Patron Edmond pointed out the Intel cache controller can only talk to 32k of SRAM. I took a look at the datsheet and I totally agree. It says one bank of 8k x 8 chips or two banks of 4k x 4 chips. Either way that's a total of 32k. It could be that the 8k x 8 chips in there are only half being used -- which means half is wasted!
Also, the 15th MB isn’t being cached because of the “ISA Memory Hole”, which is used for memory mapped ISA devices. Later caching boards sometimes have the option to enable caching this megabyte, when the user knows there are no memory mapped ISA cards.
@@EvilTurkeySlicesthe manual seems to indicate that. Says that 32 is direct mapped, but 64 is some mode I don’t recognize. Probably some long lost driver to make it work
It's midnight in the UK and I'm watching Adrian's Digital Basement. I cannot fault my life choices. I had so many chances to collect these old boards, but I always passed them on to people without a computer. Love your channel Adrian.
According to Wikipedia the 80287 and 80287XL both work with the 80386 and were the only options available when the 386 first came out; the 80387 didn't come until two years after the 386 was released.
That's right -- I've seen 386 boards like that, but I think the 386SX (which is really a glorified 286 with a 386 stuffed inside) uses some kind of different FPU package. (Rarely bought since the performance of a SX was so substandard anyway.) Somewhere around the lab I have one of these FPUs (Which I purchased off ebay years ago.) I really should make a video about that!
@@adriansdigitalbasement2 Its the XT trick again: 16bit bus with 32bit cpu (xt was 16bit cpu with 8bit bus). Its like a faster 286 or a slower 386. But being actual 386 helped in things that required the protected memory like windows 3.1 or early linux, etc.
Very special boards you have there 🤩. For Speedsys make sure to use ver. 4.78. On some chipsets speedsys really like to freeze at the memory detection. You can start with „speedsys dspdr“. This is disabling the memory type detection and u can use the program then just normally. Thanks for your videos. Cheers, Peter
Those Micronics 386 DX mobos were awesome late 80's goodness! I put an AMD DX 40 and loaded it up with like 16 Megs by the mid 90s. Used it as a Windows 3.11 machine. Totally useful.
Hi Adrian, great video, brings back some memories. Those SIPP pins you are looking for are available from several manufacturers - do a search for "leadframe pins". They are still used quite often in electronics where configurable modules get plugged/soldered into the main board. Power supply modules, hybrid circuits, and double-sided modules on single-sided boards are very common applications. Anyhow, I love the channel and the content. Cheers! p.s.: Leadframe pins are also used on small modules like the ESP-32 and Pi Pico, instead of Molex headers, to prevent them from spreading (obliterating!) the contacts in a standard socket.
@@AureliusR They are a little different, with round pins instead of square, and slightly smaller in diameter. Breadboards *used* to be made for components and small jumper wires, so the old pin-headers would damage the sockets after a while. I think that modern breadboards are made to accommodate pin headers, so it is not such a problem. However, **chip sockets** are made for very thin pins, so pin-headers *will* damage them - or at least spread the contacts out enough so that a DIP-chip won't fit tightly any more. That's why the leadframe pins work better: they are close to the thickness of a standard DIP chip component and they don't spread out the pins. Apologies for the verbose comment, that's how I roll. Cheers!
We had a 286 with a Phoenix BIOS, and I think it used a CTRL-ALT-ESC combination to access the built-in setup program. That combination would work at any time, even after the system was booted.
That cyrix chip may not actually be bad, the 486dlc was not intended as a direct drop in replacement for just any 386. It was designed for ,often later, cyrix aware or enabled motherboards. Though sometimes you may luck out and it works anyway, but often with cache and other stability issues. For a direct replacement for the 386, they had 486DRu and 486DRx series chips.
I'd like to add that these Cyrix/TI DLC CPUs are *very* hot and power-hungry. I had a board which refused to start with them until I recapped it with solid-states (original caps were regular electrolytics).
@@Arti9m though similarly labelled, the ti processors were even different iirc. I think it wasn't until the true 486s ang later where the cyrix, to, and IBM were pretty much the same, just different fabs and labelling. Could be mistaken.
Entered the PC world around 92 (after the Amiga) with the Pentium, so it's fun to see all these older boards that I never got to experience. Everything looks extra chunky! :)
Yeah, this is the era when I was learning and getting into pc's as a young kid. Hand me down 286's, 386's, and my favorite processor of them all, the 486. Learning and playing with all the old stuff helped me pass my A+ certification with flying colors when I got to high school years later.
DOS 6.22 also has the handy F8 option during startup to selectively choose which lines to run in the autoexec and config sys. This way you can choose whether to run EMM386 and/or HIMEM and so on without having to edit the startup files in some situations.
@@root42 True, that's also a handy option for commonly used configurations, especially if you frequently use 2 or more start up scenarios. The advantage of the F8 option, though, is to be able to toggle numerous permutations of existing options without further modifications to the configuration. It's mostly for those one-off scenarios and/or rarely used scenarios, and prevents expanding the config file too much. In a nutshell, it all comes down to whether you use alternate startup options frequently enough.
The best way to deal with SIPP memory these days that I found is this: get some SIMM sockets. The pins from those fit perfectly into SIPP slots, and you can install regular SIMM modules in them afterwards. This way you don't have to modify either the motherboard or the memory modules. Or if you're so inclined you can desolder the SIPP headers and solder instead SIMM sockets.
We used those Micronics boards in our frash test systems, they were fast 386s. They would be loaded with the RAM card, a DSP5600 card, WD graphics, Inmos Transputer board, 3c509 later on and our proprietary Comms card. These ate power supplies Some versions of the Dallas chips can take an external battery
Don't worry about the length, I usually watch your videos at 1.5x, sometimes 2.0x but I never skip one :) This is a promising video, looking forward to watch it later! Edit: I watched it and it was fun. At 41:40, that motherboard looks pretty, I feel like biting into, looks like chocolate :) You have much more luck with motherboards than I have, I was looking for a neat 486 motherboard and had to buy 2 broken ones before I got my hand on a cool working one. I thought I was onto something with one of the boards but I had no luck, switched some components, tested the BIOS in a different motherboard (it was okay) and nothing helped to fix that one, I think maybe the chipset was fried or something.
Also, one of the things I've had immensely good results from when cleaning boards is hot water, dish soap, and baking soda. The baking soda helps clean mechanically, and the hot soapy water cleans chemically. Combine the two and it's how I get some of my immensely disgusting boards looking a million times better. I had a dell 386 machine complete with leaky battery pack that was left in a rotting shed outdoors for 20 years, and it come back to life with nothing but cleaning. It was in such bad shape that the base was almost rusted entirely through.
My little computer shop used to sell DTK computers back in the mid 90s. They were my "regular" line. Touch computers with Asus MB were my premium, then I had my own line of custom computers that used Asus and Gigabyte MBs. Those were simpler times...
Hi Adrian. If you ever get your hands on an EISA (32 Bit) motherboard, that would be very exciting!! I had one back in the day. They seem to be very rare nowadays!
If you have the actual EISA boards to play with. It can take normal ISA IIRC but that would be boring. Oh what should i do with my 386 class single Vesa Local Bus slot motherboard... And you thought EISA was rare 🙂
I mentioned your channel in the comments on JayzTweCents, His last video he showed a C64 and I said your channel is great for trouble shooting etc.. cheers.
