The T.Res chip is actually a resistor network. Termination Resistor is what the name stands for. It only needs to be on the last drive in the chain, and also is only needed with some early floppy controllers. Newer controllers terminate the bus on the controller card.
@@PCRetroTech Yeah, the termination at each end is to reduce reflection when the signal reaches the end of the cable - it behaves like a "transmission line" so the closer you match the impedance the less reflection you get. BUT! floppy cables runs at such low enough speeds that the reflections really doesn't matter much unless you have very long cables (say 3m+), so most? all? PC floppy controller cards provides sufficient termination that it'll work with or without the other end being terminated. More people are familiar with termination-at-each-end for Parallel SCSI cabling, getting this right there is much more important since the signalling is 10-400 times faster than for floppy cables so the reflections are much more likely to cause issues. But it's still possible to sometimes get away with "incorrect termination" for short parallel SCSI cables and low (by SCSI standard) speeds, so it's logical that floppy cables are even more resilient.
@@Torbjorn.Lindgren That totally makes sense. I'm even vaguely remembering a video I saw where this was mentioned once, now that you've reminded me of the details. Anyway, thanks for the comments.
OHH, that's important to know; thanks for pointing it out. Turns out this info is actually on minus zero degrees too; look for diskette/5.25_floppy_drive_termination.jpg.
2 роки тому+5
It appears that the T.Res is on the wrong drive in his system! It is supposed to be installed only in the last drive in the daisy chain, the one at the end of the cable. Those drives from IBM have a DIP network, but some other drives used a SIP instead or a single jumper that you installed or removed. On a drive with the single jumper the resistor network was always there; the jumper lifted the ground connection to it, taking it out of the circuit.
Games that were designed to run on a wide variety of cpu classes hook the PC's interval timer circuit and use it to synchronize the screen updates. Thus you can have consistent game play across machines of different performance. Grand Prix Circuit is one of these games. Once you have a CPU that exceeds the computation power required to play the game without bogging down it spends time in an idle loop until it is required to generate the next frame.
Yes, that's correct. The other way it can be done is to use the vertical retrace of the screen, which is usually 60Hz give or take on most older machines.
@@dgmt1 For overall game speed maybe, but a lot of those games still had other unforeseen timing issues, such super fast animations, physics bugs or trying to access the sound card too fast. There were still some big name 90s games that were straight up speed sensitive too, such as Ultima VII and XCOM.
@@ozzyp97 Same on other platforms. I once tried playing an old Mac game (some Breakout clone, can't remember the name, I'd say early 90s) that ran fine on a IIcx with its 68030 on a Power Mac and it was absolutely unplayable, the ball moved lightning fast.
the bad adjustable colorburst cap could have caused excessive noise back to the CPU clock line, IBM PC and XT boards both used a 14.318 mhz crystal and I -think- it was divided down for various functions in the machine including the CPU clock IIRC.
I like this era simply because you can repair them. In later machines one simply pulls the entire board out and puts a new one in, which just doesn't seem as interesting. I'm glad you are having fun with a 5160. It's a nice machine to own.
Love the video and the topics you explore! Unfortunately, you picked one of the worst games to showcase CGA RAM speed: Grand Prix Circuit is internally framerate-locked to 10fps, so it looks the same on any 286 or higher. A much better game to test how CGA RAM speed affects gameplay is Jet 2.0, which updates the screen with REP MOVSW and can run as high as 60fps on fast hardware. Otherwise, using a benchmark program like you did in the video is the only way to reliably measure speed changes. It is difficult to find games that can reach 60fps *and* support CGA. This is because, in the era of programming for CGA, it was impossible to reach such heights, so programmers cut corners that would never cause an issue with speeds of the time, like lock the internal framerate calcs to a low rate (10 Hz, as seen in Grand Prix Circuit). Or worse, not lock the framerate at all, so you see some games become unplayable on CGA on faster systems whereas the same game works fine on EGA because the timing is locked to the page flipping rate of 60Hz; Stunt Driver is one such game that comes to mind.
Wow 10fps, really. I felt really sure it was much higher than that. It seemed butter smooth in person. And I feel like I can usually tell framerates up to about 12fps. I'm not at my computers right now but when I get back to them I think I have to check that out to figure out how I fooled myself so easily. Probably I was just focused on the video so much I didn't even notice. It's definitely a shame I made such a bad choice.
Maybe FastDoom would be a good option to test with nowadays. It supports CGA now IIRC. Also: can you query vertical refresh on the CGA? I only know how to do that on VGA.
That variable capacitor (not a variable resistor in this case, btw) was most likely used to tweak the signal coming from the crystal next to it. I assume the crystal is related to the system bus clock, and that could explain potential ISA and floppy drive issues.
Yes, definitely. I can't believe I said variable resistor and not capacitor. I guess when it fell to pieces it actually did look a lot like one of those carbon pots, but of course that makes no sense given what it is doing. I suppose it could have been part of an LCR network or something, but that would be a stretch.
Yes the composite on CGA iir is derived from the system clock. The variable cap is used to dial the system clock into spec. You can tweak that cap and watch your composite video lose color burst on the cga card. Since I believe its tweaking the system clock in real time it would do funny things to bus timing thus parallel port timing. I have a generic Turbo XT that was giving me only black and white off the composite out on the cga card and needed a tweak to get color back.
@@joshhiner729 Yes, I think the cap was partly broken when I adjusted it. I made a video with the 8088MPH demo in it on a Commodore 1084s monitor and I adjusted it to get colour composite output. Apparently I overdid it and broke the cap completely.
I read that this IBM 360K floppy T-RES terminator resistor chip should be installed on the last floppy drive in the chain and it works just like SCSI bus termination, to prevent the reflection of electrical signals from the end of the cable in order to ensure reliable operation. Cheers.
2:45 "black & red = dead" even when I wasn't so fluently speaking English back then, we've been teached this rhyme back then when installing hardware. Before ATX came out. Thanks for your content!
It's actually "Red and red and you're dead". If you think about it, you can't have black and red in the center. It's either black and black, or red and red.
I suspect that the 1/3 option in the bios is not really working and it's actually still 1/4. You could test this by meassuring the frequency directly from the ISA slot (pin B20), if you have a multimeter that can do that high a frequency, or an osciloscope.
About the dip package that one of the drives lacks, if I remember correctly, old disk drives didn't use the twisted cable to identify the two drives connected to the same cable, so I guess the component is a terminator.
No it's to terminate the bus. The drivers are in fact both configured as drive 1 with a twist on the cable, but the floppy controller card needs the termination.
Nice to see the IBM working well again! The 486+CGA combination was something I used a lot back in the day, on a VGA card though... I had a black and white monitor and the strong colors helped with the contrast in games!
Yes, I think quite a few people will have done it with a VGA card or something backwards compatible with CGA. It was nice to find I could put a genuine CGA card in this particular motherboard. It made a fun experiment.
My first "proper" graphics card was a Cirrus Logic 5429, a VESA "Windows Accelerator" with a whole 1MB of memory. I was so proud of that powerhouse of a GPU.
@@PCRetroTech Unfortunately mine is gone; it was in my 486 which I gave to my mum, then (without my knowledge) my sister organised a new PC for her and the 486 went to the tip (erm, recycling centre). If I'd known I'd have taken it back of course. Sigh. I remember on Windows 3.11 the drivers were a bit buggy with some screen artefacts adding random dots sometimes. Worked perfectly on Win95 though.
@@ian_b That's a shame. A lot of cards ended up in landfill I bet. But I suppose the rarity these days is part of what makes retrocomputing and collecting fun.
@@PCRetroTech It is sad yes. I had a sort of perception of future retro value at the time, which is why I hung onto my IBM Model 60 (a huge PS/2 tower with a 286 processor), but even I foolishly sent the IBM monitor to the recycling centre, to save a modicum of storage space. I didn't think a 12" monochrome CRT was worth keeping; nowadays of course I wish I still had it. Oh well, live and learn.
@@ian_b Yeah, I've also thrown away stuff that I didn't have space for back in the day, or which just didn't seem to be worth anything any more. Funny how value changes.
@@PCRetroTech i like the new magazine format and doing fun deep dive techy things with vintage PCs! I built my own 8088 machine, the XTjr so I learned a lot about the architecture and got kinda obsessed. If you ever do Patreon/Membership let us know as be happy to throw you a few bucks if it helps.
When my VGA monitor died, I used for a short time (few weeks I think) a original Hercules adapter. I also had a CGA card but no monitor to pair with. Luckily the Hercules I had was paired with a beautiful amber mono screen. My first shock was when I entered the BIOS setup. My motherboard had winbios (graphical UI) but it was unrecognizable with that Hercules; it was displayed in a strange text mode. The next surprise was that Win95 was not able to boot with that card. So i took another HDD with ms-dos to use. The time I spent with it, was entertaining, it was like a lottery to run programs; would it run? and if so, how will it look? I also used a CGA emulator for Hercules that allowed to run CGA games on that mono card. I also remember that the only GUI OS that worked with Hercules was Red Hat Linux (ver 7 I think it was) that I installed and ran fine with X server GUI only. Win 3.1 or 3.11 would not run in mono, only Win 3.0 was able to. The only drawback was that I had no idea of Linux and no manuals available, so no way to learn how to do something useful with it. So I returned to Windows as soon as I got a new monitor.
