VOGONS


Copam U-series 486

Topic actions

First post, by aitotat

User metadata
Rank Newbie
Rank
Newbie

uc?export=download&id=1LmVLa5chkHRnmor80RWfKA796H6ZhmN2
I recently found a very little Copam U-series (ultra slim) computer. Only information the seller gave was a photo of BIOS post screen and couple of photos of the computer and what came with it. The seller didn’t have a keyboard so the computer was stuck at battery error on the post screen.
The post screen did gave me some hints about the system. There was SHASTA-F mentioned and that was a very good indication that this was a slow 486 system with Headland chipset and likely without motherboard cache. Turned out I was correct at that assumption.
The computer came with all manuals and disks (driver/utility disks, MS-DOS 5 and Finnish version of Windows 3.1). There were manuals and disks for both 386sx variant as well as 486 variant of Copam U-series. On the photos there was part of manual shown with TVGA8900 written on it. I of course assumed there was some 8900 model Trident integrated but that manual was for the 386sx model (it has 8900C). Also the 386sx motherboard manual was shown and that let me to assume this computer was running at 16, 20 or 25 MHz. But I didn’t know that manual was for 386sx.

So when I finally got the computer (I payed 52 euros knowing only what I described above) it was just as small and cute as I hoped it to be. I opened it first before testing it. It has DS1287 RTC with build-in battery and the motherboard did not have support for external battery. The CPU was Intel 486DX 33Mhz. I assumed it would be SX-25 but that guess went wrong, but not too much. The CPU was underclocked to 25 MHz. The motherboard has jumpers to set the clock to 16/20/25 or 33 MHz. I wonder if they installed wrong CPU at the factory or did they set the jumpers wrong.

The motherboard has 2 MB RAM soldered and then there are two 72-pin SIMM slots that came with 4 MB SIMMs. So it had 10 MB total. This is also the maximum what is supported based on the manual. I did try 8 MB SIMMs without luck so the manual is correct. Based on date codes (late 92), the computer came shipped with 10 MB installed.
The only thing that did not have 92 date codes was hard disk. It is 130 MB Seagate with 13.5-93 sticker on but I think on the hard drive PCB there was 3993. Since battery was dead, the BIOS setup defaults to 40 MB drive. So likely the original hard drive was 40 MB. That sounds small for a 486.

The chipset is Headland Shasta and there are no motherboard cache or even empty sockets. Graphics chip is Headland HT216 and based on manual, it is integrated to VLB and benchmarks indicate this as well. There seems to be 1 MB ram for the HT216 but for some reason most detection programs cannot detect the size. Both the motherboard graphics and system RAM are 80 ns. The SIMMs are 70ns.

Power supply is 65W and it has 40 mm fan. It is loud (but not louder than any other old PSU fans) and I’ll replace it eventually. For now I left it as it is. The computer was very clean inside, no dust at all. But there aren’t many holes for the dust to get in. Ventilation must be poor and the CPU didn’t even have heat sink.

Powering the system gave me some problems. First this seems to be very picky about keyboard. I first tried very old one but it usually gave me errors. So did IBM Model M. There were no problem with Keytronic or later IBM keyboard.

I needed to make battery mod for DS1287 because every time I tried to save BIOS settings nothing was saved and I got the battery error again. So it was not possible to boot because all I could do was enter bios setup. This was first time I did the battery mod and it wasn’t hard at all. I’ve been avoiding motherboard with Dallas RTC but I guess now I’d rather take Dallas than leaking battery. I had two too large coin batteries that I accidently once bought. I directly soldered wires to one of those. I’ll get the usual CR2032 battery holder when those larger ones are used.
uc?export=download&id=1vOB3MQme-LvJM_BRgZWi4FiAZyzw61zN

When I finally got to boot, it turned out that floppy drive was broken! And not only that, two of my spare drives had stopped working so it took a while to fix it. I had one known broken Mitsumi drive that was dirty. I cleaned and lubricated it and got it working. This Mitsumi is likely the quietest floppy drive I’ve had so it is great on a computer that can’t have any 5,25” drives.