Awesome bunch of vintage motherboards. It always amazed me how they could make those multi-layer boards with such hair-like traces and tiny vias, and get everything to line up when the layers were put together. Wizzadry at its finest 👍
I do remember using the Harris 20 Mhz 286 and having to find ISA cards that could work with the bus sped up. Then the computer store I worked in found some (really small) 286 motherboards that had jumpers to set the clock divider for the ISA bus. Well into the 486 era there was still a small market for 286 systems.
Really enjoyed this video, just feels like sitting next to you watching you slowly go through your process, and picking up tips for my own projects along the way
I can imagine the size of the computer for that huge board. WOW. They were so big, but without impressed power compared today. Now days motherboards have shrunken extremely small. Pretty crazy. And lots more options. My first PC was not as big as those Motherboards. I wish I still had that one, and my Commodore 64. I do miss programming with my old commodore 64. I really enjoyed watching you test those Motherboards!
The coprocessor socket of the 386 board is very likely a Weitek-enabled socket. The 387 from Intel and compatible (Cyrix, ULSI...) have two pin rows/squares, the Weitek has a third row/square so a little wider socket. To support Weitek, you need the wider socket (3 rows instead of 2) and BIOS-Support (BIOS will show "Weitek Coprocessor" Option somewhere in advanced settings). The 287 on the 286 Board seems to be a little bit off to me. The 287 is actually clocked at Crystal/3 while the CPU is at Crystal/2 (meaning the 287 runs at 2/3 of the MHz of the CPU). Never heard of half.
The IBM 5162 is a 286 which runs with zero wait states... however, IBM crippled the BIOS so that it couldn't be run at a higher clock so that it doesn't make the 5170 look bad. If there is a bad cache tag chip, then that could make some parts of memory uncacheable.
It is pretty normal for 286 PCs not to have a builtin setup. You had to use an external dos app. Often setup.exe Sometimes you need to have the particular setup.exe for some boards due to extra features.
@@herrbonk3635 Unix was on the way out? So you didn't go to a university in the mid-80s? Or work at a large corporation with a huge database? Or work at a bank? Or work at a store with a large inventory spread out over several states or countries?
@@scottlarson1548 I didn't say it was "on the way out". But it seemed sooo old fashion to me, that I always wondered *_why_* it wasn't :) So totally backwards in user interaction (1950s style), but still presented as "the future". I didn't criticise that a large firm had a central data base, of course, but you didn't need Unix for that. Lots of other systems had networking. (In the early 80s, my idea of what a _personal_ computer should be was very much along the lines of today's mobiles. I was inspired by instrumentation and advanced calculators from HP, Tektronix, and all the japanese. Early GUI I've seen too, from Xerox Alto to similar attempts on home computers.)
@@herrbonk3635 So you were completely out of touch with the computer industry. People didn't need expensive GUIs to sell carpet or car parts or track their inventory or payroll. Businesses could buy a DEC terminal for a fraction of the price of a PC and it didn't need maintenance and your employees didn't play games on it all day. What you're describing wasn't even an option.
@@scottlarson1548 Yes, at least that part of the industry. I worked as an embedded systems programmer and electronics engineer in my youth (before and during university). So pretty far from that administrative world.
SIPP: Single Inline Pin Package. Also, can we just marvel at how nice the first two boards look with their respective parts color schemes? I thought the bright red resistor packs looked nice on the first Micronics board until I saw the second one with its 100% color coordinated capacitors.
@@adriansdigitalbasement2 it _had_ to be intentional. They're too coordinated to have just been colored like that by chance. I agree those would have been very high end boards, so making them look nice and coordinated was probably the start of the "gamerificiation" of modern motherboards. Glad they all still worked, too!
And incidentally almost nothing used it, so they were of very low need while costing as much as another cpu. Unless you had some specific niche that required it, like autocad. Only when it became included inside the cpu, were developers starting to make use of them.
The board with the serial and parallel ports was aimed at the upper end of the market where a scsi card would be installed that would have the floppy interface.... I remember these boards....
I have a bunch of 386s. Normally, sx would not have cache and dx would have. From my memory of those days it was just more expensive, and i remember at some point the shop i worked at would ask customers if they want it with half cache or full cache. That 287 math co was expensive and the only time i remember a client bought one was to run Autocad with a 286. Ah that DTK brought memories, i still have the computer desktop case (with the original DTK packaging box!) where that would go in. I kept the DTK case over the decades after various motherboard upgrades. That 486 with 16mb is luxurious, win95/win98 capable. Indeed post codes are different, there was documentation about them, i bet online you can still find it.
Third board is commonly found in Gateway 2000 systems before they started the PS/2 keyboard and mouse port shenanigans. That DTK 386 motherboard would be a great candidate to stick in that huge DTK 386 AT chassis you did the rust conversion on years ago!
@@lucasRem-ku6eb sadly, there's a lot of people with your trash it attitude. The vintage computing community is large and growing. Old computers and components are no longer in production so every time old systems get trashed is a lost learning opportunity for vintage computer enthusiasts.
286's had setup on floppies early on. I think I recall the IBM PC/AT needing setup floppies, and I definitely remember the PS/1 and PS/2 systems using setup disks. Without them, you were just stuck with the error. Some 386's did that for a bit too, but 'not saddling the customer with setup disks' cut the cost to support users too since they lost them all the time. On the third board, the 286-10 that you caused me to add this second comment for, the reason it might not have included the floppy/IDE connector is that back then, there was a lot of MFM/RLL stuff out there, 'special controllers' galore for odd freaky "the size of your computer" external hard disks/tape drives, SCSI, even ESDI! Ah, ESDI...that awkward step between 'they used modems to transfer data from disk to PC' and "Actual digital signals" with IDE. And I exaggerate slightly about the 'modems for data transfer' but the drives used to be stupid and BRUTALLY slow. Before SCSI was worth anything, ESDI was a good higher-performance Back then it was amazing, of course! Not loading from floppies was way faster...and floppies were way faster (and better!) than punch-cards, which were themselves better than binary-toggle per-instruction boot-up processes! Set switches, flip toggle. Set switches, enter the third byte. Repeat forever... For "the cache being missing", many of them had jumpers to set 16/32/64k, and some did run 4 chip cache modes, though I don't recall 'what density they actually were', I just know that I recall 4/4 cache/empty setups being valid, so maybe there's a missing/incorrect jumpers as you noted. I recall bad cache being either 'the end of all the cache' or 'the system didn't boot', not that it worked only half-way. Not caching 13 or 15 is ODD, though. "Not caching over 15" might make sense in a '16Mb and up' thing bug..? Strange that they don't have a cache size jumper setting, however...well, I dunno. The video isn't done yet and you're silly-tired, so we'll see what happens before the end! :) Thankfully, I started with the IBM PC, so I only know the punch-cards and 'toggle-panels' from historic videos and retro people using it later on. I did the full evolution from 'single-density floppies' to 'Woohoo!' to 'ZIP disks' to "USB thumb drives"...it's been fun to see the difference across the eras, that's for sure.
The kind of socket that the SIPPs go in is still easily available new (basically, a single row precision socket). If they are factory new and not soldered/clipped, THEIR pins fit themselves quite well. Basically, precision sockets are like LEGO bricks (they also make fantastic stackable connectors for lab modules, if you manage to solder wires to the side of the pins without melting the plastic too much!). So you can basically take that piece of socket (or pieces of precision DIP sockets even), and just solder-bridge or wire the socket ends of them to the pads on a SIMM. Presto, it can haz pins! One thing to be careful about: Best tin, solder-wick and retin anything that is gold plated (socket or pad) before soldering to it, too much gold getting dissolved into the solder can make some types of solder extra brittle!
