Mystery motherboard testing 

Ben Eater's explanation of parity and error checking is one I found very easy to understand. 24:37 ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-MgkhrBSjhag.html

Third motherboard: The clock option you changed said SCLK/2 which means divided by 2. That's why the speed dropped rather than increased.

That first 386 motherboard is a special unicorn. AMI used to ship something like a "BIOS construction kit" to their customers (mainboard manufacturers). That construction kit used chipset-based "modules" to be enabled and comprising the final chipset setup on the shipped product. The detailed register-setting was for development purposes only and was not supposed to be enabled in shipped products. Obviously I cannot tell how it got enabled but I am sure, this was not in the regular retail version. I was working for a major taiwanese mainboard manufacturer back in those days and had access to the construction kit along with the documentation to go with the corresponding chipsets, which where hard to obtain back then. Fun times, but I would keep that mainboard in a special place.

For the 386DX-25 motherboard, at 26:57: Each row is a register (obviously), the 0s and 1s you can change. R is "read-only." The ONE 82C206 register exposed is for DMA timing. 82C301 04H is for CPU timing and NMI, 05H/06H are to do with ISA access timing The 82C302 is to do with memory mapping. () is what yours is set at, [] is default. Most of your settings are at default, but it is worth noting that one of the NON-defaults is parity: yours is turned on, confirming that your ram is arranged in x9 (or else it would be throwing errors constantly) 82C206 01H: First two bits: wait-states for accessing 82C206. 00 = 1, 01 = 2, 10 = 3, [(11) = 4]. Next two bits: wait-states for 16-bit DMA cycles. [(00) = 1], 01 = 2, 10 =3, 11 = 4. Next two bits: wait-states for 8-bit DMA cycles: [(00) = 1], 01 = 2, 10 =3, 11 = 4. Next bit: Enable early DMAMEMR#. [(0) = DISABLE], 1 = ENABLE Last bit: DMA clock. [(0) = SYSCLK/2], 1 = SYSCLK 82C301 04H: First three bits: read-only Next bit: Processor Clock Select. [(0) = OSC], 1 = SYSCLK/2 Next bit: Power Fail NMI. [(0) = DISABLE], 1 = ENABLE Next bit: Bus Ready Timeout NMI. [(0) = DISABLE], 1 = ENABLE Last two bits: read-only 82C301 05H: Each pair of bits represents a number of BCLK cycles, 00 = 0, 01 = 1, 10 = 2, 11 = 3 First two bits: 32-bit memory command delay: [(00)] Next two bits: 16-bit memory command delay: [(00)] Next two bits: 8-bit memory command delay:[(01)] Last two bits: I/O Cycle command delay: [(01)] 82C301 06H: First two bits: 32-bit transfer wait states. [00 = 3], (01) = 2, 10 = 1, 11 = 0 Next two bits: 16-bit transfer wait states. [00 = 3], (01) = 2, 10 = 1, 11 = 0 Next two bits: 8-bit transfer wait states. [00 = 5], (01) = 4, 10 = 3, 11 = 2 Last two bits: AT BUS clock source. [(00) = OSC/3], 01 = OSC/2, 10 = RESERVED, 11 = ATCLK 82C302 08H: First three bits: read-only Next bit: Write-Protect 256K RAM at 00FC0000H. [(0) = R/W], 1 = RO Next bit: Disable BIOS ROM just below 16MB. [(0) = ENABLE], 1 = DISABLE Next bit: /AF32 for addresses >= 16MB. [(0) = DISABLE], 1 = ENABLE Next bit: Memory limit after reset. [0 = 256K], (1) = USE 0A-0F below Last bit: Memory interleave. [0 = DISABLE], (1) = ENABLE 82C302 09H: First bit: C0000-CFFFF Read-only: [(0) = R/W], 1 = RO Next bit: D0000-DFFFF Read-only: [(0) = R/W], 1 = RO Next bit: E0000-EFFFF Read-only: [(0) = R/W], 1 = RO Next bit: F0000-FFFFF Read-only: [(0) = R/W], 1 = RO Next bit: C0000-CFFFF RAM in place of BIOS: [(0) = DISABLE], 1 = ENABLE Next bit: D0000-DFFFF RAM in place of BIOS: [(0) = DISABLE], 1 = ENABLE Next bit: E0000-EFFFF RAM in place of BIOS: [(0) = DISABLE], 1 = ENABLE Last bit: F0000-FFFFF RAM in place of BIOS: [0 = DISABLE], (1) = ENABLE 82C302 0AH through 0FH: Each bit represents 16K of address space. Turning a bit 'ON' disables system RAM in that space, allowing I/O devices to reside there. 0AH: 040000-05FFFF 0BH: 060000-07FFFF 0CH: 080000-09FFFF 0DH: 0A0000-0BFFFF 0EH: 0C0000-0DFFFF 0FH: 0E0000-0FFFFF 82C302 10H: First two bits: DRAM bank 0/1 type. 00 = DISABLE, [(01) = 256K], 02 = 1M, 03 = RESERVED Last six bits: DRAM bank 0/1 starting address bits A25-A20. [(000000)] 82C302 11H: First bit: DRAM bank 0/1 RAS precharge: 0 = 3xCLK2, [(1) = 5xCLK2] Next bit: DRAM bank 0/1 wait states: 0 or [(1)] Last six bits: read-only 82C302 12H: First two bits: DRAM bank 2/3 type. 00 = DISABLE, [(01) = 256K], 02 = 1M, 03 = RESERVED Last six bits: DRAM bank 2/3 starting address bits A25-A20. [(000000)] 82C302 13H: First bit: DRAM bank 2/3 RAS precharge: 0 = 3xCLK2, [(1) = 5xCLK2] Next bit: DRAM bank 2/3 wait states: 0 or [(1)] Last six bits: read-only 82C302 28H: First bit: Parity disable. (0) = Parity, [1 = No parity] Last 7 bits: read-only 82C302 2AH: First 7 bits: read-only Last bit: First 256K RAM. 0 = Disable, [(1) = Enable]