That's a pretty funny story. I guess at least you managed to learn about which software would and would not work with Hercules. It's amazing how old technology just stops being useful at some point, even though one of the major selling points of a lot of PCs is that they are supposedly "future proof".
I always love this kind of mix of non-era part. Way back I ran an ATi Small Wonder with an amber monochrome in a Pentium 3 board. It ran as a Linux server for a while. Later I used it as a networked music player using mpxplay and a dos smb client.
I think mixing parts like this can be fun. Part of retrocomputing is about experiencing things as they were back in the day for nostalgia, and part of it is trying new things that we couldn't try back in the day. Both are fun.
T-RES stands for terminating resistor and it is used to tell the PC bus where the last drive on that control line is. So if you have two floppies, if memory serves, the resistor block should be on the B drive. If you have an XT with one floppy drive then the only floppy drive, A, should have it installed. Hope this helps.
To my knowledge, CGA either Black White Cyan & Magenta, or Black Yellow Red & Green colour pallettes, how was the game at 4:24 showing Black White Cyan and RED? How was it showing that combination of colours? Also the demo with spinning shapes is more than 4 colours, how it that possible?
There was an undocumented CGA palette. Wikipedia has a description of this. As for the demo that appeared to have more than 4 colours, it is an illusion. When you set the background colour to something other than black and then alternate pixel colours, you get the appearance of more than 4 colours. That demo was written by me and I have a video on the channel about how it was done.
You may need to move that T-Res chip to the B: drive if you want to continue using it. Although it may not be making much difference yet, that chip, which is a set of 150 ohm (or somewhere about) resistors is used for terminating the floppy drive cable and prevent signals from bouncing around. That may also have something to do with issues you've had out of it in the past. Check the Shugardt floppy drive standard for specifications.
My 286 PC has some similar issues to your 5150 where programs/games will just crash constantly in random places, seemingly even more so if you have more ISA cards installed. Really wish I had the skills to identify the issue and fix it, there's no RAM issues detected or anything like that when I run something like PC-CHECK.
I believe the reason why you only get a few different benchmark results on the CGA card is the wait states. The algorithm (implemented by the discrete logic on the CGA card) is basically: 1. wait 1 hchar, 2. wait for the next lchar boundary, 3. wait for the next CPU cycle. Step 3 will be negligible on a 486 so how fast you can access CGA RAM is limited by how well the access pattern of the benchmark code fits the CGA wait state algorithm. With a different benchmark you might get different results as to which set of FSB divisors give the same performance as each other.
Exactly. The CGA card runs at a dot clock of 14.318 MHz in text mode. A character is 8 pixels, so the CGA card runs at 1.79 megachars per second. You should get a video memory access time slot per char, so if your write cycles are short enough and in quick enough succession, you should be able to fill the video RAM at 1.79 MB/s. That's obviously way faster than the IBM PC could handle. If the processor could saturate the bus doing block fills (spoiler: The 8088 can't, the V20 can), the FSB would allow 1.19MB/s, so you would miss every other opportunity to access video RAM and lock in at half the theoretical 1.79 MB/s, i.e. 895 KB/s. The 486 chipset throws in a lot of wait states for 8 bit memory access, so it obviously misses the 895KB/s rate and instead only uses every third chance to access the video memory. This results in a theoretical rate of 597KB/s. The lower actual rate (583KB/s) might be due to RAM refresh cycles. They need to be visible on the ISA bus for compatibilty reasons, as some cards refresh their local memory based on ISA refresh cycles (mostly ISA memory cards). Some mainboards allow changing the number of wait states for 8-bit cycles using the BIOS setup, but that would be in the "advanced chipset options", and obviously you BIOS doesn't allow it. I checked the SiS471 configuration file for CTCHIPZ, and it seems waitstates for 8-bit cycles are software configurable as 4, 7, 11 and 16 ISA clocks. As I don't know what's the base line (i.e. how many clocks "0 waitstates" is supposed to take), I can't directly translate that to data rates. What we do know, though, is that at 16MHz ISA clock, you are close to "every other cycle" instead of locking in at "every third cycle", which you would lock in at 11MHz and 8MHz. We know you can't run at 16 extra bus clocks, because 16 clocks per cycle would mean a theoretical maximum of 500KB/s at 8MHz, which your board exceeds. Getting 600KB/s at 8MHz would require no more than 13 bus clocks per ISA cycle. Not getting to 808 KB/s at 11 MHz also indicates a likely total time of 13 bus clocks, so your BIOS likely sets up the ISA bridge at 11 "recovery clocks" that get added to the AT minimum bus cycle duration of 2 clocks. Setting your chipset to 4 bus clock periods recovery time should allow you to reach the extremely mad data rate of 1.7MB/s on a classic CGA card. Get CTCHIPZ here: www.rumil.de/uwsysuti.html
@@tw11tube The wait state problem is why OPL3 and Adlib cards sometimes play garbage noise on a faster machine. MIDI setup also can get goofy as well, particularly the MPU-401 systems.
Check with modern vga if it goes faster, if not, the game has fps control, or the speed of the physics in the game are not related to the rendering process, they are independent. This is done in games mostly playable on a network.
On the answer to the question "Why" there is an 8 bit port.. The simple answer is: BEFORE everything was integrated in the motherboard, you really needed those expansion slots.. plenty of them... and.. 1. Cost.. and 2. Motherboard real estate. Well.. the CGA is ONE card that actually DEMANDS it, because of the form-factor going down so a 16 bit slot would be in the way, but it wasn't alone. .. and.. if you buy a card (with special functionality) for one computer.. you pretty much don't want to be FORCED to give it up for your new computer. F.ex. I have a blind relative that had a speech-synthesis specific card in his computer. That was a full-length 8 bit card.. It cost a shitload... and I don't even think they ever made it in a 16 bit version, because suddenly people just had to have a sound-card and with a Soundblaster 16- and with the speed of computers, and available memory: By the time you had those- it was all done in software. .. but there was also a shitload of 8 bit cards that.. TECHNICALLY could've been mounted in a 16 bit port, but they still were only the 8 bit cards. Most people think of the ps/2 mouse as standard. That was not yet the standard on 486. 486:es were still being created when it finally DID become standard so newer motherboards may have a couple of ps/2 ports for keyboard and mouse, but most people had to settle for something else. Serial mice?.. they actually grew in popularity fairly late, so it was VERY common that the people who manufactured mice sold them together with a dedicated isa-card for them.. my first mouse was of the brand "Mighty Mouse", and it was such a bundle which had to come with it's own card and special drivers. LATER serial port mice became popular, but tbh.. If you had a modem (almost ALL modems during the 486 era were external and connected through a serial port.. I SUSPECT due to the interference problems).. Then I also got an MM-401 (mpu-401 compatible) midi-card. The thing is; We had a Roland Sound-canvas module.. I used it with my organ, OR with my computer. It was the second revision which is when it finally came with a MT-32 sound-set that you could switch to if you wanted to play any actual game.. While people were listening to FM-music while playing games, I had a full symphony orchestra while playing Might and Magic 3, or Sierra games.. (If memory serves; Even doom did support Roland sound).. Of course one wanted a soundblaster anyway.. or more exactly: Also.. because it played recorded sounds, so all the sound-effects in most games.. So.. Just sound took two slots.. Then of course.. there was the advent of CD-rom drives for your computer.. hint: They were NOT IDE compatible at the time.. you needed a special card to be able to use it... Since the actual transfer speed was 150kb/s.. you didn't need much bandwidth: Save money by only supporting the 8 bit bus. That said: A LOT of sound-cards had various CD-rom interfaces on them. I bought a bundle from Creative-labs, so I had a Soundblaster that ONLY had the ONE interface, but it WAS the right one (fortunately; I know people who bought the card and the drive separately, so DESPITE there being one or even as many as THREE different interface ports on the sound-card they STILL had to mount a separate card for their cd-rom drive).. Of course.. I was NOT helped by the fact that my soundcard supported it. Not much anyway, because I needed my CD drive to be external. Which WAS what the bundle was for, BUT.. it came with an extra card in order to move the soundblaster pro's internal port to an external connector. While we still were using a 286 we also had gotten a VGA card.. more exactly an Ultra-VGA card.. same amount of memory, but supported some extra modes.. most notably 256 colours in 640x400 mode (Our windows looked better than anyone else's and they had 386:es, so we really could make use of coreldraw).. And that was of course also an 8 bit card... the 640x400x8bit was of course nothing to play games on, but still.. we kept that card for a LONG time.. It was when we updated our first 486sx25 to a 486dx33 we finally got a trident 16 bit card. So.. cards I had in the same 486 that was 8 bits: Graphics, MIDI-card, Mouse-controller card, CD-Rom controller, and the serial/parallell port cards. I THINK The IDE was 16 bit, though.. while we had a 386 we were still using our harddrives from when we had a 286 and they were both MFM-drives and the controller for that was 8 bit.. Just to say; We had started with a 20 Meg drive.. then bought a 40 meg drive.. and kept the 20 meg drive.. BUT.. then we bought a different controller that used a different scheme for packing the data, so we literally had one 30 meg and one 60 meg drive.. "for free".. where the biggest drive anyone could have if they intentionally splurged was 120 meg.. but those two drives died after a while. So.. yes.. During the era of 486.. I DARE say that pretty much 100% of people had AT LEAST one 8 bit card in their computer.. AND a lot of people had many cards in their computer.. so.. doing ONE card so it CAN handle odd format 8 bit cards wouldn't even be a "lost port" for the people who didn't use a CGA or some other card that literally required such port, because it would still just be just another port that wouldn't be in use.