I'd look for datasheets for the cache controller (jumper settings) and verify them against the jumper pins. Experienced a few mobos where the manual have been completely wrong (if you even were to find them for this particular board) :)
Indeed the datasheet for the controller indicates it only supports 32k - so the 64k on there is being partially wasted. Someone clearly had money to burn :-)
First PC I built was a clone XT board with a new crystal to match my NEC V20 CPU, in a 5150 case and with a green screen, all bits bought from auctions. First HDD was ST-225. Then 286, 386SX, many 486s, after which it became my work rather than a hobby. Fun times.
The extra pins are for a different type of FPU, specifically the WEITEK FPU which connects to the CPU differently from a standard 387 or compatible. In fact there are boards that have instead 2 FPU sockets. Due to how they interface differently with the CPU it should be possible to actually have both a Weitek FPU AND and 80387 FPU on the system at the same time and even use them simultaneously though I never got nobody to confirm this (although electrically this is possible) and worse is that if you want to test simultaneous use of the two FPUs you'll need to make the software yourself as AFAIK there is no software that will use both simultaneously. Either or yes, and, not really which is a pitty. For example, CYRIX went to crap because intel CPUs could basically do integer and floating point math simultaneously (similar to a dual core) which the CYRIX ones could not (AMD also could do that by the way) and QUAKE actually took advantage of this feature which meant math wise QUAKE would run twice as fast on Intel/AMD VS running on Cyrix. Now, imagine we could have it use all the math processing devices hehe :) So Integer math on the main CPU, then 2x floating point on both FPUs.
If my memory serves me right those Micronics boards were considered very high quality boards in their day so I am not surprised it still works. Even still, battery corrosion kills most boards whether they are good quality or not.
Now I wish I would have keep all my old computers. My most odd was the Intel PR440FX board with dual Pentium Pro's running Windows NT. That was a rock solid machine at the time. Had SCSI HDDs and DAT tape drive backup, and 256MB of DRAM at the time and dual 200MHz Pentium Pro's. I then did an upgrade and got the Pentium II Overdrive 333MHz CPUs. Ah those were the days back then!
Hi Adrian, great video, I also have a DTK 2500, I could never get XTIDE to run, it just hangs. I originally got it to harvest the DRAM as these are also used on many Amiga 2000 RAM cards, and was way cheaper than buying the DRAM chip on their own, but decided I liked the board. There is a video on my channel where I installed the board into a full tower case (Retro 386 SX 25 PC Build) I later released it is a 386DX, but no one noticed 🙂
Well, they certainly were surprising to see them all working for the most part, even the bent one, I thought for sure that Bender there would have needed some form of repairs or something to get it to wrk, but, nope, it's a runner, unlike modern PCBs that fail if you so much as breathe near them... :P
A lot of earlier 386 boards did make use of 80287 coprocessors. The 80386 came out two(?) years or so before the 80387 was ever made, so people who did CAD and stuff needed SOMETHING. Something like the 80287XL's which were '387sx chips in a 287 socket' also existed, I think? I don't remember when those came out, but the 287XL's were a lot faster than the original 287's. Any copro beat none for the right workloads, of course. Like POV-Ray! Oh, the memories... CAD too, I'm sure. Lower-cost motherboards "later in the early-years of the 386 era" cheerily used 287's since they were sorta-almost-common around then, but I think most everything used a 287 until the 387's finally came out, if I'm remembering things properly! And for the retro-lulz, I still remember my joy at affording a 486DLC to upgrade my 386 with. It was so much faster with just that bit of cache in the CPU! And then I bought a 486/25 like a month later, but hey.
It was extremely common for 386 mainboards to have external cache back in those days. If a board didn't (or if the cache chips were bad), believe me, you knew it.. because it ran like an 8086 if it booted up at all.
They may be hopelessly under-powered but some of those early boards showed a lot of creativity where it comes to adding functions on a relatively low tech board.
Low tech?? It's more or less exactly the same technology! From the 8088 PC/XT (or even Z80 CP/M machines) to the Pentium. Only more and more integrated.
As for MB #3. IIRC there were some MB's that were socketed for BOTH the 386DX and the 486. This would have been when the 486 first came out and was still quite $$$$. I think there were also some 486 "over drive" chips with 386 pin outs to upgrade a 386 to a 486. So that smaller socket could have been for a 387 FPP. That last MB can't be a 386sx, those chips were not socketed, they came in a package that had to be soldered to the MB. There were some 386SX "over drive" chips that were really PC boards that plugged into a 286 socket that had a 386SX soldered to them. Did someone sell a 286 mb that came with such a chip? Since there were also 486 overdrive chips in 386sx clothing, on could have upgraded a 286 to a 486 sorta-kinda. The 386sx would work with the 287 chip, as the 386DX would work with the 287, but it PREFERRED the 387. The motherboard could have had sockets for BOTH types of co processors, but please only install ONE thank you!!! Win-386 could have referred to the OS that was installed. Yes, Windows-386 (version before Windows 3.1) would work on the 286 processor (poorly!). You can make some RubeGoldberg SIPP pins by using either Swiss Pins or Swiss Sockets (Augat), Adafruit stocks them. The pins will solder right to the SIMMs, the sockets will require a small bit of wire pushed into them first. (www.adafruit.com/product/3647?gclid=Cj0KCQiA9YugBhCZARIsAACXxeL_-ReeKoTLODQAvLm7eBnpIMqeTy1V4qE06aBexsyRd98cZueOHJQaAvZMEALw_wcB) IIRC, the smaller pinout for the FPP was the same as the PGA version of the 287, and the larger pinout was for the 387. OR that's the pinout for a 387sx made to go with the 386sx.
Yes, the "MOS-techology" firm had problems with boron contamination in their process. This would often break down the IC after just a few years running hot. Commodore really bought the worst semiconductor manufacturer there was :)
@@kaitlyn__L Haha, perhaps. But were they really cheaper to produce than ICs from, say, Mostek or Syntertek, or japanese, or korean manufacturers? I belive the key was that Commodore (a calculator company before PET and VIC) wanted a lot of custom designs. That was cheaper when you own the firm, of course.
@@herrbonk3635 I wasn’t around for it, but I’d read that MOS had much higher yields at the time due to their proprietary post-lithography bodge technique. Which contributed also to the frailty! As they turned failed chips into marginally-working chips. But the aspect you point out, that was also true!
with the last mobo and the cache it could just be the way the BIOS is set up to address it. I noticed the writes were performing fine (the blue line on the graph) so it may just be the way its set for reading from the cache that needs adjusting. Its been decades since I've had to play with that sort of thing.
Back in the old times, when OS2 (the IBM OS that was co-developed with Microsoft), there was 15/16 MB Memory hole option in the BIOS settings. Maybe, that motherboard has that enabled by default, so it is not caching in the region. Sounds plausible.
The jump from the 286 to the 386 was long (the 386 had an on-chip MMU and could mode switch easily, so could multitask DOS programs), but the jump from the 386 to the 486 was prodigious. The 386 had a very brief moment in the sun. I would team this killer board with an EGA video card and have an example of the best you could get before the 486 and VGA appeared. Put it in a period-correct case. The most significant decision in PC history was IBM's insistence on running OS/2 on the 286. Microsoft correctly realized that the 386 would solve the main problem of OS/2 easily, running legacy DOS software alongside the graphical UI more or less transparently. IBM's approach was to target advanced bus mastering hardware that didn't yet exist - Microchannel. It is a fascinating story.