If you remember one of the original IBM ads for the PS/2: "How ya gonna do it? Ya gonna PS two it!" Brilliant. Just Brilliant.

The extra set of static RAM on your second 386 was the cache tag RAM. The tag RAM contains the address that's being cached. The other cache RAM contains the data being cached. Because the location being cached needs to be looked up *very* quickly tag RAM is faster and closer to the CPU to have the shortest PCB trace length.

You nailed it Adrian. I worked in corporate computing from the eighties. People used to say Wow when they got a new computer. The last significant change was with SSD's... they produced a noticeable speed bump.

30 years on and Adrian's enthusiasm is amazing.

That weird XT clone BIOS also misspelled version as "verson". Parity RAM was nearly universal in PCs until the Socket 7 era. The 80387 math co-processor was very expensive when it was introduced (over $1000) so that's why the first 386 motherboard you showed lets you use an 80287 instead.

42:55 back when Moore's Law ruled the earth. Now I update my PC about every 5-7 years and hope I can get 2x improvement. Went from a 4-core 3.2ghz Haswell to 6-core 4.1ghz Comet Lake last November.

Whenever you get back to the XTs... there's a new open source BIOS project for the PC and XTs called GLaBIOS that might be fun to play around with on some of those. They also have another project that provides support for XT real-time clock cards entirely in ROM. Great video, sorry the XTs went 0 for 2 though!

Third motherboard: It complained about CMOS and XCMOS errors. Entering the standard setup and saving cleared the CMOS warning but not the XCMOS warning. Fourth motherboard: The faster SRAM is for the "tag" (keeping track of what is in the cache), the slower is the actual data for the cache. This is a nearly universal cost optimization, the tag is needed much earlier in the process so it gets faster (more expensive) SRAM, TC55417P is a 16Kx4 so it has 16KB tag and 64kB cache size which matches cachechk.