Yeah the CGA card was special in that it wouldn't go in a 16 bit slot, but most other 8 bit cards would. And it is uncommon to have an 8 bit only slot on a 486. But yes, there were lots of 8 bit cards still around for sure.
@@rogerlundstrom6926 I don't know why the discrepancy, but on theretroweb there are 669 socket 3 boards identified, only 91 have an 8 bit ISA slot. I have a large box of 486 boards. Only one of mine had an 8 bit slot.
@@PCRetroTech .. Well.. The discrepancy can be one of many.. ONE is that the amount of models itself doesn't show anything about how things are in the wild.. F.ex.. I once saw a 486-motherboard.. with only two 16-bit slots.. which DIDN'T have IDE nor ports on the motherboard.. so.. one multi-IO-card, and one graphics-card.. then.. you couldn't add anything. I assume it was cheap at purchase. It didn't seem to be brand name, which is why I THINK it was intended to be used in computers sold to people who only needed to provide a lot of users with a lot of computers at a cheap price, such as poorly funded schools, or extra cheap corporations.. Though it MIGHT have been that the motherboard was created for a specific case, where you were supposed to put in 2 16-bit bus-risers where the extra slots were in different directions.. That may be TECHNICALLY possible, but most corporations that does this even back then tended to use their own designed edge-connector instead of a 16-bit-bus.. So whatever it was.. I bet it was only sold in very very small numbers, but if using "number of models" as a metrics.. well.. IF that mother-board was in the list.. It would've counted as "1" just as.. actually USEFUL motherboards. It isn't unusual that there are local deviations that MAY depend on specifics such as what is popular in one culture, or even depending on one specific importer or manufacturer that exist in the nation in question. I live in Sweden, you don't, therefore.. It's kind of like how the local travel agent tries to find deals and then try to make it sweeter by negotiating, meaning that people living in one city are fairly likely to meet each-other when they go to vacation in a foreign nation. Things I can mention as examples is that I have never ever seen a Tandy computer in real life, so whenever I see 3-voice pc-speaker as a choice I gawk. and.. as I pointed out; There was a financial advantage if you could get away with only sending an 8 bit bus instead of going full 16 bit, so it's FAIRLY likely that a motherboard that has 3 16-bit slots and 1 8-bit slot was cheaper than if you had a motherboard that had 4 16-bit slots.. and.. in "pre-made" computers the person BUYING the computer won't even expect it to be cheaper, so the people actually buying the motherboards simply make more money by choosing the cheaper option if they don't even have a reason to buy a more expensive version, and this kind of is even MORE cost-effective if you yourself IS the manufacturer of the board.. and.. when selling to end users; Unless they have a REASON to want more 16-bit options, they would gladly settle for a board that has ~enough~ 16 bits option and put whatever peripheral-board they had (of which they most likely had at least one) in the 8-bit slot... This isn't exactly an argument that "shows" that it is so, but at the very least it show that the basic rule of "The consumer" and "marketing" means it makes sense... only people who doesn't have a single board that is 16 bit may see a reason to ~intentionally~ look for a motherboard without an 8-bit slot...
Some of the new retro games that run under DOS target the improbable combination of 486 with CGA (e.g. Castle of Viana). Now you have the ideal system for them! It was nice to see the IBM back to being a happy computer in the second half. Hopefully it will get its voice back soon.
Nice! I hadn't seen Castle of Viana, and that's precisely the sort of thing that will make an interesting follow up. And I do hope to fix the sound issue in a later video too. Thanks for the comment!
@@michalzustak8846 It's just not optimised well. People write stuff for emulators, which are 386 minimum and they do things the easy way, e.g. BIOS routines and high level languages instead of doing assembly language and the like.
@@michalzustak8846 Of course you can slow it down, but then these games are not playable. Viana requires about a 486 75MHz to be playable because it is not optimised. It was written as part of the CGA Game Jam competition where there is very limited time to write the game, so certainly not enough time to optimise the code.
I have a retro computer that is rather a Mac Plus than an IBM with a similar symptoms in the built-in display screen, which maybe might just have had some dirt between solder contacts of the HV output to the CRT causing it to ripple and flicker. I’ve never examined it internally, in spite that there was no compromise the last time it was used.
Looks like that floppy chip in question would be a terminating resistor pack. Likely should be installed on the last drive in the chain. Maybe early floppy setups were noisy and needed a t-res? Ive never heard of PC's needing a terminating resistor on a floppy chain as a requirement.
I think the variable capacitor adjusts the OSC signal on the bus, which can be used for all sorts of things, including affecting the DRAM refresh timing on some motherboards. That could explain why you were getting what looked like ram faults in addition to weird behaviour from expansion cards making use of that signal.
It was added to adjust the chroma signal output by composite CGA, which is locked to the system crystal. This explains why the video system was affected when it was loose and jostling aroung when the side of the system unit was smacked. Also see Geoff Reed's comment below.
@@JimLeonard Yeah this much I knew, since I had previously adjusted it to show 8088MPH. I've also talked to Reenigne a few times about it. I took the embarrassing section out of the video where I called it a variable resistor. But unfortunately I can't cut the multiple places where I call it a pot. But In a later video I'll make a section with viewer comments and clear that up. I had always thought it was a potentiometer, thought that was very odd, and even had go through my mind when making the video "I wonder if it is a variable capacitor", then forgot to check. So this one was a self-inflicted error.
Hi, nice video. About the CGA snow effect, i know it's due to the contemporary access to video memory by cpu and cga. I think it's because while the cpu is accessing the memory the cga get bad readings or it is blocked from fetching display data. I also know that not all CGAs had this problem. Do you know how they fixed the issue in snowless cga cards?
One (very late) comment on the racing game. It appears to use a graphics technique sometimes called beam racing. In order to horizontally scroll different regions of the screen at different rates, the game will update the frame buffer pointer register in the 6845 CRT controller at different points in the frame. This must be done continuously at the frame rate and with consistent timing during the relevant line blanking interval in order to ovoid very obvious jitter. This would probably place an upper limit on the game performance unless it was recoded for faster CPUs. If interested, UA-camr James Sharman has a video dedicated to this topic in his VGA From Scratch mini-series (an amazing DIY hardware and software project).
Grand Prix Circuit uses some tricks to make it easier to manage the sprites when they run off the bottom. There are two mirrors on the car. Those are the only regions where an oncoming car sprite can pass. If it intersects any other part of the dash, a crash occurs. That gives a hint that they draw the sprite for the oncoming car and then cover it up with the mirror, which has to be redrawn every game frame (game frames are a lot less often than CRT frames). Probably when the oncoming cars need to be drawn, they just wait for vertical retrace and draw those sprites first so that they are drawn in time. That is a kind of 'racing the beam', but not a very sophisticated kind. Most games could not use very sophisticated tricks due to the large variation in CGA hardware and PCs on which it had to run. You'll find a video on my channel where I analyse this game frame by frame. I think some of my conclusions are actually incorrect, as I was anticipating that they did something more sophisticated than they really did. But I did manage to recreate part of the game myself, albeit not the part we are talking about. My engine was many times faster, and nowadays I could do much better again. So it's certainly not a particularly sophisticated game engine.
Largely because the resolution was taken from the NTSC television standard (262 scanlines per CRTC frame) and because if you want to use more pixels you need more video RAM. If you wanted to use more pixels you'd need to increase the video RAM from 16kb to 32kb and that would have been extremely expensive back in the day for just a few more pixels. You could set up the CRTC controller to use more of the pixels so long as you reuse video RAM more than once for part of the image, but this is difficult to do on the fly. Some demos like 8088MPH and Area5150 do this.
0:10 Your hand shows the older 8/16-bit *emulated* PC/AT-bus (black slots) as well as the never 32-bit bus (brown slots) going *directly* to the 486 CPU. Cards normally use either one of these, not both.
The 8 bit slots have loads of uses - there are other expansion cards other than video cards you can plug into them! Scanner cards, i/o cards, modems, specialist interface cards, etc, and why would you want to waste a 16bit slot on an 8bit card!
Don't you think that some games would sync to the frame rate, and thus not run "too fast"? They might do better, finding they have more time for computation, but still present the game world at the proper speed. For all but the most trivial games, I doubt that it's just looping as fast as it can and thus the write speed of the CGA memory and registers bring it back to original speed.
What would be funny is to find the newest computer you can that would take that card. You might need a slot extender to make it fit onto newer boards. I wonder if it can be made to work on PCI only boards using a PCI to ISA adaptor ? if that exists
@@PCRetroTech Yep that's why I think the idea would be really fun to explore. Having such a limited card (in bus speed, capabilities) and trying it on faster and faster machines; how far can you go!?