Nah, the 386 was basically just a 32-bit version of the 286. Not much faster at the same clock, i.e. on the ubiquitous 16-bit code base of the time. And the 286 had an MMU as well. But you are totally right about multitasking of 8086-code, or switching between protected mode and "real" mode. The 286 was not intended for that, at all. The jump from 386 to 486 was a doubling of performance per clock, largely even in floating point. Very good indeed! But the only "prodigious" about it was perhaps that it scared the believers of the RISC-church :)
Check that the correct tag RAM chip is installed. Someone may have attempted to upgrade the cache to 64k and only replaced the memory chips, but not the tag.
With the last board that was having the cache issues, I wonder if it is an issue with the memory it's trying to cache? It's possible one of those SIMM's has a marginal chip on it that passes the memory check upon boot but fails when you actually try to use it. You could try testing the ram or moving the SIMM's around and see if the issue moves in the cache check.
Don't worry. Speedsys is buggy as hell in my experience. It doesn't work on a lot of boards/CPUs that are otherwise totally stable. If DooM runs it's totally fine but without a local bus video card you won't get the full 35FPS.. Yeah the DX4-100 has 16k of cache which was a pretty nice upgrade compared to 'normal' 486's 8k. Especially on machines without any L2 cache like the Presario you meant to mention (but called it a Deskpro) that made quite a noticeable difference I've got an AMD 5x86-133 Evergreen CPU upgrade in my Presario and it's a big step up compared to the stock SX-33! A lot of 386DX boards had cache on them.. My Deskpro 386 which is a 1988 machine has a 385 and 32k of cache on board for instance. Even some 386SX machines have cache on board. I sold an NCR 386SX-20 3315 desktop today which had 16k of cache which is indeed quite rare for an SX class machine 👍 The coprocessor PGA socket on that 386 board has that extra row of pins because it was meant for a Weitek Abacus coprocessor as wel as a 387 class FPU unit ;-) Peter from CPU Galaxy made quite a nice video about this very interesting piece of hardware, the Weitek Abacus 3167 👌
Wow I thought I’d never be so excited to see post codes as I was when you powered on that dx100. As for the 286 board, I thought 286 cpus where long straight looking not square. I had an early one maybe?
That Cyrix486dlc is one odd creature. Ran across several over the years. However, I remember that they had been extremely picky with the clock speed set to exactly what it was paired with before they would post. With that being a 40Mhz, try it in a board set to 40 before you can it if you hadn't already? Or melt down the pins and make a nice gold C64 badge or something!
Unlike Intel, those did not use much gold (if any at all). And if i recall correctly, the 40mhz would run fine at 33mhz, just not the other way around. I had the 40mhz part for some years, did all my Simcity 2000 plays with it and many other games, in the super odd 386 motherboard that sported a single vesa local bus which i found a "svga" to use it with, ran windows 3.1 with 16/24bit colors. That was before moving to a proper motherboard for 486DX and TWO VLB slots... Before the Pentium era and PCI slots.
You seemed genuinely surprised about the dead Cyrix CPU. As I recall, they had a reputation for running really hot (basically overclocked from the factory in an attempt to compete with Intel) and would have a very short life if not properly cooled. If they were run hard for a long time, they could also die due to electromigration.
Electromigration from the excess heat yeah, they needed active cooling (aka fan), you could absolutely not attempt to overclock it, but underclock was fine. If that cpu was ok, running at 33mhz would have been no problem. The Intels were famous for allowing overclock, but my trauma from Cyrix put me away from overclocking ever after. Ah the days the Chinese would paint a 75mhz pentium as 90mhz pentium. Put some cellotape and it would rip off, those were the days... or not, i avoided Intel cpus during the pentium (3v) days because it was too difficult here to get a genuine intel and the amd/cyrix being un overclockable made them safer to buy, if you did not attempt to overclock, that is.
Yet another great video!! I do have a question and that is, how long do you have to wait for Deoxit to dry before you can power up the computer. I've been wondering about that for some time. Also I was wondering if they sell Deoxit that's not in a spray bottle. I'm asking as I have a few vintage computers that could use some Deoxit but not wanting the spraying mess.
Element 14 had a whole video about Deoxit and all of its variants both in terms of composition and application. Adrian usually collects spray deoxit into a small container when the plastic spray bit breaks though 😂
![Adrian's Digital Basement ][](http://yt3.ggpht.com/ytc/AIdro_k7CrQRe71qN6c0Ua7A1P_I8KkxjqLa70u0zYeCvdvnKw=s48-c-k-c0x00ffffff-no-rj)
Patron Edmond pointed out the Intel cache controller can only talk to 32k of SRAM. I took a look at the datsheet and I totally agree. It says one bank of 8k x 8 chips or two banks of 4k x 4 chips. Either way that's a total of 32k. It could be that the 8k x 8 chips in there are only half being used -- which means half is wasted!
Also, the 15th MB isn’t being cached because of the “ISA Memory Hole”, which is used for memory mapped ISA devices. Later caching boards sometimes have the option to enable caching this megabyte, when the user knows there are no memory mapped ISA cards.
Ah right -- so that's be design. I do recall seeing that option before on some motherboards, and always wondered about it.
Wonder if that PAL chip is supposed to do some sort of bank switching to support the full 64KB of cache.
the Tag RAM also has to be compatible with the amount of cache installed.
@@EvilTurkeySlicesthe manual seems to indicate that. Says that 32 is direct mapped, but 64 is some mode I don’t recognize. Probably some long lost driver to make it work
The phrase "That's not very cache memory of you" came up in my brain during this episode. Not gonna question it lol
i just had a deja vu of hearing that before while reading your comment
Lmfao
I'll now go to the attic. See if there's any cache in it.
It's midnight in the UK and I'm watching Adrian's Digital Basement. I cannot fault my life choices. I had so many chances to collect these old boards, but I always passed them on to people without a computer. Love your channel Adrian.
According to Wikipedia the 80287 and 80287XL both work with the 80386 and were the only options available when the 386 first came out; the 80387 didn't come until two years after the 386 was released.
That's right -- I've seen 386 boards like that, but I think the 386SX (which is really a glorified 286 with a 386 stuffed inside) uses some kind of different FPU package. (Rarely bought since the performance of a SX was so substandard anyway.) Somewhere around the lab I have one of these FPUs (Which I purchased off ebay years ago.) I really should make a video about that!
@@adriansdigitalbasement2 Its the XT trick again: 16bit bus with 32bit cpu (xt was 16bit cpu with 8bit bus). Its like a faster 286 or a slower 386. But being actual 386 helped in things that required the protected memory like windows 3.1 or early linux, etc.
I have a Compaq SLT 386s/20 portable which is a 386SX. It has a 387SX FPU, the packaging is similar to that used by the 286 CPUs.
Having started with a VIC--20 myself, you are the Bob Ross of Computers inasmuch as artistry, and yet the Human Wikipedia of ideas.
Very special boards you have there 🤩. For Speedsys make sure to use ver. 4.78. On some chipsets speedsys really like to freeze at the memory detection. You can start with „speedsys dspdr“. This is disabling the memory type detection and u can use the program then just normally. Thanks for your videos. Cheers, Peter
Those Micronics 386 DX mobos were awesome late 80's goodness! I put an AMD DX 40 and loaded it up with like 16 Megs by the mid 90s. Used it as a Windows 3.11 machine. Totally useful.
oh god she's barely warm time for a heat sink to cool her down before she bursts into fire!!🤣
Hi Adrian, great video, brings back some memories. Those SIPP pins you are looking for are available from several manufacturers - do a search for "leadframe pins". They are still used quite often in electronics where configurable modules get plugged/soldered into the main board. Power supply modules, hybrid circuits, and double-sided modules on single-sided boards are very common applications. Anyhow, I love the channel and the content. Cheers!
p.s.: Leadframe pins are also used on small modules like the ESP-32 and Pi Pico, instead of Molex headers, to prevent them from spreading (obliterating!) the contacts in a standard socket.