16:29 "IBM COMPATIBLE BIOS VERSON 1.2" - yes, it says VERSON

That first 386 was a 20Mhz, not 33. The 33 is actually the double sigma that (allegedly) indicated 100% test of that part.

The first 386 M/B had a A80386DX-20, so a 20MHz not 33MHz CPU.

@21:22 The first commercially-available 386 PC was the Compaq Deskpro 386 in September of 1986 (according to Wikipedia). This was about 11 months before IBM released their first 386 machine, the PS/2 Model 80 which, as you mentioned, did have Micro Channel Architecture (MCA). But IBM wanted outrageous licensing fees in order for manufacturers to include MCA in their machines, and MCA wasn't available at launch time for the Deskpro 386 anyway. So the Deskpro 386 had only ISA slots. Later, a consortium of computer manufacturers, Intel and Microsoft (I think) started creating 32-bit expansion slot standards. If memory serves, EISA was the first, but wasn't very common. It wasn't until the advent of VLB and PCI that 32-bit expansion buses became commonplace. But, by that time, the 486 was the mainstream processor.

Toward the end, you summed up why I loved growing up in the late 80s & 90s. Each little upgrade was massively significant. It all dried up in the mid-2000s and just isn't exciting any more.

@33:00 The one highly integrated chip is the ubiquitious 82c206: That's the two DMA controllers, the two interrupt controllers, the timer chip and the real time clock integrated into one chip that is connected to the ISA bus. So basically a AT-on-a-chip minus the processor-to-ISA adaption (which is provided by Intels 82288). The '206 is used on a lot of 286, 386 and 486 motherboards, but there needs to be further support for the processor frontside bus (bringing it down to ISA), in this case, it's all PAL and discrete logic, so this board is likely older than integrated 386 chipsets.

For the XT motherboard, you could try flashing a eprom with the Landmark XT Diagnostic Rom. I used one when repairing my XT class NCR computer. As long as the board is not completely locked up, it should help figure out where the boot process is failing.

That A80386DX-20 is most places of the world is a 20MHz part, not a 33MHz part 🤪.

There is just something so soothing about seeing old hardware cared for and restored lol

Might have been mentioned, but XT class machines use a different port for boot-codes (0x60) than AT class machines (0x80 for IBM and a few others for Award/AMI/Compaq BIOSes). A POST card won't support the XT boot-code IO port, so unless the BIOS you got use one of the AT-ports it won't tell you more than what's shown on the status LEDs. Also, for EGA or VGA you have to select "other/no video" on the DIP switches. Otherwise it will not even acknowledge the BIOS extension on your graphics card.

I could watch videos like this everyday.

I think I might know the problem with the Sun Up XT clone. It still has "Verson 1.2" which despite the reassuring "System Already" is actually code for "All your base are belong to us!"

That variable capacitor next to the 8284 to draw the quartz is so ... quaint. I love it! :-)

Was keeping this video for my lazy Sunday afternoon watching but ... Old motherboards got me. Its so interesting. I had my hands on many back then and just kept working. Did not give any thought. Thanks for these detailed videos.

I 'freaking' love PC archaeology!

64K Cache on such an old looking 386DX board is petty cool! Maybe Adrian should note down on a post-it was tag-ram is. Some Motherboards have no documentation on what it needs but he could just write down a table for every amount of cache.

I use this kind of videos as a podcast while I'm cleaning my home

The last 386 motherboard you tested may have an external set up utility to set the chipset registers. I have a 386DX-20 motherboard where I have to run a utility called, setup386, to enable the shadow RAM. This made a HUGE improvement in performance on that board.