I wanted larger and quieter hard drive so I settled to a 20 GB Maxtor. It is not quiet at all but I didn’t have anything more suitable at the moment so I’ll keep this temporarily. The BIOS does not support drives over 512 MB, as expected from 1992 BIOS. I could have used some drive overlay software but instead I installed XT-CF. This way I can have XTIDE Universal BIOS and CF slot at the back of computer to easily transfer data. This computer does not have any optical drive that could have been used for that.
uc?export=download&id=170ptl3P2UaXn1aoJZq1Nc9jG-ARouxnb

Now, with case still open but computer functional, is a good time to do some benchmarks. I decided to test every supported clock rates starting from 16 MHz to 33 MHz and then with 486DX2 50 MHz and 66 MHz (way too late I realized I should have also tested DX2-33 and DX2-40, well maybe some other time).

I used Phils 386 Benchmark pack. I also ran Doom with min and max detail and Speedsys from DOS Benchmark pack.
Here are the results.
And here are Speedsys results.

I'm not stopping here. There is a reason why chose that old IBM PS/2 VGA monitor for this computer. And why I decided to keep the CPU underclocked to 25 MHz. More to come.

Edit: Added Speedsys results and reduced picture sizes.

Reply 1 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

HT216 seems to be fast enough for this system, thanks to VLB. VGA compatibility is also good. Tested with Keen 4 and everything works without need to turn on fixes from game settings. Compatibility is not perfect and Vinyl goddess from mars has problems. But the biggest issue is RAMDAC and very poor signal it generates. I first tried with 17" Viewsonic E70. Text modes and hires-graphic modes (16 color?) seems to be fine but 320x200 graphics modes, no matter if 16 or 256 color, are like looked behing a grid. 17" CRT monitors have visible scanlines and this computer produces horizontal lines between pixels.

Next I tried 14" ICL Monitor and it doesn't display scanlines since it is nice and small. This monitor is manufactured in october 93 so less than year older than this computer. However the vertical lines are present and they look like vertical scanlines. I tried to took pictures but that didn't work. The problem wasn't seen on pictures.

So even older CRT next, Nec Multisync 2A. Now the picture was good, at least as good as this monitor displays. The ICL has far better quality. Next I decided to try ISA graphics card. I happen to have hi-end #9 ISA graphics card based on S3 928. I did the testing with ICL monitor. There is a jumper on the motherboard to disable integrated graphics. Good thing I had the manual since there was no indication on the motherboard what the jumper does. The far sharper image quality was immediately obvious when I powered the computer. I also tested the performance and the ISA S3 was not much slower. I would have replaced the XT-CF with S3 but it does not fit. I need the other ISA slot for sound card and the bottom ISA slot have size limitations caused by hard disk and other SIMM. I'm sure there are smaller and good ISA graphics card but I did not have any. But no matter, there is still one small CRT left to test with.

uc?export=download&id=1XEYI9oCb2B_uVG-exxo4GQmdEEaQptQL

This one is IBM 8513, the original 12" VGA monitor. I really like this monitor because image quality is excellent. It has very good colors. In fact I have only one CRT that can be said to have as good colors and it is 19" Dell with Trinitron tube. So I decided to use the IBM since it looks nice with this computer and I have long hoped to put the 12" IBM to a good use. It is just too good to be simply stored without any use.

Finally about the sound card. The purpose for this system is to have Sound Blaster Pro 1.0 CT1330A. I've wanted to put that dual OPL 2 sound card to some of my retro systems but I didn't want to replace my good quality SB16, 32 or AWE32 models. But of course things did not go as well as I hoped.

Reply 2 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

OPL2 is known to be speed sensitive so I thought this would be a very good system for it because this would be just about as fast as it can be before there will be problems with speed sensitive games (not necessarily OPL2 related). Since this computer does not have turbo button (it only has turbo led), any DX2 is going to be too fast. And I have no real need for DX2 system since I already have a very good DX4 system.

I’ve been under assumption for a long time that 486DX-33 would be the fastest possible without speed issues. Turned out I was wrong. I tested with two games that I’ve known to be very problematic: Cycles and Indy last crusade (I tested with VGA version). On my DX4 system I need to turn off turbo AND L1 cache when normally it is enough to just disable turbo. And doing those both slow down more than necessary.