Do you have a link? Ebay came up blank. Maybe something at Mouser or Digikey?
I've always just known those as 0.1" pin headers.
@@AureliusR They are a little different, with round pins instead of square, and slightly smaller in diameter. Breadboards *used* to be made for components and small jumper wires, so the old pin-headers would damage the sockets after a while. I think that modern breadboards are made to accommodate pin headers, so it is not such a problem. However, **chip sockets** are made for very thin pins, so pin-headers *will* damage them - or at least spread the contacts out enough so that a DIP-chip won't fit tightly any more. That's why the leadframe pins work better: they are close to the thickness of a standard DIP chip component and they don't spread out the pins. Apologies for the verbose comment, that's how I roll. Cheers!
We had a 286 with a Phoenix BIOS, and I think it used a CTRL-ALT-ESC combination to access the built-in setup program. That combination would work at any time, even after the system was booted.
That cyrix chip may not actually be bad, the 486dlc was not intended as a direct drop in replacement for just any 386. It was designed for ,often later, cyrix aware or enabled motherboards. Though sometimes you may luck out and it works anyway, but often with cache and other stability issues. For a direct replacement for the 386, they had 486DRu and 486DRx series chips.
I'd like to add that these Cyrix/TI DLC CPUs are *very* hot and power-hungry. I had a board which refused to start with them until I recapped it with solid-states (original caps were regular electrolytics).
@@Arti9m though similarly labelled, the ti processors were even different iirc. I think it wasn't until the true 486s ang later where the cyrix, to, and IBM were pretty much the same, just different fabs and labelling. Could be mistaken.
Were there any other direct replacement 386 chips that offered performance increases?
@@rgi9509 I'm not sure honestly. I'd expect there would be.
Entered the PC world around 92 (after the Amiga) with the Pentium, so it's fun to see all these older boards that I never got to experience. Everything looks extra chunky! :)
Yeah, this is the era when I was learning and getting into pc's as a young kid. Hand me down 286's, 386's, and my favorite processor of them all, the 486.
Learning and playing with all the old stuff helped me pass my A+ certification with flying colors when I got to high school years later.
why off topic, Amega was already dead in 1995, why you cry i need old garbage, what you run on it ???
This sounds funny but those board are beautiful works of art.
DOS 6.22 also has the handy F8 option during startup to selectively choose which lines to run in the autoexec and config sys. This way you can choose whether to run EMM386 and/or HIMEM and so on without having to edit the startup files in some situations.
Also MS DOS 6.22 supports custom boot menus for this case
@@root42 True, that's also a handy option for commonly used configurations, especially if you frequently use 2 or more start up scenarios. The advantage of the F8 option, though, is to be able to toggle numerous permutations of existing options without further modifications to the configuration. It's mostly for those one-off scenarios and/or rarely used scenarios, and prevents expanding the config file too much.
In a nutshell, it all comes down to whether you use alternate startup options frequently enough.
The best way to deal with SIPP memory these days that I found is this: get some SIMM sockets. The pins from those fit perfectly into SIPP slots, and you can install regular SIMM modules in them afterwards. This way you don't have to modify either the motherboard or the memory modules. Or if you're so inclined you can desolder the SIPP headers and solder instead SIMM sockets.
As others have said, the length of the video is fine. It's total fun!
The '90s-ish big, hunky boards are complete nostalgia for me, they're what got me into computers at such a young age, this was a fun watch!
We used those Micronics boards in our frash test systems, they were fast 386s. They would be loaded with the RAM card, a DSP5600 card, WD graphics, Inmos Transputer board, 3c509 later on and our proprietary Comms card. These ate power supplies
Some versions of the Dallas chips can take an external battery
Don't worry about the length, I usually watch your videos at 1.5x, sometimes 2.0x but I never skip one :) This is a promising video, looking forward to watch it later!
Edit: I watched it and it was fun. At 41:40, that motherboard looks pretty, I feel like biting into, looks like chocolate :) You have much more luck with motherboards than I have, I was looking for a neat 486 motherboard and had to buy 2 broken ones before I got my hand on a cool working one. I thought I was onto something with one of the boards but I had no luck, switched some components, tested the BIOS in a different motherboard (it was okay) and nothing helped to fix that one, I think maybe the chipset was fried or something.
5 sec, and i pressed stop, more than enough, he does not understand it !
Also, one of the things I've had immensely good results from when cleaning boards is hot water, dish soap, and baking soda.
The baking soda helps clean mechanically, and the hot soapy water cleans chemically.
Combine the two and it's how I get some of my immensely disgusting boards looking a million times better. I had a dell 386 machine complete with leaky battery pack that was left in a rotting shed outdoors for 20 years, and it come back to life with nothing but cleaning. It was in such bad shape that the base was almost rusted entirely through.
Fun fact, soap also cleans mechanically :) but at a much smaller scale, dismantling grease and oil
My little computer shop used to sell DTK computers back in the mid 90s. They were my "regular" line. Touch computers with Asus MB were my premium, then I had my own line of custom computers that used Asus and Gigabyte MBs. Those were simpler times...
Hi Adrian. If you ever get your hands on an EISA (32 Bit) motherboard, that would be very exciting!! I had one back in the day. They seem to be very rare nowadays!
They were always pretty rare, never got far outside the server & similar market.
If you have the actual EISA boards to play with. It can take normal ISA IIRC but that would be boring. Oh what should i do with my 386 class single Vesa Local Bus slot motherboard... And you thought EISA was rare 🙂
will definitely be remembering the AT standoff/BIC pen trick, hate trying to get those off with tweezers
I have and so have others I know overclocked 386/25 to 33 with a heatsink/fan with no problem back in the day.
I mentioned your channel in the comments on JayzTweCents, His last video he showed a C64 and I said your channel is great for trouble shooting etc.. cheers.
Awesome bunch of vintage motherboards. It always amazed me how they could make those multi-layer boards with such hair-like traces and tiny vias, and get everything to line up when the layers were put together. Wizzadry at its finest 👍
Seeing all those motherboards always brings back memories of my early PC days 😊
I do remember using the Harris 20 Mhz 286 and having to find ISA cards that could work with the bus sped up. Then the computer store I worked in found some (really small) 286 motherboards that had jumpers to set the clock divider for the ISA bus. Well into the 486 era there was still a small market for 286 systems.
9:44 That's an MR-BIOS! If it's not yet backed up/archived, please back it up.
Really enjoyed this video, just feels like sitting next to you watching you slowly go through your process, and picking up tips for my own projects along the way
I can imagine the size of the computer for that huge board. WOW. They were so big, but without impressed power compared today. Now days motherboards have shrunken extremely small. Pretty crazy. And lots more options. My first PC was not as big as those Motherboards. I wish I still had that one, and my Commodore 64. I do miss programming with my old commodore 64. I really enjoyed watching you test those Motherboards!
That first full size 386 motherboard looks amazing!
love the prosaic 'battery is dead' message, that was pretty ahead of its time 😅
POST = Power-On Self-Test. It's not only what it really means, but it's also self-explanatory.
The coprocessor socket of the 386 board is very likely a Weitek-enabled socket. The 387 from Intel and compatible (Cyrix, ULSI...) have two pin rows/squares, the Weitek has a third row/square so a little wider socket. To support Weitek, you need the wider socket (3 rows instead of 2) and BIOS-Support (BIOS will show "Weitek Coprocessor" Option somewhere in advanced settings).