It's so nice to have those ISA/PCI slot test cards. They saved me so much time, when troubleshooting motherboards. Last time I had one with a reset switch that got stuck in the on position, after I changed CPU's, and I falsely assumed that the processor was responsible for the board dying. Turned out, everything was fine, which I realized when the card wasn't outputting clock cycles. A Swiss army knife of base diagnostics.

makes me want to get my XT clone out of storage and see if I can bring it back to life.

Enjoying your channel more and more Adrian!

Love the percussive maintenance on the XTIDE card. XD

I love your troubleshooting and repair videos! Love your channels Adrian!

It's better to have shorts in your tantalums than tantalums in your shorts.

I did not realize how fast time flew by with this video! great work as always :)😊

I dont know why but just those boards look like works of art.. thanks for the great video.

That 1st 386 DX25/33 motherboard was the same one I had in my first serious business PC I leased in 1987 for *NIX & DOS development. Previously had used TRS-80 model 1, Apple ][ and IBM PC clones. I had the extra ram card with another 64mb ram. Had 2x20MB hard drives. Power beast considering most people were still buying 8/10mhz PC and 16/20Mhz AT clones. Cost in 1987 was $10K aud, but earned me over 30k year for the 2 years I used it. I replaced it with a Wang Mini in 1990. Now retired I program asrduino and ARM SBCs that can outprocess any old AT/PCs

Hey Adrian, would be awesome if you did a vid of a complete build for one of those 386 boards, tricking it out with all kinds of good stuff in a big ol' case 😁👍

My favourite content this. Love seeing old random gear tinkered with. Top vid.

System Already! Love it!

I used to have a 386sx 25mhz. i ran Doom in a tiny pixelated window. bet that last 386DX PC could run it well!

Like taking a trip back in time to the good old days!

I loved the departure from all the Apple II videos, looking forward to when you have a Coleco Adam to work on.

Welcome back! Missed you on Saturday - glad you're back from travels!

I remember buying a i386DX running at 16Mhz and putting 4 MBytes in it back in 1988-9. People thought I was nuts because RAM cost a fortune at the time. I was able to run Desqview and run my editor, compiler, and test my programs all side-by-side. It was totally kick-ass.

Ohhh.. i saw a couple of utils I haven't seen for AGES - I used them too (like everybody else). That was a great trip down memory lane!

I remember upgrading from an 486 DX2 66Mhz to a Pentium III 800E when i was young. The first thing i got amazed at was that i could move running programs around in windows without the PC freezing up. Windows 95 on a 486 was pain, even with my 24MB of memory. Browsing websites was so slow. Every time i scrolled down, it had to slowly redraw the entire window. The Pentium III handled Windows 2000 without any effort and websites which were crammed with multimedia content ran just fine. The next thing i remembered testing out was video playback. I had windows 95B running on my 486 and it could barely play MPEG video at like 64x96p or something like that and heavy compression. The pentium III even played MPEG2 at crazy resolutions. I believe i even pushed it to 1280x720 with full 50hz because i live in PAL region. I believe i used this PC until i had the Core2Duo, because i didn't want a Pentium IV. From Pentium III to Core2Duo was another big step, but not as big as the previous one with the 486.

Trying to get the "System already!" board / BIOS up and running would have been really interesting

Thanks so much for just making great videos and content and not complaining about subscriber numbers and views numbers,. you are awesome !!!

15:11 I love how even in the early 80s Changlish problems still were a thing, System Already! 🤣🤣

Looking forward to more repair videos 👍🏻👊🏻

I may be wrong but that first motherboard may need jumper settings to work properly with that VGA card, Hercules and CGA should be ok as is though.

Newer Motherboards have just two custom chips i.e. Northbridge and Southbridge chips and just a few regulators for the various voltages required, now the PSU just supplies the 12V and an Intermediary unit generates the other voltages which makes for a smaller PSU.

I love it, "System Already!"

One major issue I face is storage of boards, they take up a LOT of room, especially if set up in "testbench" configuration!

Nice to now, when you where working in a computer store, I was working for a computer company “Sunny Tech” ;-) I assembled about a 100 computer a day.