There was some problems from the very beginning. 33 MHz was certainly too much. Sometimes 25 MHz would work and sometimes not. It was odd so I also did some testing with SB Pro 2 CT1600 and SB16 SCSI CT1770. Not much difference between cards. Eventually I found out that INDY fails every time after I tested with Cycles first. Likely Cycles leaves registers to some bad state.

Anyway, 486-33 is too fast no matter if CT1330A or CT1600 is used. Indy works fine with 25 MHz but Cycles need internal cache disabled. I haven’t tested much but so far only Cycles has failed with 25 MHz so I left the CPU at that. Before closing the cover I noticed that there is some leaking below largest capacitors on the CT1330A. I’ll need to replace capacitors soon.

Reply 4 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

I scanned the manual and made backup from utility disk (=HT216 drivers). I took a hi-res scan from the manual cover. It has photo of the motherboard. I can try to take a actual photo when I get the new hard drive and fan. I'll backup the BIOS then also.

Reply 7 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie
evasive wrote on 2021-05-08, 21:35:

Great! But is also supports 16 and 20 MHz FSB.
I made a backup of the BIOS ROM. It is here. The motherboard has only one (128k) eprom and it also contains VGA bios (first 32k). Actual bios is in the last 64k.

My next improvement for this little computer is to try include XTIDE Universal BIOS in the unused 32k. I hope it works. That way I can save one ISA slot. Perhaps i can move CF reader to the front panel or perhaps i don't need it at all. I could use serial port transfer for example.

I'm also going to try to replace RAMDAC. I have started noticing the "vertical scanlines" even with the IBM monitor. But only on grey and it is not too bad but I'll try to make this computer as good as possible.

I already did some improvements. I replaced the PSU fan. The new one is one of those quiet 50mm Sunon maglev fans. It was the quietest (i think) 3200 rpm model. The old one was 40mm fan but i was able to get the 50mm fit nicely. I used elastic glue since screw holes are for 40 mm fans. This is a huge improvement but I ended adding 150 ohm resistor to make it even more quieter. It is now way quieter than the Maxtor hard disk. There didn't seem to be any heating issues but I'm going to test more to be sure.

Most of my time was spend on a little side project. I'm having problems with Sound Blaster Pro mixer, so i made another system for testing it. Well, the SBP mixer worked just fine on the other machine. But while testing the SBP, i decided to find fastest CPU speed for both Cycles and Indy to work. The other system has Opti 495SX based hybrid 386/486 motherboard. I tested it with Amd 386DX and Cyrix 486DLC processors. Fastest CPU speed for Sound Blaster pro 1 was 25 MHz. That was true for both 386DX and 486DLC. Cycles didn't work with 33 MHz but Indy did work with 33 and 40 MHz 386. DLC-33 required L1 or L2 to be disabled for Indy.

Sound Blaster Pro 2 worked with all 386 clock rates without problems. Cycles worked with DLC-33 but indy required L1 or L2 to be disabled.

I made a lot of benchmarks but they are still on a paper at the moment. I'm going to put them on here eventually for comparison. I think Headland sure is a fast chipset even without L2 cache.

Reply 8 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie
chrismeyer6 wrote on 2021-05-06, 18:02:

Without a turbo button on the case it probably uses a keyboard shortcut to enable/disable turbo mode.

I tried LCtrl-LAlt-GreyPlus and LCtrl-LAlt-GreyMinus combinations but they did nothing. Are there possibly other combinations to try? Turbo wasn't mentioned on the manuals at all.

Reply 9 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

XTIDE Univeral BIOS now successfully added to system rom. I simply placed it after VGA bios and it now appears at C800 segment as it should. I wondered why I couldn't use C800 with XT-CF (I had to use D000 instead) but the reason is that first 64k of the 128k system EPROM is mapped at C000 (VGA BIOS) and the remaining to F000 (main BIOS). Of course the VGA BIOS is only 32k as usual so there were 32k of zeroes from C800 to CFFF. There would have been plenty of space for VESA extensions, for example, but it was just unused. Well, now XTIDE Universal BIOS is there (using 10k).