_Before I say anything else, a brief apology to PCRetroTech for leaving you hanging on the CGA improper on stock thing: It likely (my current guess: 95%) would not after all work as I expected, though there might still be an outside chance (5%). I've not followed up and not gotten back to you because I got sidetracked and overwhelmed, including mentally, and I guess that latter part now includes that there's a war - almost beyond comprehension. In principle I still would like to get back to you on that one day, but in the meantime you're probably reasonably safe in assuming I was once again talking nonsense._ On plugging stock CGA into newer boards: Been there, done that, didn't know about the whole what's clocking what when I did it. The interesting thing is that at least well into the nineties and noughties, boards routinely used to have that one 8-bit ISA slot (and I wouldn't exclude the possibility that perhaps some ISA-compatible boards still do), though I'm not sure the board designers always understood why. That's because when I last plugged my IBM CGA into a maybe 386 or 486/Pentium board (hard to remember which), there were a bunch of ICs and various sufficiently tall components behind the 8-bit ISA slot that actually physically blocked things, so the CGA could not be fitted into that slot. What I ended up doing was, I cut out the very back at the end of one of the 16-bit slots, and then I used some simple insulating material, it might have been just a piece of paper, and wrapped that around the rear bottom edge of the CGA card, so I could physically push the card into both the non-insulated 8-bit part as well as the insulated 16-bit part of that slot - so that it fit but nothing in the 16-bit extension slot made contact. It wasn't pretty, it required cutting out the back of that 16-bit slot, but it worked. There was nothing else in the way behind that particular 16-bit ISA slot. I don't think the single 8-bit-only ISA slot that routinely was on boards back then was there for any other reason than to accommodate large cards whose PCB dimensions resembled that original CGA. But I think including one 8-bit ISA slot was one of those things everybody was doing, but not everybody who was doing it remembered why they were still doing it, and that resulted in mainboards whose inclusion of an 8-bit ISA slot was kind of useless -- like the board I had. I guess you could still justify that 8-bit slot as "Hey, you can fit a short ISA card in here without the slot taking up more space on the board", but then, why not just add another PCI slot instead, because we're prolly well into that era at that point? It's a nice little litmus test: If there isn't a "firebreak" of only low-profile components behind that 8-bit ISA slot, the board designers didn't know what they were doing. Which was probably common, I'd wager.
Thanks for the update on the CGA Improper mode. And yes, I too have been annoyed by slots in line with components which jam against full length cards. This board was interesting precisely because it allowed the full length card to be inserted and also worked with a CGA card as primary video device, including BIOS support for it. It was also very interesting to be able to adjust the speed of the ISA bus in the BIOS. I have very many late 486 boards, and I believe it is the only one with these features.
Grand Prix Circuit has a conceptual framerate limit or something. If you use the successor game, Grand Prix: The Cycles, you get an improvement in framerate/smoother gameplay.
one drive in the chain must be terminated - it is PRE-Standard IDE and most an MFM-Controller, right ? look in the Techsheet of the Drive and you will find the TERMINATION RESISTORs
Lucky me! OLDest x86 is a 1994 Tulip Socket 3! 1981? hmm, can't get any closer than my C64 'breadbox' with Grey Fs...main use = 1987 to early 1993. (BUT still alive!)
Nice! I have a few C64's lying around, but I need to fix two of them and don't find much time for playing on them. I think my oldest computer so far is a TI 99/4A.
So I watched over half of this and I guess I missed why you shouldn't put a 1981 graphics card in a 1994 PC except for some CPU speed issues? It looks like it worked just fine except for the required tinkering for the bus speed. Is there some reason I couldn't put a Hercules in any ISA slot and use it with appropriate software on any PC?
No those are fixed. You can change the background colours or use a hacked up text mode to get all 16 colours at once, but other than that there's not much else unless you are prepared to use a composite monitor.
Possibly you are talking about Zaxxon? You mean the one where you are flying over a flat platform that scrolls diagonally down the screen. That's Zaxxon (so-called because it is axonometric projection.)
@@PCRetroTech Nah, lol, back when I had a 486 that used this type of board. It looked exactly like this one only it had 4 four 30 pin simms, and two 72 pin simms. But it had that single 8 bit slot exactly like that. Board makers did that for real estate reasons, in this case a chip of some type. (If they even added the 8th slot, by the end of AT most did not) While you could put an 8 bit video card in it. That wasn't its primary purpose. There were a lot of 8 bit ISA boards back then. So back in the mid 90s that slot is where my 14.4K Internal Modem lived. But there were a lot of other 8 bit cards that were short that could fit in that slot and be out of the way.
Hi There is limitation to the ISA bus, it does not always run at 4.0mhz, they are fairly robust and can be pushed and run up to about 20mhz This is what the clone market done, as they pushed the XT into the Turbo XT territory, they also found tweaking the bus helped, as you mentioned no good having the cpu doing 8/12/16/20 mhz is the bus is fixed at the original slow speed. As for software / games. The original PC/XT were all the same speed 4.7mhz, so you just coded, there was no need to think of CPU speed With XT Turbo, AT / 386 days, old programs were just running to fast - like your zaxxon, the game has done all the clock cycle calculations and yes you were to slow on the keyboard, all those non-movements from you means crash Try playing space invaders or frogger, you are dead by the first or second keypress So programmers wrote some test code, either hardware probing or some nested loops and get the elapsed time / clock ticks taken - then factor that as a percentage of the XT, this gives you some software wait states This is where turbo C and turbo program programs have that divide by zero crash, as the nested loops on newer machines are done under 1 second, as a integer it is divide by zero There are some good old tools and test programs for video cards and hardware rating, no 2 give the same result :) i find nortons sys info ( version 5 ?? ) fairly good Regards George
Hi George, thanks for your comments. Indeed, some games use the CRT frames to time things and they behave much more reliably across different machines. The ones that just go flat out will simply run too fast on later computers. It's quite interesting to find the games that work well on the later machines.
That pronunciation is subjective and depends what country you are from. Wikipedia originally had your pronunciation, but it was challenged and removed ages ago.
The T.Res chip is actually a resistor network. Termination Resistor is what the name stands for. It only needs to be on the last drive in the chain, and also is only needed with some early floppy controllers. Newer controllers terminate the bus on the controller card.
Ah that will explain why everything still works. Thanks!
@@PCRetroTech Yeah, the termination at each end is to reduce reflection when the signal reaches the end of the cable - it behaves like a "transmission line" so the closer you match the impedance the less reflection you get.
BUT! floppy cables runs at such low enough speeds that the reflections really doesn't matter much unless you have very long cables (say 3m+), so most? all? PC floppy controller cards provides sufficient termination that it'll work with or without the other end being terminated.
More people are familiar with termination-at-each-end for Parallel SCSI cabling, getting this right there is much more important since the signalling is 10-400 times faster than for floppy cables so the reflections are much more likely to cause issues. But it's still possible to sometimes get away with "incorrect termination" for short parallel SCSI cables and low (by SCSI standard) speeds, so it's logical that floppy cables are even more resilient.
@@Torbjorn.Lindgren That totally makes sense. I'm even vaguely remembering a video I saw where this was mentioned once, now that you've reminded me of the details. Anyway, thanks for the comments.
OHH, that's important to know; thanks for pointing it out. Turns out this info is actually on minus zero degrees too; look for diskette/5.25_floppy_drive_termination.jpg.
It appears that the T.Res is on the wrong drive in his system! It is supposed to be installed only in the last drive in the daisy chain, the one at the end of the cable. Those drives from IBM have a DIP network, but some other drives used a SIP instead or a single jumper that you installed or removed. On a drive with the single jumper the resistor network was always there; the jumper lifted the ground connection to it, taking it out of the circuit.
The RED on the IBM Monitor looks really good.
Yes, I've noticed that.
Games that were designed to run on a wide variety of cpu classes hook the PC's interval timer circuit and use it to synchronize the screen updates. Thus you can have consistent game play across machines of different performance. Grand Prix Circuit is one of these games. Once you have a CPU that exceeds the computation power required to play the game without bogging down it spends time in an idle loop until it is required to generate the next frame.
Yes, that's correct. The other way it can be done is to use the vertical retrace of the screen, which is usually 60Hz give or take on most older machines.
is this inside a memory disambiguation chip? because CPU would need to dump a buffer IC about store to load forwarding the "clock frame"
I think this is the reason we'll never see such technical review on youtube to get tenths thousands of views
@@dgmt1 For overall game speed maybe, but a lot of those games still had other unforeseen timing issues, such super fast animations, physics bugs or trying to access the sound card too fast. There were still some big name 90s games that were straight up speed sensitive too, such as Ultima VII and XCOM.
@@ozzyp97 Same on other platforms. I once tried playing an old Mac game (some Breakout clone, can't remember the name, I'd say early 90s) that ran fine on a IIcx with its 68030 on a Power Mac and it was absolutely unplayable, the ball moved lightning fast.
the bad adjustable colorburst cap could have caused excessive noise back to the CPU clock line, IBM PC and XT boards both used a 14.318 mhz crystal and I -think- it was divided down for various functions in the machine including the CPU clock IIRC.