I do like this kind of video.

Great Videos 😍

I'd just like to point out that the 1st 386 was actually a 386-20 not 33 (20:26) and the math co was a 387-25. They got the CPU to run at the math co speed of 25 MHz.

If I remember correctly, the BIOS in the original PC didn't support booting from hard disks. That was one of the "extensions" in the PC XT where a ROM in the hard disk controller could patch the INT 13h handler so code for floppies would now also work for the hard disk. For the AT the on board BIOS came pre-patched but with a limited list of hard disks it could handle. Things were a lot better by the time we got IDE.

Only on Adrian's Digital Basement can you be excited that something isn't an easy fix and requires a separate repair video.

"System Already!" I giggled at that.

Many 90s and 00s ATX cases will Baby-AT and full size AT motherboards by relocating the standoffs. I have a Baby-AT Socket 3 build in an ATX mid-tower with ATX power supply. All the expansion slots line up, and I sourced an ATX I/O shield for AT motherboards. I'm using an ATX power supply and I have soft power working with a latching relay circuit I assembled. The case's momentary power button toggles the relay on and off, grounding the PWR ON line from the PSU to turn the system on.

2:22 Unless there is a problem with that CF to IDE adapter it should draw power from pin 20 on its own so no need for that extra wire to bring the 5V to it, closing that jumper should be sufficient. I have the same adapters and on motherboards where pin 20 is physically present in the IDE connector I usually connect it to 5V with a wire underneath the motherboard for this reason, it makes powering them neater

Oh, man... the nostalgia is rich here!

At the bottom of the screen, the date range seems to be indicated implying this board might be Y2K compliant already back in '89? Would have liked to watch a post 2000 year being entered to check that.

That might not be parity memory, but ECC on the 386 board. That means several parity bits per byte. ECC can detect a double bit error and CORRECT a single bit error. BTW you noticed that it had an 80287 coprocessor, not an 80387? on the last board, does the cache have parity?

Would be cool to do a series on a 386DX PC build of you find a nice case.

About your comment regarding the 25x single-threaded CPU speed-up between 1981 & 1989, and how modern CPUs don't have that sort of gain from generation to generation --- I'm currently running a 2014 Xeon E5-2667 v2 (8-Cores/16-Threads, 3.3GHz / 4.0GHz Boost), which is listed as having a PassMark CPU Benchmark Single-Threaded Score = 2005. The 2014 Core i7-4790K had highest LGA1150 Single-Threaded Score = 2465. The Core i9-14900K is the highest new Intel CPU, with a Score = 4791. The new Apple M3 Max 16 Core has a Score = 4795. So, in the last 9 years, the Intel Single-Threaded Score has gone up less than 2x.

Since there was a big price difference I struggled between a 386 DX33 and a 486 back in the day after a few month I managed to find a reasonable priced 486 SX25 with 4 MB Ram and a 210 MB Harddisk, just bought a mobile 10TB HD, how times have changed!

I remember not being able to keep up with CPU peeds, even just reading about them. They got so fast so quickly. in the late 80s through to 2000. Having said that my mums old PC scores about 3000 in passmark, its a 4th gen intel, whereas my new 13th gen intel CPU gets 47000, but it has a LOT more cores.

26:47 The first ever PC we had in the family was a 386 from "Akhter Computers" and it had that BIOS too! I remember scratching my head for quite a while working out what those "Advanced" numbers did, as the machine came with almost no documentation for this part of it. Pretty sure that BIOS is identical!

42:45 Moore's Law baby. lol

I seem to recall having problems using a VGA card with my 5150 motherboard; with an original 8088 CPU the VGA card didn't work however with a NEC V20 CPU it worked fine. I came to the conclusion that the V20 has some additional instructions that the VGA BIOS needed to work.

IBM was big on support, so they probably supported users up until 1987 with all their PC models.