The 287 on the 286 Board seems to be a little bit off to me. The 287 is actually clocked at Crystal/3 while the CPU is at Crystal/2 (meaning the 287 runs at 2/3 of the MHz of the CPU). Never heard of half.
The IBM 5162 is a 286 which runs with zero wait states... however, IBM crippled the BIOS so that it couldn't be run at a higher clock so that it doesn't make the 5170 look bad.
If there is a bad cache tag chip, then that could make some parts of memory uncacheable.
It is pretty normal for 286 PCs not to have a builtin setup. You had to use an external dos app. Often setup.exe
Sometimes you need to have the particular setup.exe for some boards due to extra features.
Was about to say this.
"Gsetup.exe" is used on old HP-Vectras
Early ones had dip-switches :)
was just going to say the same, have had a couple like that over the years.
@@herrbonk3635 Those would be XTs, the IBM AT and clones had none - Generally.
Around 1999 we were running Unix machines with 16 MB of RAM and 20 users on them. It's amazing to see that much RAM on a little PC board.
Unix and terminals were totally backwards already in the mid 80s though :)
@@herrbonk3635 Unix was on the way out? So you didn't go to a university in the mid-80s? Or work at a large corporation with a huge database? Or work at a bank? Or work at a store with a large inventory spread out over several states or countries?
@@scottlarson1548 I didn't say it was "on the way out". But it seemed sooo old fashion to me, that I always wondered *_why_* it wasn't :) So totally backwards in user interaction (1950s style), but still presented as "the future".
I didn't criticise that a large firm had a central data base, of course, but you didn't need Unix for that. Lots of other systems had networking.
(In the early 80s, my idea of what a _personal_ computer should be was very much along the lines of today's mobiles. I was inspired by instrumentation and advanced calculators from HP, Tektronix, and all the japanese. Early GUI I've seen too, from Xerox Alto to similar attempts on home computers.)
@@herrbonk3635 So you were completely out of touch with the computer industry. People didn't need expensive GUIs to sell carpet or car parts or track their inventory or payroll. Businesses could buy a DEC terminal for a fraction of the price of a PC and it didn't need maintenance and your employees didn't play games on it all day. What you're describing wasn't even an option.
@@scottlarson1548 Yes, at least that part of the industry. I worked as an embedded systems programmer and electronics engineer in my youth (before and during university). So pretty far from that administrative world.
SIPP: Single Inline Pin Package. Also, can we just marvel at how nice the first two boards look with their respective parts color schemes? I thought the bright red resistor packs looked nice on the first Micronics board until I saw the second one with its 100% color coordinated capacitors.
Indeed. Have to wonder if that was intentional? It was a sure thing that both of those big boards were really really expensive.
Back in the early 90s me and my dad tried to upgrade the 386 to 4 megs, but we missalinged the modules by one pin which fried the board 😥
@@adriansdigitalbasement2 it _had_ to be intentional. They're too coordinated to have just been colored like that by chance. I agree those would have been very high end boards, so making them look nice and coordinated was probably the start of the "gamerificiation" of modern motherboards. Glad they all still worked, too!
@@belzebub16 wow, that's awful. I'm surprised it fried the board, honestly. I'd expect the module to die, at least, but the board is quite extreme.
Back in the day, I never had a MATHCO. They were beyond my meager means. But it's amazing how common they appear to be these days.
And incidentally almost nothing used it, so they were of very low need while costing as much as another cpu. Unless you had some specific niche that required it, like autocad. Only when it became included inside the cpu, were developers starting to make use of them.
The board with the serial and parallel ports was aimed at the upper end of the market where a scsi card would be installed that would have the floppy interface.... I remember these boards....
The 286 looks like one i had in the days. It required a setup program loaded from disk.
I have a bunch of 386s. Normally, sx would not have cache and dx would have. From my memory of those days it was just more expensive, and i remember at some point the shop i worked at would ask customers if they want it with half cache or full cache. That 287 math co was expensive and the only time i remember a client bought one was to run Autocad with a 286. Ah that DTK brought memories, i still have the computer desktop case (with the original DTK packaging box!) where that would go in. I kept the DTK case over the decades after various motherboard upgrades. That 486 with 16mb is luxurious, win95/win98 capable. Indeed post codes are different, there was documentation about them, i bet online you can still find it.
Third board is commonly found in Gateway 2000 systems before they started the PS/2 keyboard and mouse port shenanigans.
That DTK 386 motherboard would be a great candidate to stick in that huge DTK 386 AT chassis you did the rust conversion on years ago!
I really enjoy these test videos. Keep em coming!! ;)
trash it ? why you need these ????
@@lucasRem-ku6eb sadly, there's a lot of people with your trash it attitude. The vintage computing community is large and growing. Old computers and components are no longer in production so every time old systems get trashed is a lost learning opportunity for vintage computer enthusiasts.
Nice, I use to work and repaired these motherboards. Most of the time it had buffer problems.
I love these PC motherboard inspection/repairs you do. Really interesting
286's had setup on floppies early on. I think I recall the IBM PC/AT needing setup floppies, and I definitely remember the PS/1 and PS/2 systems using setup disks. Without them, you were just stuck with the error. Some 386's did that for a bit too, but 'not saddling the customer with setup disks' cut the cost to support users too since they lost them all the time.
On the third board, the 286-10 that you caused me to add this second comment for, the reason it might not have included the floppy/IDE connector is that back then, there was a lot of MFM/RLL stuff out there, 'special controllers' galore for odd freaky "the size of your computer" external hard disks/tape drives, SCSI, even ESDI! Ah, ESDI...that awkward step between 'they used modems to transfer data from disk to PC' and "Actual digital signals" with IDE. And I exaggerate slightly about the 'modems for data transfer' but the drives used to be stupid and BRUTALLY slow. Before SCSI was worth anything, ESDI was a good higher-performance Back then it was amazing, of course! Not loading from floppies was way faster...and floppies were way faster (and better!) than punch-cards, which were themselves better than binary-toggle per-instruction boot-up processes! Set switches, flip toggle. Set switches, enter the third byte. Repeat forever...
For "the cache being missing", many of them had jumpers to set 16/32/64k, and some did run 4 chip cache modes, though I don't recall 'what density they actually were', I just know that I recall 4/4 cache/empty setups being valid, so maybe there's a missing/incorrect jumpers as you noted. I recall bad cache being either 'the end of all the cache' or 'the system didn't boot', not that it worked only half-way. Not caching 13 or 15 is ODD, though. "Not caching over 15" might make sense in a '16Mb and up' thing bug..? Strange that they don't have a cache size jumper setting, however...well, I dunno. The video isn't done yet and you're silly-tired, so we'll see what happens before the end! :)
Thankfully, I started with the IBM PC, so I only know the punch-cards and 'toggle-panels' from historic videos and retro people using it later on. I did the full evolution from 'single-density floppies' to 'Woohoo!' to 'ZIP disks' to "USB thumb drives"...it's been fun to see the difference across the eras, that's for sure.
awesome video, can't wait to see what is going on with the cache -- really, half is wasted ? gee
Those are all awesome. I really enjoyed this production. I would love to buy that 486 off of you.
The kind of socket that the SIPPs go in is still easily available new (basically, a single row precision socket). If they are factory new and not soldered/clipped, THEIR pins fit themselves quite well. Basically, precision sockets are like LEGO bricks (they also make fantastic stackable connectors for lab modules, if you manage to solder wires to the side of the pins without melting the plastic too much!). So you can basically take that piece of socket (or pieces of precision DIP sockets even), and just solder-bridge or wire the socket ends of them to the pads on a SIMM. Presto, it can haz pins! One thing to be careful about: Best tin, solder-wick and retin anything that is gold plated (socket or pad) before soldering to it, too much gold getting dissolved into the solder can make some types of solder extra brittle!