Thank you for the awesome video again. I just acquired an IBM PC 5160 and all repairs regarding that era of hardware is very interesting indeed.
I like this era simply because you can repair them. In later machines one simply pulls the entire board out and puts a new one in, which just doesn't seem as interesting. I'm glad you are having fun with a 5160. It's a nice machine to own.
Love the video and the topics you explore!
Unfortunately, you picked one of the worst games to showcase CGA RAM speed: Grand Prix Circuit is internally framerate-locked to 10fps, so it looks the same on any 286 or higher. A much better game to test how CGA RAM speed affects gameplay is Jet 2.0, which updates the screen with REP MOVSW and can run as high as 60fps on fast hardware. Otherwise, using a benchmark program like you did in the video is the only way to reliably measure speed changes.
It is difficult to find games that can reach 60fps *and* support CGA. This is because, in the era of programming for CGA, it was impossible to reach such heights, so programmers cut corners that would never cause an issue with speeds of the time, like lock the internal framerate calcs to a low rate (10 Hz, as seen in Grand Prix Circuit). Or worse, not lock the framerate at all, so you see some games become unplayable on CGA on faster systems whereas the same game works fine on EGA because the timing is locked to the page flipping rate of 60Hz; Stunt Driver is one such game that comes to mind.
Wow 10fps, really. I felt really sure it was much higher than that. It seemed butter smooth in person. And I feel like I can usually tell framerates up to about 12fps. I'm not at my computers right now but when I get back to them I think I have to check that out to figure out how I fooled myself so easily. Probably I was just focused on the video so much I didn't even notice. It's definitely a shame I made such a bad choice.
Maybe FastDoom would be a good option to test with nowadays. It supports CGA now IIRC.
Also: can you query vertical refresh on the CGA? I only know how to do that on VGA.
@@root42 You can, using the same port read mechanism.
@@JimLeonard awesome. Thanks for the clarification.
Its good to know a 486 board still works with CGA card. I have a card that needs to be tested so will give it a whirl.
Lets take a moment to appreciate all the cool hardware in the background!!! The kitchen workshop is fantastic Will!
It sure is fun having lots of parts to play with. That's an advantage of where I live. Lots of cheap old parts.
That variable capacitor (not a variable resistor in this case, btw) was most likely used to tweak the signal coming from the crystal next to it. I assume the crystal is related to the system bus clock, and that could explain potential ISA and floppy drive issues.
Yes, definitely. I can't believe I said variable resistor and not capacitor. I guess when it fell to pieces it actually did look a lot like one of those carbon pots, but of course that makes no sense given what it is doing. I suppose it could have been part of an LCR network or something, but that would be a stretch.
Yes the composite on CGA iir is derived from the system clock. The variable cap is used to dial the system clock into spec. You can tweak that cap and watch your composite video lose color burst on the cga card. Since I believe its tweaking the system clock in real time it would do funny things to bus timing thus parallel port timing. I have a generic Turbo XT that was giving me only black and white off the composite out on the cga card and needed a tweak to get color back.
@@joshhiner729 Yes, I think the cap was partly broken when I adjusted it. I made a video with the 8088MPH demo in it on a Commodore 1084s monitor and I adjusted it to get colour composite output. Apparently I overdid it and broke the cap completely.
I read that this IBM 360K floppy T-RES terminator resistor chip should be installed on the last floppy drive in the chain and it works just like SCSI bus termination, to prevent the reflection of electrical signals from the end of the cable in order to ensure reliable operation. Cheers.
Interesting. I better check it is actually the last one in the chain!
2:45 "black & red = dead" even when I wasn't so fluently speaking English back then, we've been teached this rhyme back then when installing hardware. Before ATX came out.
Thanks for your content!
Yes, I've heard that on Adrian Black's channel I think.
@@PCRetroTech Yes, he said that, too!
It's actually "Red and red and you're dead". If you think about it, you can't have black and red in the center. It's either black and black, or red and red.
I suspect that the 1/3 option in the bios is not really working and it's actually still 1/4. You could test this by meassuring the frequency directly from the ISA slot (pin B20), if you have a multimeter that can do that high a frequency, or an osciloscope.
That's possible, though there could be an issue with CGA wait states that makes it seem that way, as another commenter mentioned below.
About the dip package that one of the drives lacks, if I remember correctly, old disk drives didn't use the twisted cable to identify the two drives connected to the same cable, so I guess the component is a terminator.
That seems to be the case. Thanks for the comment!
No it's to terminate the bus. The drivers are in fact both configured as drive 1 with a twist on the cable, but the floppy controller card needs the termination.
@@booboo699254 Oh I misread that bit. Yes, the cable has a twist.
Nice to see the IBM working well again! The 486+CGA combination was something I used a lot back in the day, on a VGA card though... I had a black and white monitor and the strong colors helped with the contrast in games!
Yes, I think quite a few people will have done it with a VGA card or something backwards compatible with CGA. It was nice to find I could put a genuine CGA card in this particular motherboard. It made a fun experiment.
My first "proper" graphics card was a Cirrus Logic 5429, a VESA "Windows Accelerator" with a whole 1MB of memory. I was so proud of that powerhouse of a GPU.
I think I still don't have a 5429 in my collection. I've got a 5428, but the 5429 seems to be a bit rarer.
@@PCRetroTech Unfortunately mine is gone; it was in my 486 which I gave to my mum, then (without my knowledge) my sister organised a new PC for her and the 486 went to the tip (erm, recycling centre). If I'd known I'd have taken it back of course. Sigh.
I remember on Windows 3.11 the drivers were a bit buggy with some screen artefacts adding random dots sometimes. Worked perfectly on Win95 though.
@@ian_b That's a shame. A lot of cards ended up in landfill I bet. But I suppose the rarity these days is part of what makes retrocomputing and collecting fun.
@@PCRetroTech It is sad yes. I had a sort of perception of future retro value at the time, which is why I hung onto my IBM Model 60 (a huge PS/2 tower with a 286 processor), but even I foolishly sent the IBM monitor to the recycling centre, to save a modicum of storage space. I didn't think a 12" monochrome CRT was worth keeping; nowadays of course I wish I still had it.
Oh well, live and learn.
@@ian_b Yeah, I've also thrown away stuff that I didn't have space for back in the day, or which just didn't seem to be worth anything any more. Funny how value changes.
Man, I used to play with Fractint on my 486 back in the 90's. So many memories.
It was one of my favourite programs as a kid. I even tried to write a program to draw the Mandelbrot set myself. Those were fun days!
This is quickly becoming one of my fav channels, great video!
Wow, thanks!
@@PCRetroTech i like the new magazine format and doing fun deep dive techy things with vintage PCs! I built my own 8088 machine, the XTjr so I learned a lot about the architecture and got kinda obsessed. If you ever do Patreon/Membership let us know as be happy to throw you a few bucks if it helps.
@@mogwaay Thanks. I'm still looking into a Patreon. I might be able to do that next year or so.
When my VGA monitor died, I used for a short time (few weeks I think) a original Hercules adapter. I also had a CGA card but no monitor to pair with. Luckily the Hercules I had was paired with a beautiful amber mono screen.
My first shock was when I entered the BIOS setup. My motherboard had winbios (graphical UI) but it was unrecognizable with that Hercules; it was displayed in a strange text mode. The next surprise was that Win95 was not able to boot with that card. So i took another HDD with ms-dos to use. The time I spent with it, was entertaining, it was like a lottery to run programs; would it run? and if so, how will it look? I also used a CGA emulator for Hercules that allowed to run CGA games on that mono card. I also remember that the only GUI OS that worked with Hercules was Red Hat Linux (ver 7 I think it was) that I installed and ran fine with X server GUI only. Win 3.1 or 3.11 would not run in mono, only Win 3.0 was able to. The only drawback was that I had no idea of Linux and no manuals available, so no way to learn how to do something useful with it. So I returned to Windows as soon as I got a new monitor.
That's a pretty funny story. I guess at least you managed to learn about which software would and would not work with Hercules. It's amazing how old technology just stops being useful at some point, even though one of the major selling points of a lot of PCs is that they are supposedly "future proof".
I always love this kind of mix of non-era part. Way back I ran an ATi Small Wonder with an amber monochrome in a Pentium 3 board. It ran as a Linux server for a while. Later I used it as a networked music player using mpxplay and a dos smb client.
I think mixing parts like this can be fun. Part of retrocomputing is about experiencing things as they were back in the day for nostalgia, and part of it is trying new things that we couldn't try back in the day. Both are fun.
@@PCRetroTech I run an ET4000 in my 5150 just so I could have VGA on it.
T-RES stands for terminating resistor and it is used to tell the PC bus where the last drive on that control line is. So if you have two floppies, if memory serves, the resistor block should be on the B drive. If you have an XT with one floppy drive then the only floppy drive, A, should have it installed. Hope this helps.
Thanks. That is what I am getting from others too. So I think there is a concensus on this now.
I like your new format. Please keep up the great content!
I should have done it sooner!