Reply 12 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

uc?export=download&id=1_fzLxqBwiSJnjiVBKVE4mi94narxWkG-

Hard disks and adapters arrived. Why 1,8" hard disk? (I've never even tried those before). I really like Hitachi microdrives but they are no longer available, at least not at any reasonable price. Hitachi microdrives are small, noiseless and have never given any problems. They work just like any IDE hard disk as master or slave drive.

I know many of you prefer CF or SD cards and sure, they have superior seek times. But bad CF cards are simply not usable as hard disks. I had more or less problems with (cheap) CF cards when I first time tested those Hitachi microdrives. So I'll stay away from those, at least for now.

These 1,8" hard disks use the same CF interface as microdrives (and CF cards of course). But these are wider so IDE-to-CF adapter cannot be used (but most can be modified by removing the plastic sides from the CF connector). I ordered proper adapters but these are essentially IDE-to-CF adapters without those plastic parts. So it is very easy to connect the hard disk wrong way!!!

I'm glad I was careful. I looked from Toshiba datasheet where the pin1 is since nothing is printed on the hard disk. It was easy to find it there. But it was on the other side where the CF connector had the mark! It turned out the datasheet had IDE pinout but the CF pinout is different. The picture shows the correct way to install the hard disk. I did a quick test and XTIDE Universal BIOS detected the drive. But I need to mount this thing to something before continuing. There is nothing supporting the hard disk and CF pins are very fragile at the moment. Maybe I'll try to make some sort of 3,5" drive with CF slot for microdrives and CF cards (slave drive). But it needs to be good looking before I'm going to put it on this nice Copam.

Reply 14 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

Just a quick update on the progress. I'm waiting to get new RAMDAC from a Chinese seller and it seems to take a long time to arrive. But I did test Orpheus today. If real Sound Blaster Pro gives problems (the mixer issue) then what would be better alternative than Orpheus? Well, GUS Extreme could be but I don't have one (actually I used to have one). Then again, the Orpheus has things the GUS Extreme doesn't, like real OPL3 and intelligent mode MPU. And I have GUS ACE waiting if new RAMDAC helps. Otherwise I have to rely on ISA graphics card.

Orpheus seems to work just fine, except Cycles does not work at all. Disabling internal cache does not help. I also tested Indy Last Crusade and it behaves just like with real Sound Blaster Pro. With CPU at 25 MHz it works fine but set clock to 33 MHz and sound is garbage. So late ISA era sound chip did not help here (but is the real Yamaha OPL3 on the Orpheus actually much different? Maybe I should try with CrystalFM). I was hoping it would allow me to use 33 MHz.

I did one experiment unrelated to sound. There is JPX jumper on the motherboard. Manual says it is keylock and after placing jumper on it I can say that the manual is correct. I was hoping it would have been for turbo button.

Reply 16 of 42, by keropi

User metadata
Rank l33t++
Rank
l33t++

I can try this Cycles game tomorrow on my 386DX+Orpheus and see what happens - is it the one darry linked?
garbled music in Indy LC means you have the upatched game and you need to grab the "486 patch" to update the sound driver... this is unrelated to sound chip used - game's driver just go nuts on speedy systems.

🎵 PCMIDI mpu site buy+info
🎧 WIP Orpheus soundcard site
💻 WTB Amstrad PC7486SLC-33 system

Reply 17 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

I had completely forgotten about the Indy 486 patch. Thank you for mentioning it. I now remembered I've tested it long time ago. I'll (re)test that. Maybe I can keep the clock at 33 MHz after all (but I'll need to test other speed sensitive games too. I'll gladly stay at 25 MHz if it is more compatible).

But about Cycles. It is one of those games that sounded really awesome when I first heared it with Sound Blaster 2.0 (my first sound card). So maybe that is the reason why it has become a habit to always test sound cards and systems with that game. And it is very speed sensitive to get Adlib sounds working so it has been great game to test if system can be slowed down enough with turbo and/or disabling L1.

