serial ports (although only the chip for the first port is populated) and 1 game port:

It has no branding other than a "T-ZONE" label, whatever that is. I was unable to find anything similar in TH99, but leaving the jumpers as they were set when I got it (shown in the picture) everything seems to work properly.

Originally the card came with a rechargeable barrel battery that was already leaking, so I had to remove it, clean the corrosion and scrape+tin a few traces which I later covered with solder mask. I then found a CR2032 battery holder which happens to fit EXACTLY with the shape that was silkscreened on the board, remove the diode used to charge the battery and installed it in another position so that it now prevents the coin cell from being charged. Looking at the silkscreen, it appears both rechargeable and non-rechargeable battery configurations are supported. I also added a current-limiting resistor in case the diode fails, as recommended in some CR2032 datasheets.

The RTC works fine and keeps time very well with the coin cell, when it comes to syncing the MS-DOS data/time all the standard RTC programs for 8167 clock chips seem to work with it. Personally I'm using the Strickland RTC programs.

The floppy drive is an EPSON 5.25'' double density drive. Nothing terribly interesting, other than it works and seems really clean, with very little use:

Next: hard drive!

Part 4: hard drive

When I got this computer, the guy who sold it to me told me it came with a Seagate MFM drive (an ST-225 if I recall correctly) but that he couldn't "get it to work". I was willing to give it a try, but he wanted an extra $60 to include the drive in the deal. I declined since that's about what I paid for the entire PC, and the likelihood of the drive being dead was high. He did however part with the MFM controller, a Seagate ST11M:

So, having a controller but no drive when I started to put the build together, I was looking into solutions based on the XT-IDE Universal BIOS. I don't use CF cards for other purposes than testing since I like spinning drives in my builds, and I had some low capacity IDE drives that could fit this PC.

However, a few months ago this popped up on an auction site:

It was about $25 shipped, listed as "untested" which usually means "completely dead" (especially for MFM drives) but the person selling it had no other computer products on sale and did not want a small fortune for the drive, so they seemed genuine about the "untested" status. Also, this drive supports auto-park unlike the ST-225, so the user forgetting to park the heads before the drive was retired was not a concern. So I took the plunge.

When it arrived I feared the worst: the drive was poorly packaged and was loose inside the box, with only a layer of bubblewrap protecting it. To my amazement, when I gave it a try it started without missing a beat! I then ran a LLF with the ST11M controller and it formatted without bad sectors!! I'm really happy about this, this guy is a true survivor! I've been wanting to play with these older drives for a while now, so this is great and it fits like a glove in this build.

NOTE: I had to run the LLF utility in non-turbo mode (4.77MHz), otherwise at 10MHz the controller would freak out and the whole system became unstable, even after a reset: memory count detecting less RAM than usual, random lock-ups, etc. Strange. It only seems to happen when the LLF utility is run though, since going back to full speed after formatting everything works properly. Has anyone experienced something like this before? I am able to replicate this consistently: just run the utility using debug at 10MHz (even without the drive connected) and after exiting everything is messed up unless I shutdown.

Now I have all the parts I need, but before assembling everything I have to tackle the PSU!

Part 5: power supply refurb

The power supply is a standard XT-style unit, claiming an output capacity of 15V on +5V, 6A on +12V and 0.5V on both negative rails:

My initial inspection showed that it was very clean inside, and in good working condition. Voltages seemed stable under load and no strange noises or explosions occurred. I was pleased to notice that the manufacturer used 18AWG wires.

However, given that it used some garbage-tier PCE-TUR caps, I decided a recap with good quality Japanese parts would be a good idea. I also made some improvements to make it as reliable as possible. Here's a list of the updates:

• Replaced all electrolytic caps (I used an assortment of Rubycon, Nichicon and Chemi-con for the replacements)
• Replaced the 16A 40V Schottky rectifier in the +5V line with a 30A 45V part
• Replaced the 2 FR diodes in a bracket used for +12V rectification with a fast recovery diode array in TO-220 package
• Replaced the generic X/Y caps in the input filter with safety-rated parts

With these updates I'm confident the PSU can achieve its rated specs, even though it will be limited to what the main transformer can do. The rated 150W seems easily achievable, even with the original parts.

Here's a couple of pics of the innards after the updates (unfortunately I forgot to take pictures of the PSU in its "before" state):

The next part will be the last one, with the final build and maybe a benchmark or two!

Very nice! You seem to have a lot of knowledge about these parts and electric engineering.