To my knowledge, CGA either Black White Cyan & Magenta, or Black Yellow Red & Green colour pallettes, how was the game at 4:24 showing Black White Cyan and RED? How was it showing that combination of colours? Also the demo with spinning shapes is more than 4 colours, how it that possible?
There was an undocumented CGA palette. Wikipedia has a description of this. As for the demo that appeared to have more than 4 colours, it is an illusion. When you set the background colour to something other than black and then alternate pixel colours, you get the appearance of more than 4 colours. That demo was written by me and I have a video on the channel about how it was done.
You may need to move that T-Res chip to the B: drive if you want to continue using it. Although it may not be making much difference yet, that chip, which is a set of 150 ohm (or somewhere about) resistors is used for terminating the floppy drive cable and prevent signals from bouncing around. That may also have something to do with issues you've had out of it in the past. Check the Shugardt floppy drive standard for specifications.
That CPU install blew my mind. Those pins and not needing a cooler
A DX4 should have at least a heatsink. I kind of cringed when he installed it without one 😬
My 286 PC has some similar issues to your 5150 where programs/games will just crash constantly in random places, seemingly even more so if you have more ISA cards installed. Really wish I had the skills to identify the issue and fix it, there's no RAM issues detected or anything like that when I run something like PC-CHECK.
I got really lucky here I think. If it hadn't been so simple I guess I might not even have gotten a video made about it.
I believe the reason why you only get a few different benchmark results on the CGA card is the wait states. The algorithm (implemented by the discrete logic on the CGA card) is basically: 1. wait 1 hchar, 2. wait for the next lchar boundary, 3. wait for the next CPU cycle. Step 3 will be negligible on a 486 so how fast you can access CGA RAM is limited by how well the access pattern of the benchmark code fits the CGA wait state algorithm. With a different benchmark you might get different results as to which set of FSB divisors give the same performance as each other.
Interesting. I hadn't thought of the wait states being a limitation. That could well be it. Thanks.
Exactly. The CGA card runs at a dot clock of 14.318 MHz in text mode. A character is 8 pixels, so the CGA card runs at 1.79 megachars per second. You should get a video memory access time slot per char, so if your write cycles are short enough and in quick enough succession, you should be able to fill the video RAM at 1.79 MB/s. That's obviously way faster than the IBM PC could handle. If the processor could saturate the bus doing block fills (spoiler: The 8088 can't, the V20 can), the FSB would allow 1.19MB/s, so you would miss every other opportunity to access video RAM and lock in at half the theoretical 1.79 MB/s, i.e. 895 KB/s.
The 486 chipset throws in a lot of wait states for 8 bit memory access, so it obviously misses the 895KB/s rate and instead only uses every third chance to access the video memory. This results in a theoretical rate of 597KB/s. The lower actual rate (583KB/s) might be due to RAM refresh cycles. They need to be visible on the ISA bus for compatibilty reasons, as some cards refresh their local memory based on ISA refresh cycles (mostly ISA memory cards).
Some mainboards allow changing the number of wait states for 8-bit cycles using the BIOS setup, but that would be in the "advanced chipset options", and obviously you BIOS doesn't allow it. I checked the SiS471 configuration file for CTCHIPZ, and it seems waitstates for 8-bit cycles are software configurable as 4, 7, 11 and 16 ISA clocks. As I don't know what's the base line (i.e. how many clocks "0 waitstates" is supposed to take), I can't directly translate that to data rates. What we do know, though, is that at 16MHz ISA clock, you are close to "every other cycle" instead of locking in at "every third cycle", which you would lock in at 11MHz and 8MHz. We know you can't run at 16 extra bus clocks, because 16 clocks per cycle would mean a theoretical maximum of 500KB/s at 8MHz, which your board exceeds. Getting 600KB/s at 8MHz would require no more than 13 bus clocks per ISA cycle. Not getting to 808 KB/s at 11 MHz also indicates a likely total time of 13 bus clocks, so your BIOS likely sets up the ISA bridge at 11 "recovery clocks" that get added to the AT minimum bus cycle duration of 2 clocks.
Setting your chipset to 4 bus clock periods recovery time should allow you to reach the extremely mad data rate of 1.7MB/s on a classic CGA card. Get CTCHIPZ here: www.rumil.de/uwsysuti.html
@@tw11tube The wait state problem is why OPL3 and Adlib cards sometimes play garbage noise on a faster machine. MIDI setup also can get goofy as well, particularly the MPU-401 systems.
Check with modern vga if it goes faster, if not, the game has fps control, or the speed of the physics in the game are not related to the rendering process, they are independent. This is done in games mostly playable on a network.
Loved the CGA demonstration! Great work here. Thanks for sharing!!
Glad you enjoyed it!
@@PCRetroTech I'm going to tweet it out to my audience, they will love it!!!
@@RetroTechChris Thanks very much!
On the answer to the question "Why" there is an 8 bit port.. The simple answer is: BEFORE everything was integrated in the motherboard, you really needed those expansion slots.. plenty of them... and.. 1. Cost.. and 2. Motherboard real estate.
Well.. the CGA is ONE card that actually DEMANDS it, because of the form-factor going down so a 16 bit slot would be in the way, but it wasn't alone. .. and.. if you buy a card (with special functionality) for one computer.. you pretty much don't want to be FORCED to give it up for your new computer. F.ex. I have a blind relative that had a speech-synthesis specific card in his computer. That was a full-length 8 bit card.. It cost a shitload... and I don't even think they ever made it in a 16 bit version, because suddenly people just had to have a sound-card and with a Soundblaster 16- and with the speed of computers, and available memory: By the time you had those- it was all done in software.
.. but there was also a shitload of 8 bit cards that.. TECHNICALLY could've been mounted in a 16 bit port, but they still were only the 8 bit cards. Most people think of the ps/2 mouse as standard. That was not yet the standard on 486. 486:es were still being created when it finally DID become standard so newer motherboards may have a couple of ps/2 ports for keyboard and mouse, but most people had to settle for something else. Serial mice?.. they actually grew in popularity fairly late, so it was VERY common that the people who manufactured mice sold them together with a dedicated isa-card for them.. my first mouse was of the brand "Mighty Mouse", and it was such a bundle which had to come with it's own card and special drivers. LATER serial port mice became popular, but tbh.. If you had a modem (almost ALL modems during the 486 era were external and connected through a serial port.. I SUSPECT due to the interference problems)..
Then I also got an MM-401 (mpu-401 compatible) midi-card. The thing is; We had a Roland Sound-canvas module.. I used it with my organ, OR with my computer. It was the second revision which is when it finally came with a MT-32 sound-set that you could switch to if you wanted to play any actual game.. While people were listening to FM-music while playing games, I had a full symphony orchestra while playing Might and Magic 3, or Sierra games.. (If memory serves; Even doom did support Roland sound).. Of course one wanted a soundblaster anyway.. or more exactly: Also.. because it played recorded sounds, so all the sound-effects in most games.. So.. Just sound took two slots..
Then of course.. there was the advent of CD-rom drives for your computer.. hint: They were NOT IDE compatible at the time.. you needed a special card to be able to use it... Since the actual transfer speed was 150kb/s.. you didn't need much bandwidth: Save money by only supporting the 8 bit bus. That said: A LOT of sound-cards had various CD-rom interfaces on them. I bought a bundle from Creative-labs, so I had a Soundblaster that ONLY had the ONE interface, but it WAS the right one (fortunately; I know people who bought the card and the drive separately, so DESPITE there being one or even as many as THREE different interface ports on the sound-card they STILL had to mount a separate card for their cd-rom drive).. Of course.. I was NOT helped by the fact that my soundcard supported it. Not much anyway, because I needed my CD drive to be external. Which WAS what the bundle was for, BUT.. it came with an extra card in order to move the soundblaster pro's internal port to an external connector.
While we still were using a 286 we also had gotten a VGA card.. more exactly an Ultra-VGA card.. same amount of memory, but supported some extra modes.. most notably 256 colours in 640x400 mode (Our windows looked better than anyone else's and they had 386:es, so we really could make use of coreldraw).. And that was of course also an 8 bit card... the 640x400x8bit was of course nothing to play games on, but still.. we kept that card for a LONG time.. It was when we updated our first 486sx25 to a 486dx33 we finally got a trident 16 bit card.
So.. cards I had in the same 486 that was 8 bits: Graphics, MIDI-card, Mouse-controller card, CD-Rom controller, and the serial/parallell port cards. I THINK The IDE was 16 bit, though.. while we had a 386 we were still using our harddrives from when we had a 286 and they were both MFM-drives and the controller for that was 8 bit.. Just to say; We had started with a 20 Meg drive.. then bought a 40 meg drive.. and kept the 20 meg drive.. BUT.. then we bought a different controller that used a different scheme for packing the data, so we literally had one 30 meg and one 60 meg drive.. "for free".. where the biggest drive anyone could have if they intentionally splurged was 120 meg.. but those two drives died after a while.
So.. yes.. During the era of 486.. I DARE say that pretty much 100% of people had AT LEAST one 8 bit card in their computer.. AND a lot of people had many cards in their computer.. so.. doing ONE card so it CAN handle odd format 8 bit cards wouldn't even be a "lost port" for the people who didn't use a CGA or some other card that literally required such port, because it would still just be just another port that wouldn't be in use.