But back to Copam and turbo. I wonder if I really need turbo switch support after all. There are only two jumpers to set the CPU clock. I wonder what happens if I change them while system is on?

Reply 18 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

I did more tests, now with DX2 underclocked to 33 and 40 MHz. I updated the spreadsheet. Not much surprises there. As expected, the DX2/33 is close to DX-25. On some benchmarks it is slower but on some faster. But perhaps a little surprise is that DX2/40 couldn't beat DX-33 at every test, although it was faster in most. 25 MHz VLB clock is probably the biggest reason.

But what about compatibility? Is the DX2/33 still too fast, after all, it is somewhere between DX-25 and DX-33 being closer to DX-25. Well since the problem is software timing loops the CPU internal clock matters most. After all, the code is run from L1 cache so slower external bus does not matter there once the code is loaded. I tested with Sound Blaster Pro 2.0 CT1600 since that was what I was using from the start. Indy Last Crusade did not work without the 486 patch. But the 486 patch did help here. I of course also tested Cycles and I was surprised that it worked with DX2/33 and DX2/40 as well. I had to test with DX2/50 and DX2/66 also. It worked with 50 but failed at 66. I wonder why it now worked with 50 MHz? Perhaps I didn't test it previously because CT1330A was more important.

I wanted to test a third CPU. I've had UMC U5S-Super33 for many years but I've never tested it. It had horribly bend legs. It took a lot of time to get them straight but eventually I managed to get it installed. I jumpered the motherboard to 486SX but sadly it did not post with U5S installed. The U5S is two years later than the motherboard so no BIOS or motherboard support. I thought it wouldn't matter. I need to test the U5S on a more modern motherboard to make sure it is working. It would have been interesting to see would it have been slow enough at 25 MHz and how well it would have performed.

I also did test the "turbo switch" meaning changing CPU clock while the system is on. Usually that is not possible at all. Motherboards have ISA/VLB bus divider settings likely with jumpers or at least on the BIOS setup. But not this one. I think it is a Headland feature to configure them automatically. That still doesn't mean clock can be changed while power is on. The automatic configuration likely would be done only once while powering the system.

But it actually works! I tested with topbench to easily see realtime change in performance. And toggling the (temporary) switch between 25 and 33 MHz showed the same scores as seen in the spreadsheet. I tested by powering the system first with 25 MHz and then redo the test when powering with 33 MHz. Both worked. Great!

But there are limitations. First, the turbo led naturally won't change but it could be rewired to the turbo switch. But perhaps bigger problem is that perhaps the best turbo combination would be DX2/33 <-> DX2/66. But that requires change of two jumpers and the DX2/33 is not as compatible as I hoped. Although Indy works with 486 patch, there might be some game that does not have such patch. And this system was about compatibility anyway. I'm sure there would be a switch to allow changing both jumpers at once or as an alternative I could set the clocks to DX2/33 <-> DX2/50. That would be possible with one jumper and since Cycles did work with 50 MHz that might not be bad at all. But still, DX instead of DX2 feels much more comfortable and is more era correct CPU for this computer.

Reply 19 of 42, by aitotat

User metadata
Rank Newbie
Rank
Newbie

I had to test the switch a bit more because previous testing did not reveal anything about ISA dividers. After all there is no VLB divider since it runs at the same external clock as RAM. I first thought it would be easy to test with ISA graphics card and Landmark because the chr/ms gives very good indication about VLB speed. But instead I tested hard disk transfer rate since I assume that is connected to ISA since there is no VLB IDE controller on the motherboard.

First I got 2,6 MB/s on whatever clock I last left the computer. Flip the switch and now I got slightly less than 2 MB/s. It was obvious at this point that ISA divider did not change. Just to be sure I powered off the system and restarted it without touching the clock switch. Now I got slightly less than 2,3 MB/s with correct ISA divider. Flip the switch again and I got almost 3 MB/s. So definitely the ISA was overclocked at that rate.

So this cannot be used as a regular turbo switch but I could place it to the back of the computer. That way it would not get switched while power is on but I would have easy access to alternative clock rate if needed.