Thanks! Even though I did some basic electronics stuff early on in college my whole career has always been on the software engineering side. So regarding electronics I've learned a few things on my own as a hobbyist. My brother is an EE though, so he's a good source of info on the more complex stuff.

The good thing about these XT/AT power supplies you find in clone PCs is that they all pretty much share the same topology, based on the TL494 control IC. So once you get an understanding of how they work, the improvements you can make are usually the same, and you end up being limited by the custom components such as the transformer that are not easily replaceable with off-the-shelf parts.

mkarcher wrote on 2023-02-22, 18:08:

Finally, some people consider the 8-pixel character box of the EGA an advantage, because it allows tiling the screen with the 25%, 50% and 75% dithering characters without vertical bars (as you would get it on VGA cards with their 9-pixel character box). The VGA can easily be switched to an "EGA compatiblity mode" using the EGA 8x14 character box instead of the VGA 9x16 character box, even without enabling a register-level EGA emulation mode. There is a BIOS call to choose between "CGA 8x8 character box", "EGA 8x14 character box" and "VGA 9x16 character box" for text modes. This call is persistent until reboot because the requested character box size is stored in the BIOS data area RAM, and not cleared by video mode switches.

Can you provide more information on this BIOS call please? I greatly prefer the EGA character box. I found on one of my setups that I can force it (and by extension a 800px/line horizontal total scan out, as opposed to VGA's 900) by using the DOS MODE command to set a 43 line mode, and it persists even after reverting to standard 25 lines.

maxtherabbit wrote on 2023-03-05, 23:11:

Can you provide more information on this BIOS call please? I greatly prefer the EGA character box. I found on one of my setups that I can force it (and by extension a 800px/line horizontal total scan out, as opposed to VGA's 900) by using the DOS MODE command to set a 43 line mode, and it persists even after reverting to standard 25 lines.

http://www.techhelpmanual.com/170-int_10h_12h … text_modes.html to configure 350 scan eines.

This call is generally used to select 80x43, because you configure that mode by setting the EGA character box 80x25 mode and then requesting a switch to an 8x8 font. You report the 80x43 trick only works in one card, though. Pehaps I'm mistaken that this setting persists?

Part 6: sound card and benchmarks

Originally I wasn't planning to add a sound card to this build, considering that most games supporting Adlib/SB or MIDI I'm interested in happen to run quite poorly on an XT, even on the 10MHz V20. Since this machine will be dedicated pretty much entirely to 80's stuff, here's an overview of the games I will be installing:

• Sierra's AGI catalog (only played a couple of these early ones, so I want to start here and move on chronologically through all the "Quest" series)
• Early Lucasfilm adventure games (Maniac Mansion, Zak McKracken and Indiana Jones and the Last Crusade)
• RPGs such as M&M 1 and 2 or early SSI Gold Box games like Pool of Radiance (never played any of these back then, so I also want to start from the beginning)
• A couple of classic racing sims such as Test Drive 1 (most of them are really slow though)
• A few arcade/platformers (Thexder comes to mind)
• Other miscellaneous stuff I'm familiar with (Pirates!, SimCity, etc.)

This will be enough to fill the MFM drive I think 😃 .

Anyway, while browsing around I found the excellent AGI-SB tool that can patch Sierra's AGI games with Adlib/SB support to give them a much needed upgrade from the grating PC Speaker "music". I have to say, this is a huge improvement! I know one of the new ISA Tandy sound cards would be ideal, but for now this is close enough. So, having Adlib support on Sierra's games plus a few others that already support it like Indy, I can justify adding an Adlib card to the build! Well, kind of:

Yes, it's a clone but it sounds great and was sitting there unused, also I kind of need my kidneys and my house so I won't sell them to buy a real one 😛.

Now, with all the components I can finally complete the build! I will update the first post with some more pics and the final configuration. To wrap things up, here are benchmarks I ran using CheckIt and Landmark. Nothing too surprising as these XT clones are all pretty similar in performance:

CheckIt v3.0

2.37 times IBM PC-XT speed seems correct, with turbo off (4.77MHz) I get 1.13 times PC-XT which is also expected due to the V20 being slightly faster than the 8088 at the same clock speed

This ST-251 is a speed demon! 😛

Landmark System Speed Test Version 6.00

Like a 6MHz AT when running at full speed and like a 3MHz AT with turbo off.

Seems very in line with my XT results. Very nice!