Yeah the CGA card was special in that it wouldn't go in a 16 bit slot, but most other 8 bit cards would. And it is uncommon to have an 8 bit only slot on a 486. But yes, there were lots of 8 bit cards still around for sure.
@@PCRetroTech
@@rogerlundstrom6926 I don't know why the discrepancy, but on theretroweb there are 669 socket 3 boards identified, only 91 have an 8 bit ISA slot. I have a large box of 486 boards. Only one of mine had an 8 bit slot.
@@PCRetroTech .. Well.. The discrepancy can be one of many.. ONE is that the amount of models itself doesn't show anything about how things are in the wild..
F.ex.. I once saw a 486-motherboard.. with only two 16-bit slots.. which DIDN'T have IDE nor ports on the motherboard.. so.. one multi-IO-card, and one graphics-card.. then.. you couldn't add anything. I assume it was cheap at purchase. It didn't seem to be brand name, which is why I THINK it was intended to be used in computers sold to people who only needed to provide a lot of users with a lot of computers at a cheap price, such as poorly funded schools, or extra cheap corporations.. Though it MIGHT have been that the motherboard was created for a specific case, where you were supposed to put in 2 16-bit bus-risers where the extra slots were in different directions.. That may be TECHNICALLY possible, but most corporations that does this even back then tended to use their own designed edge-connector instead of a 16-bit-bus.. So whatever it was.. I bet it was only sold in very very small numbers, but if using "number of models" as a metrics.. well.. IF that mother-board was in the list.. It would've counted as "1" just as.. actually USEFUL motherboards.
It isn't unusual that there are local deviations that MAY depend on specifics such as what is popular in one culture, or even depending on one specific importer or manufacturer that exist in the nation in question. I live in Sweden, you don't, therefore.. It's kind of like how the local travel agent tries to find deals and then try to make it sweeter by negotiating, meaning that people living in one city are fairly likely to meet each-other when they go to vacation in a foreign nation. Things I can mention as examples is that I have never ever seen a Tandy computer in real life, so whenever I see 3-voice pc-speaker as a choice I gawk.
and.. as I pointed out; There was a financial advantage if you could get away with only sending an 8 bit bus instead of going full 16 bit, so it's FAIRLY likely that a motherboard that has 3 16-bit slots and 1 8-bit slot was cheaper than if you had a motherboard that had 4 16-bit slots.. and.. in "pre-made" computers the person BUYING the computer won't even expect it to be cheaper, so the people actually buying the motherboards simply make more money by choosing the cheaper option if they don't even have a reason to buy a more expensive version, and this kind of is even MORE cost-effective if you yourself IS the manufacturer of the board.. and.. when selling to end users; Unless they have a REASON to want more 16-bit options, they would gladly settle for a board that has ~enough~ 16 bits option and put whatever peripheral-board they had (of which they most likely had at least one) in the 8-bit slot... This isn't exactly an argument that "shows" that it is so, but at the very least it show that the basic rule of "The consumer" and "marketing" means it makes sense... only people who doesn't have a single board that is 16 bit may see a reason to ~intentionally~ look for a motherboard without an 8-bit slot...
Some of the new retro games that run under DOS target the improbable combination of 486 with CGA (e.g. Castle of Viana). Now you have the ideal system for them! It was nice to see the IBM back to being a happy computer in the second half. Hopefully it will get its voice back soon.
Nice! I hadn't seen Castle of Viana, and that's precisely the sort of thing that will make an interesting follow up. And I do hope to fix the sound issue in a later video too. Thanks for the comment!
Are there any other games that do this? Why would anyone write a 486+CGA game?
@@michalzustak8846 It's just not optimised well. People write stuff for emulators, which are 386 minimum and they do things the easy way, e.g. BIOS routines and high level languages instead of doing assembly language and the like.
@@PCRetroTech Even say DOSbox set to 400 cycles which is as slow as a 4.77 Mhz PC?
@@michalzustak8846 Of course you can slow it down, but then these games are not playable. Viana requires about a 486 75MHz to be playable because it is not optimised. It was written as part of the CGA Game Jam competition where there is very limited time to write the game, so certainly not enough time to optimise the code.
I'm sure you already know but the mystery "chip" is a terminator block.
termination resistor
I have a retro computer that is rather a Mac Plus than an IBM with a similar symptoms in the built-in display screen, which maybe might just have had some dirt between solder contacts of the HV output to the CRT causing it to ripple and flicker. I’ve never examined it internally, in spite that there was no compromise the last time it was used.
Could be worth a try. Thanks!
There is a reason the "TURBO" button existed on the PC/AT computers. I'm sure this board had the turbo button connection on the header.
It did have a turbo jumper.
Looks like that floppy chip in question would be a terminating resistor pack. Likely should be installed on the last drive in the chain. Maybe early floppy setups were noisy and needed a t-res? Ive never heard of PC's needing a terminating resistor on a floppy chain as a requirement.
Yeah it was a surprise to me too.
I have two of those lovely 486 boards with pci, one is bricked, from a short circuit on the back. If someone wants to have it for spares let me know 😂
I think the variable capacitor adjusts the OSC signal on the bus, which can be used for all sorts of things, including affecting the DRAM refresh timing on some motherboards. That could explain why you were getting what looked like ram faults in addition to weird behaviour from expansion cards making use of that signal.
Yes, I suspect it just glitches a whole bunch of stuff.
It was added to adjust the chroma signal output by composite CGA, which is locked to the system crystal. This explains why the video system was affected when it was loose and jostling aroung when the side of the system unit was smacked. Also see Geoff Reed's comment below.
@@JimLeonard Yeah this much I knew, since I had previously adjusted it to show 8088MPH. I've also talked to Reenigne a few times about it.
I took the embarrassing section out of the video where I called it a variable resistor. But unfortunately I can't cut the multiple places where I call it a pot. But In a later video I'll make a section with viewer comments and clear that up. I had always thought it was a potentiometer, thought that was very odd, and even had go through my mind when making the video "I wonder if it is a variable capacitor", then forgot to check. So this one was a self-inflicted error.
@@PCRetroTech The important thing is that you restored a PC to working order! (I've always called it a pot too)
@@JimLeonard Yes, I am pretty happy about that.
Hi, nice video. About the CGA snow effect, i know it's due to the contemporary access to video memory by cpu and cga.
I think it's because while the cpu is accessing the memory the cga get bad readings or it is blocked from fetching display data.
I also know that not all CGAs had this problem. Do you know how they fixed the issue in snowless cga cards?
I believe they used dual ported RAM or SRAM or just faster RAM. I've not looked into it in detail, but I think there were multiple strategies.
One (very late) comment on the racing game. It appears to use a graphics technique sometimes called beam racing. In order to horizontally scroll different regions of the screen at different rates, the game will update the frame buffer pointer register in the 6845 CRT controller at different points in the frame. This must be done continuously at the frame rate and with consistent timing during the relevant line blanking interval in order to ovoid very obvious jitter. This would probably place an upper limit on the game performance unless it was recoded for faster CPUs. If interested, UA-camr James Sharman has a video dedicated to this topic in his VGA From Scratch mini-series (an amazing DIY hardware and software project).
Grand Prix Circuit uses some tricks to make it easier to manage the sprites when they run off the bottom. There are two mirrors on the car. Those are the only regions where an oncoming car sprite can pass. If it intersects any other part of the dash, a crash occurs. That gives a hint that they draw the sprite for the oncoming car and then cover it up with the mirror, which has to be redrawn every game frame (game frames are a lot less often than CRT frames).
Probably when the oncoming cars need to be drawn, they just wait for vertical retrace and draw those sprites first so that they are drawn in time. That is a kind of 'racing the beam', but not a very sophisticated kind. Most games could not use very sophisticated tricks due to the large variation in CGA hardware and PCs on which it had to run.
You'll find a video on my channel where I analyse this game frame by frame. I think some of my conclusions are actually incorrect, as I was anticipating that they did something more sophisticated than they really did. But I did manage to recreate part of the game myself, albeit not the part we are talking about. My engine was many times faster, and nowadays I could do much better again. So it's certainly not a particularly sophisticated game engine.
@PCRetroTech Interesting, thanks for taking the time to reply. I'll look up your other video.
Maybe that potentiometer had shorted something. Or Maybe it caused interference through capacitance.
Apparently it is a variable capacitor and when it broke it sent spikes through the tuned circuit it was part of whenever touched.
@@PCRetroTech a variable capacitor? Never seen one of those before! They must be quite exotic parts!
@@JohnSmith-iu8cj Not really. They were pretty common in the before the 90's at least.
As I had a 486SX back then, I wasn't surprised.
why do people always use those old CRTs with shrunk down picture? why isnt it streched to fill whole screen?
Largely because the resolution was taken from the NTSC television standard (262 scanlines per CRTC frame) and because if you want to use more pixels you need more video RAM. If you wanted to use more pixels you'd need to increase the video RAM from 16kb to 32kb and that would have been extremely expensive back in the day for just a few more pixels.