I think those blue tantalum capacitors are pretty cool, from an aesthetic point of view.
I'd look for datasheets for the cache controller (jumper settings) and verify them against the jumper pins. Experienced a few mobos where the manual have been completely wrong (if you even were to find them for this particular board) :)
Indeed the datasheet for the controller indicates it only supports 32k - so the 64k on there is being partially wasted. Someone clearly had money to burn :-)
@@adriansdigitalbasement2 Well my idea sucked lol. Happy you found some info at least :)
POST codes are sent out on I/O port 80. POST cards were sometimes referred to as "Port 80 cards".
First PC I built was a clone XT board with a new crystal to match my NEC V20 CPU, in a 5150 case and with a green screen, all bits bought from auctions. First HDD was ST-225. Then 286, 386SX, many 486s, after which it became my work rather than a hobby. Fun times.
The extra pins are for a different type of FPU, specifically the WEITEK FPU which connects to the CPU differently from a standard 387 or compatible. In fact there are boards that have instead 2 FPU sockets.
Due to how they interface differently with the CPU it should be possible to actually have both a Weitek FPU AND and 80387 FPU on the system at the same time and even use them simultaneously though I never got nobody to confirm this (although electrically this is possible) and worse is that if you want to test simultaneous use of the two FPUs you'll need to make the software yourself as AFAIK there is no software that will use both simultaneously. Either or yes, and, not really which is a pitty. For example, CYRIX went to crap because intel CPUs could basically do integer and floating point math simultaneously (similar to a dual core) which the CYRIX ones could not (AMD also could do that by the way) and QUAKE actually took advantage of this feature which meant math wise QUAKE would run twice as fast on Intel/AMD VS running on Cyrix. Now, imagine we could have it use all the math processing devices hehe :) So Integer math on the main CPU, then 2x floating point on both FPUs.
Oh yeah, i have a faint memory of that, never saw one in reality. Weitek...
On that 286 board it's possible that the company installed a 386sx upgrade module in the PLCC socket, hence the sticker.
Remember the days of sending POSTcards to each other. The problem was remembering the correct POSTcode.
If my memory serves me right those Micronics boards were considered very high quality boards in their day so I am not surprised it still works. Even still, battery corrosion kills most boards whether they are good quality or not.
Now I wish I would have keep all my old computers. My most odd was the Intel PR440FX board with dual Pentium Pro's running Windows NT. That was a rock solid machine at the time. Had SCSI HDDs and DAT tape drive backup, and 256MB of DRAM at the time and dual 200MHz Pentium Pro's. I then did an upgrade and got the Pentium II Overdrive 333MHz CPUs. Ah those were the days back then!
My first PC (after my IIe) was a 25 MHz 386 with cache. It outperformed a lot of 33 MHz systems.
Hi Adrian, great video, I also have a DTK 2500, I could never get XTIDE to run, it just hangs. I originally got it to harvest the DRAM as these are also used on many Amiga 2000 RAM cards, and was way cheaper than buying the DRAM chip on their own, but decided I liked the board. There is a video on my channel where I installed the board into a full tower case (Retro 386 SX 25 PC Build) I later released it is a 386DX, but no one noticed 🙂
25 upped to 33 isn't much of an overclock. It's also possible it's a 33 labled as a 25. Intel did that from time to time to meet demand on lower skus.
Well, they certainly were surprising to see them all working for the most part, even the bent one, I thought for sure that Bender there would have needed some form of repairs or something to get it to wrk, but, nope, it's a runner, unlike modern PCBs that fail if you so much as breathe near them... :P
Modern PCBs fail because of BGA chips and crap lead free solder. The slightest knock will crack the tiny solder balls.
A lot of earlier 386 boards did make use of 80287 coprocessors. The 80386 came out two(?) years or so before the 80387 was ever made, so people who did CAD and stuff needed SOMETHING. Something like the 80287XL's which were '387sx chips in a 287 socket' also existed, I think? I don't remember when those came out, but the 287XL's were a lot faster than the original 287's. Any copro beat none for the right workloads, of course. Like POV-Ray! Oh, the memories... CAD too, I'm sure. Lower-cost motherboards "later in the early-years of the 386 era" cheerily used 287's since they were sorta-almost-common around then, but I think most everything used a 287 until the 387's finally came out, if I'm remembering things properly!
And for the retro-lulz, I still remember my joy at affording a 486DLC to upgrade my 386 with. It was so much faster with just that bit of cache in the CPU! And then I bought a 486/25 like a month later, but hey.
It was extremely common for 386 mainboards to have external cache back in those days. If a board didn't (or if the cache chips were bad), believe me, you knew it.. because it ran like an 8086 if it booted up at all.
The 386SX can actually use the 80287 FPU, not usual, but would work.
27:52 "Build-in" sounds like something that comes out of the box but is not yet implemented as in "to build in" vs "was built in" lol :D
Those Cyrix 486DLC 40 chips are pretty rare. make a good shelf piece.
They may be hopelessly under-powered but some of those early boards showed a lot of creativity where it comes to adding functions on a relatively low tech board.
Low tech?? It's more or less exactly the same technology! From the 8088 PC/XT (or even Z80 CP/M machines) to the Pentium. Only more and more integrated.
As for MB #3. IIRC there were some MB's that were socketed for BOTH the 386DX and the 486. This would have been when the 486 first came out and was still quite $$$$. I think there were also some 486 "over drive" chips with 386 pin outs to upgrade a 386 to a 486. So that smaller socket could have been for a 387 FPP.
That last MB can't be a 386sx, those chips were not socketed, they came in a package that had to be soldered to the MB. There were some 386SX "over drive" chips that were really PC boards that plugged into a 286 socket that had a 386SX soldered to them. Did someone sell a 286 mb that came with such a chip? Since there were also 486 overdrive chips in 386sx clothing, on could have upgraded a 286 to a 486 sorta-kinda. The 386sx would work with the 287 chip, as the 386DX would work with the 287, but it PREFERRED the 387. The motherboard could have had sockets for BOTH types of co processors, but please only install ONE thank you!!!
Win-386 could have referred to the OS that was installed. Yes, Windows-386 (version before Windows 3.1) would work on the 286 processor (poorly!).
You can make some RubeGoldberg SIPP pins by using either Swiss Pins or Swiss Sockets (Augat), Adafruit stocks them. The pins will solder right to the SIMMs, the sockets will require a small bit of wire pushed into them first.
(www.adafruit.com/product/3647?gclid=Cj0KCQiA9YugBhCZARIsAACXxeL_-ReeKoTLODQAvLm7eBnpIMqeTy1V4qE06aBexsyRd98cZueOHJQaAvZMEALw_wcB)
IIRC, the smaller pinout for the FPP was the same as the PGA version of the 287, and the larger pinout was for the 387. OR that's the pinout for a 387sx made to go with the 386sx.
Isn't it amazing how robust and durable mainboards can be when they're not made by Commodore? :P
Yes, the "MOS-techology" firm had problems with boron contamination in their process. This would often break down the IC after just a few years running hot. Commodore really bought the worst semiconductor manufacturer there was :)
@@herrbonk3635 but they made cheap chips! ;) that’s all Commodore cared about!