Triage is the name of the game. Find out what works, and kick the failures back into the bin for future work. Time is too precious to waste on weird PC clones.

Thank you for your videos, I like it much. 😊 If you have shorted tantal caps would you replace it by tantals again or just electrolyts?

18:06 “VERSON” 1.2 😂

I have the manual for another 386 board with that chipset which explains all those extended settings. I am in the process of preparing a nice scan for upload on theretroweb, but I can send you the raw scan if you like

I'd like to have 640k on-board on my 5150 but I don't want to do it with my existing original mainboard. It's apparently not too difficult to do; it involves a slight logic hack.

As for the speed improvements over the years currently - well, compare GPUs, not CPUs (at least not x86 ones), because that's where the progress is made.

I like video like this

well you know GUIs arrive and need that performance boost... I remember Windows arriving and our techs complaining that the CLI was so much faster ... hahaha

For scrubbing motherboards, do you just use dishsoap, or something special? Also do you rinse them with tap water, or do you use distilled water?

@ 21:13, when you zoom in, there is a socket labelled 80287; would there/could there actually be a 2nd math co-processor installed? Also, @ 23:54 there is labels on the board that say "80287 Installed" and "80387 Installed" by your thumb.

I have a 286 laptop from 1987 that has 1 MB RAM on the bottom(below the desktop PC ISA cards) in DIP chips. I wonder how much memory can I install in there by replacing those chips?

26:55 Funny how the first wait state option is "I/O" (letter O) and the other three are "I/0" (digit zero).

The XT clone motherboard, I saw the extra psu input and wondering if maybe the power is somehow a pass through for the drives such as the floppy drives. Maybe it was hooked up to a bay of floppies and that's why it needed the extra connection? Just a thought, I'm not familiar with those motherboards It was just the first think that came to mind.

The 386 looks like my first 386 motherboard

Cache isn't directly mapped to a specific memory page. What it caches can be anywhere most recently accessed in memory. So "the first 64K of memory" is not correct. The cache controller keeps track of which memory space specific pages of cache map to specific pages of memory. So that 64K of cache handles that 4MB of memory you had installed.

That 386 bios looks a lot like my Atari PC4 'NEAT' bios

So load up the Supersoft diag and run it on the IBM and Clone board. Its prob RAM that is the issue on the clone. there is sometime another Jumper that needs to be changed but supersoft will run with the bare minimum. Awesome Stuff. Keep it going. (We also missed your normal sunday vid on the main channel)

The floppy drive you tried with the XT clone board looked like a 1.2M drive, but did the XT support them, or just the standard 360K drives? It was interesting you mention with the last board that it's running the 80ns memory at 70ns, the last batch of SIMMs I soldered up had 60ns chips, but my SIMCHECK said they tested fine at 45ns! I really want the SIMM adapter for my RAMCHECK so I can actually confirm that speed with something more "modern". But otherwise it seems like the manufacturers were pretty conservative with the speed ratings of their memory chips. I can imagine it could probably have lead to dodgy vendors buying slower memory and putting it into faster systems like this.

15:41 "IBM COMPATIBLE BIOS VERSON 1.2" "VERSON" lol

Come on!!!! System Already!

Doesn't the original PC require negative 5 volts? Is the ATX/AT adaptor supplying that? I had one of those XT IDE cards and it failed. I've sort of fallen out of the hobby in general due to how common hardware failures have become. As far as the second 386, it's mostly TTL, either discrete or PAL as you point out. That one chip is sort of a PC-on-a-chip that combines the timer and DMA and other chips that you'd find running an XT or AT (AT in this case). The two 20NS chips are probably cache tags, good for 50 mhz. The rest are the actual cache chips, good for 28Mhz at zero wait states.

Is it bad that I'm quite happy the 5150 board failed so we might get a bigger repair video on it in future 😂

I wonder if that long ISA slot is an early attempt of expanding ISA bus to 32 bits.