You could set up the CRTC controller to use more of the pixels so long as you reuse video RAM more than once for part of the image, but this is difficult to do on the fly. Some demos like 8088MPH and Area5150 do this.
0:10 Your hand shows the older 8/16-bit *emulated* PC/AT-bus (black slots) as well as the never 32-bit bus (brown slots) going *directly* to the 486 CPU. Cards normally use either one of these, not both.
The 8 bit slots have loads of uses - there are other expansion cards other than video cards you can plug into them! Scanner cards, i/o cards, modems, specialist interface cards, etc, and why would you want to waste a 16bit slot on an 8bit card!
Don't you think that some games would sync to the frame rate, and thus not run "too fast"? They might do better, finding they have more time for computation, but still present the game world at the proper speed.
For all but the most trivial games, I doubt that it's just looping as fast as it can and thus the write speed of the CGA memory and registers bring it back to original speed.
Terminating resistor package.
Thanks.
What would be funny is to find the newest computer you can that would take that card. You might need a slot extender to make it fit onto newer boards. I wonder if it can be made to work on PCI only boards using a PCI to ISA adaptor ? if that exists
Interesting idea. Even with an adapter it would still be restricted to ISA bus speed. The memory on the card would have a limit on performance also.
@@PCRetroTech Yep that's why I think the idea would be really fun to explore. Having such a limited card (in bus speed, capabilities) and trying it on faster and faster machines; how far can you go!?
_Before I say anything else, a brief apology to PCRetroTech for leaving you hanging on the CGA improper on stock thing: It likely (my current guess: 95%) would not after all work as I expected, though there might still be an outside chance (5%). I've not followed up and not gotten back to you because I got sidetracked and overwhelmed, including mentally, and I guess that latter part now includes that there's a war - almost beyond comprehension. In principle I still would like to get back to you on that one day, but in the meantime you're probably reasonably safe in assuming I was once again talking nonsense._
On plugging stock CGA into newer boards: Been there, done that, didn't know about the whole what's clocking what when I did it. The interesting thing is that at least well into the nineties and noughties, boards routinely used to have that one 8-bit ISA slot (and I wouldn't exclude the possibility that perhaps some ISA-compatible boards still do), though I'm not sure the board designers always understood why.
That's because when I last plugged my IBM CGA into a maybe 386 or 486/Pentium board (hard to remember which), there were a bunch of ICs and various sufficiently tall components behind the 8-bit ISA slot that actually physically blocked things, so the CGA could not be fitted into that slot. What I ended up doing was, I cut out the very back at the end of one of the 16-bit slots, and then I used some simple insulating material, it might have been just a piece of paper, and wrapped that around the rear bottom edge of the CGA card, so I could physically push the card into both the non-insulated 8-bit part as well as the insulated 16-bit part of that slot - so that it fit but nothing in the 16-bit extension slot made contact. It wasn't pretty, it required cutting out the back of that 16-bit slot, but it worked. There was nothing else in the way behind that particular 16-bit ISA slot.
I don't think the single 8-bit-only ISA slot that routinely was on boards back then was there for any other reason than to accommodate large cards whose PCB dimensions resembled that original CGA. But I think including one 8-bit ISA slot was one of those things everybody was doing, but not everybody who was doing it remembered why they were still doing it, and that resulted in mainboards whose inclusion of an 8-bit ISA slot was kind of useless -- like the board I had. I guess you could still justify that 8-bit slot as "Hey, you can fit a short ISA card in here without the slot taking up more space on the board", but then, why not just add another PCI slot instead, because we're prolly well into that era at that point?
It's a nice little litmus test: If there isn't a "firebreak" of only low-profile components behind that 8-bit ISA slot, the board designers didn't know what they were doing. Which was probably common, I'd wager.
Thanks for the update on the CGA Improper mode. And yes, I too have been annoyed by slots in line with components which jam against full length cards. This board was interesting precisely because it allowed the full length card to be inserted and also worked with a CGA card as primary video device, including BIOS support for it. It was also very interesting to be able to adjust the speed of the ISA bus in the BIOS.
I have very many late 486 boards, and I believe it is the only one with these features.
Grand Prix Circuit has a conceptual framerate limit or something. If you use the successor game, Grand Prix: The Cycles, you get an improvement in framerate/smoother gameplay.
chip on the floppy is resistor network (terminator)
one drive in the chain must be terminated - it is PRE-Standard IDE and most an MFM-Controller, right ?
look in the Techsheet of the Drive and you will find the TERMINATION RESISTORs
Will it worck in a soket 7 theoreticly
I would imagine so if it had an 8 bit ISA slot. But I've not seen a board that has one.
Lucky me! OLDest x86 is a 1994 Tulip Socket 3!
1981? hmm, can't get any closer than my C64 'breadbox' with Grey Fs...main use = 1987 to early 1993. (BUT still alive!)
Nice! I have a few C64's lying around, but I need to fix two of them and don't find much time for playing on them. I think my oldest computer so far is a TI 99/4A.
So I watched over half of this and I guess I missed why you shouldn't put a 1981 graphics card in a 1994 PC except for some CPU speed issues? It looks like it worked just fine except for the required tinkering for the bus speed. Is there some reason I couldn't put a Hercules in any ISA slot and use it with appropriate software on any PC?
is there anyway to override the coiors in the palettes?
No those are fixed. You can change the background colours or use a hacked up text mode to get all 16 colours at once, but other than that there's not much else unless you are prepared to use a composite monitor.
My guess would be that the blue chip is a terminator of sorts.
Yep, seems to be.
What is the name of the first game? childhood memories, excellent video.
Possibly you are talking about Zaxxon? You mean the one where you are flying over a flat platform that scrolls diagonally down the screen. That's Zaxxon (so-called because it is axonometric projection.)
So why shouldn't I put a graphics card from 1981 in a PC from 1994 ?
I have a OPTi VLB/ISA board that is from 1998 surprisingly, it was sold as a industrial board. it has the AMI WinBIOS
Yeah, they kept making them pretty late.
@@PCRetroTech 486 had a long lifespan, up until 2007 when they stopped making 486s, same with 386.
@@pentiummmx2294 Yeah a lot of old military hardware use them I think.
@@PCRetroTech industrial and POS systems used 486 as well i think.
I had a board almost like that. I stuck my internal modem in that slot.
You still use an internal modem? What do you connect to?
@@PCRetroTech Nah, lol, back when I had a 486 that used this type of board. It looked exactly like this one only it had 4 four 30 pin simms, and two 72 pin simms. But it had that single 8 bit slot exactly like that. Board makers did that for real estate reasons, in this case a chip of some type. (If they even added the 8th slot, by the end of AT most did not) While you could put an 8 bit video card in it. That wasn't its primary purpose. There were a lot of 8 bit ISA boards back then. So back in the mid 90s that slot is where my 14.4K Internal Modem lived. But there were a lot of other 8 bit cards that were short that could fit in that slot and be out of the way.
@@Alcochaser Yeah I very much doubt that had this particular application in mind. :-)
That's a variable capacitor btw
Doh! Of course it is. How else would it tune the frequency. What a shame I missed that. Thanks!
Hi
There is limitation to the ISA bus, it does not always run at 4.0mhz, they are fairly robust and can be pushed and run up to about 20mhz
This is what the clone market done, as they pushed the XT into the Turbo XT territory, they also found tweaking the bus helped, as you mentioned no good having the cpu doing 8/12/16/20 mhz is the bus is fixed at the original slow speed.
As for software / games. The original PC/XT were all the same speed 4.7mhz, so you just coded, there was no need to think of CPU speed
With XT Turbo, AT / 386 days, old programs were just running to fast - like your zaxxon, the game has done all the clock cycle calculations and yes you were to slow on the keyboard, all those non-movements from you means crash
Try playing space invaders or frogger, you are dead by the first or second keypress
So programmers wrote some test code, either hardware probing or some nested loops and get the elapsed time / clock ticks taken - then factor that as a percentage of the XT, this gives you some software wait states
This is where turbo C and turbo program programs have that divide by zero crash, as the nested loops on newer machines are done under 1 second, as a integer it is divide by zero
There are some good old tools and test programs for video cards and hardware rating, no 2 give the same result :) i find nortons sys info ( version 5 ?? ) fairly good
Regards
George
Hi George, thanks for your comments. Indeed, some games use the CRT frames to time things and they behave much more reliably across different machines. The ones that just go flat out will simply run too fast on later computers. It's quite interesting to find the games that work well on the later machines.
Would the MDA card work?
I didn't try it, but I don't see any reason why not.
Never heard it called a four eight six.
Termination resistor
Turn off cache in BIOS and even zaxon will run fine.
Hey mate.. using a red line on your thumbnail makes people think they already watched
Why on earth would I want to do that!?
YOUR FLOPPY CABLES ARE CONNECTED WRONG, REVERSE THEM.
Always nice to see a video that shows the complete opposite of the video title, just so I get to exercise my downvote muscle.
For the love of god it's "four eighty-six", not "four-eight-six".
That pronunciation is subjective and depends what country you are from. Wikipedia originally had your pronunciation, but it was challenged and removed ages ago.