@@kaitlyn__L Haha, perhaps. But were they really cheaper to produce than ICs from, say, Mostek or Syntertek, or japanese, or korean manufacturers? I belive the key was that Commodore (a calculator company before PET and VIC) wanted a lot of custom designs. That was cheaper when you own the firm, of course.
@@herrbonk3635 I wasn’t around for it, but I’d read that MOS had much higher yields at the time due to their proprietary post-lithography bodge technique. Which contributed also to the frailty! As they turned failed chips into marginally-working chips.
But the aspect you point out, that was also true!
@@kaitlyn__L Sounds like a pretty expensive (labour intense) way to improve yields though!
28:40 Cathode Ray Dude should look into that, could be the first implementation of something like InstantON :D
Just a hunch, but if I remember right with some of those old 286-ish machines you had to press ctrl alt insert during boot to get into the bios.
In the 90's, Cyrix "tried" to emulate the Intel's, but weren't up to spec. Later versions seemed to be ok. Same with AMD.
Speedsys worked right at the end there just as you turned it off to recopy over :)
1hr+ video, im saving this one for yard work :-).
40:56 It just occurred to me... I think that's an XT form factor motherboard, possibly intended as an upgrade part.
with the last mobo and the cache it could just be the way the BIOS is set up to address it. I noticed the writes were performing fine (the blue line on the graph) so it may just be the way its set for reading from the cache that needs adjusting. Its been decades since I've had to play with that sort of thing.
Back in the old times, when OS2 (the IBM OS that was co-developed with Microsoft), there was 15/16 MB Memory hole option in the BIOS settings. Maybe, that motherboard has that enabled by default, so it is not caching in the region. Sounds plausible.
I don't know about the Motherboard but I Cry . getting one working in a PC makes me Happy :) QC
That last motherboard booted into DOS super fast.
when my battery gently leaks :)
The jump from the 286 to the 386 was long (the 386 had an on-chip MMU and could mode switch easily, so could multitask DOS programs), but the jump from the 386 to the 486 was prodigious. The 386 had a very brief moment in the sun. I would team this killer board with an EGA video card and have an example of the best you could get before the 486 and VGA appeared. Put it in a period-correct case. The most significant decision in PC history was IBM's insistence on running OS/2 on the 286. Microsoft correctly realized that the 386 would solve the main problem of OS/2 easily, running legacy DOS software alongside the graphical UI more or less transparently. IBM's approach was to target advanced bus mastering hardware that didn't yet exist - Microchannel. It is a fascinating story.
Nah, the 386 was basically just a 32-bit version of the 286. Not much faster at the same clock, i.e. on the ubiquitous 16-bit code base of the time. And the 286 had an MMU as well. But you are totally right about multitasking of 8086-code, or switching between protected mode and "real" mode. The 286 was not intended for that, at all.
The jump from 386 to 486 was a doubling of performance per clock, largely even in floating point. Very good indeed! But the only "prodigious" about it was perhaps that it scared the believers of the RISC-church :)
I remember COAST. Cache on a Stick!
And the fake COAST scandal!
Check that the correct tag RAM chip is installed. Someone may have attempted to upgrade the cache to 64k and only replaced the memory chips, but not the tag.
With the last board that was having the cache issues, I wonder if it is an issue with the memory it's trying to cache? It's possible one of those SIMM's has a marginal chip on it that passes the memory check upon boot but fails when you actually try to use it. You could try testing the ram or moving the SIMM's around and see if the issue moves in the cache check.
26:45 - LOL I have the same DTK BIOS - it accepted type 47 so I was like.... yeah it works.
What a weird BIOS tho
Don't worry. Speedsys is buggy as hell in my experience. It doesn't work on a lot of boards/CPUs that are otherwise totally stable. If DooM runs it's totally fine but without a local bus video card you won't get the full 35FPS..
Yeah the DX4-100 has 16k of cache which was a pretty nice upgrade compared to 'normal' 486's 8k. Especially on machines without any L2 cache like the Presario you meant to mention (but called it a Deskpro) that made quite a noticeable difference I've got an AMD 5x86-133 Evergreen CPU upgrade in my Presario and it's a big step up compared to the stock SX-33!
A lot of 386DX boards had cache on them.. My Deskpro 386 which is a 1988 machine has a 385 and 32k of cache on board for instance. Even some 386SX machines have cache on board. I sold an NCR 386SX-20 3315 desktop today which had 16k of cache which is indeed quite rare for an SX class machine 👍
The coprocessor PGA socket on that 386 board has that extra row of pins because it was meant for a Weitek Abacus coprocessor as wel as a 387 class FPU unit ;-) Peter from CPU Galaxy made quite a nice video about this very interesting piece of hardware, the Weitek Abacus 3167 👌
Wow I thought I’d never be so excited to see post codes as I was when you powered on that dx100. As for the 286 board, I thought 286 cpus where long straight looking not square. I had an early one maybe?
That Cyrix486dlc is one odd creature. Ran across several over the years. However, I remember that they had been extremely picky with the clock speed set to exactly what it was paired with before they would post. With that being a 40Mhz, try it in a board set to 40 before you can it if you hadn't already? Or melt down the pins and make a nice gold C64 badge or something!
Unlike Intel, those did not use much gold (if any at all). And if i recall correctly, the 40mhz would run fine at 33mhz, just not the other way around. I had the 40mhz part for some years, did all my Simcity 2000 plays with it and many other games, in the super odd 386 motherboard that sported a single vesa local bus which i found a "svga" to use it with, ran windows 3.1 with 16/24bit colors. That was before moving to a proper motherboard for 486DX and TWO VLB slots... Before the Pentium era and PCI slots.
This video sure was a treat! 👍🏼🥳
Region 0xd000000 is at 13MB and 0xf00000 is at 15MB so both those must be configured to be noncacheable no matter how the jumpers are set.
I think Zilog is to thank for the inport and outport opcodes. I might be wrong, but i don't think so.
Someone send this man a static duster for his screens! Lol. I'm like 30 videos in, and every monitor is covered in dust.
jumper removed for full speed meaning that was where you plugged a switch in for the turbo button to function on it
You seemed genuinely surprised about the dead Cyrix CPU. As I recall, they had a reputation for running really hot (basically overclocked from the factory in an attempt to compete with Intel) and would have a very short life if not properly cooled. If they were run hard for a long time, they could also die due to electromigration.
Electromigration from the excess heat yeah, they needed active cooling (aka fan), you could absolutely not attempt to overclock it, but underclock was fine. If that cpu was ok, running at 33mhz would have been no problem. The Intels were famous for allowing overclock, but my trauma from Cyrix put me away from overclocking ever after. Ah the days the Chinese would paint a 75mhz pentium as 90mhz pentium. Put some cellotape and it would rip off, those were the days... or not, i avoided Intel cpus during the pentium (3v) days because it was too difficult here to get a genuine intel and the amd/cyrix being un overclockable made them safer to buy, if you did not attempt to overclock, that is.
try to use the ctrl-alt-esc or ctrl-alt-ins to enter the bios setup on that motherboard.
Yet another great video!! I do have a question and that is, how long do you have to wait for Deoxit to dry before you can power up the computer. I've been wondering about that for some time. Also I was wondering if they sell Deoxit that's not in a spray bottle. I'm asking as I have a few vintage computers that could use some Deoxit but not wanting the spraying mess.
Element 14 had a whole video about Deoxit and all of its variants both in terms of composition and application.
Adrian usually collects spray deoxit into a small container when the plastic spray bit breaks though 😂
Doesnt the TURBO button switch between the 2 crystals? at 51:56 you have a ghost on your screen.
G2 and Video 7 merged to form Headland Technologies.