VOGONS


Suntac 80286 Mainboards

Topic actions

Reply 100 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t

I think it unlikely you have 2 incombatible VGAs.

I did a quick look and found two 8900C. I am sure I also have a D but my collection is not organized very well...I will report later if it works.

EDIT: Both TVGA8900C dont work, but I didnt test them in another board before.
There is a long break at
1A 19
and post sequence ends at
30 24
No display on VGA.

EDIT2: Tried the ET4000 again, works. Exchanged against the TVGA again - also works now. Strange...

Indeed, both TVGA work now. Maybe the board didnt like the cold boot, 1st time I turned it on today.

Reply 101 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

Phoenix BIOS goes up to code 00 so seems to boot, does 2 beeps at the end (twice I think). Seems to initialize the keyboard too. But no matter what I set on the jumpers, there is no video output. In fact this BIOS completly ignores the card, inserted or not there is no difference in boot sequence.

Award dies at code 26, but before that stops briefly at 1F and 25, these are memory tests. At 26 it does 2 short beeps, twice, and that's it, but somewhere around 25 keyboard is enabled and I can turn caps/numlock on/off. 8042 also responds to keylock (I used a jumper to simulate it) by disabling the keyboard. But as with Phoenix, there is no video output, card can be missing and there's no difference.

I've tried my own "BIOS" but it seems just after cold boot it's not possible to access ISA slots memory - weird. Perhaps some internal chipset registers must be configured first. I tried to dump video ROM contents via the test card port but all I get is 0xff. It's super annoying to have to reprogram 2 chips to try new code. But at the very least there is something to do here.

Your boot sequence with not working card is exactly like mine. The long break at 1A 19 is where the beeps happen.

Reply 102 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t

Crazy idea for this: If the board comes up with Phoenix you could connect a floppy, boot to DOS and make a dump of RAM/ROM to the disk to see whats going on there...
If this works, you could also try to put a copy of the Video-ROM to the floppy and try to initialize manually from RAM 😉 Think its not hardwired to C000?
Could save lots of work instead of making custom ROMs...

Reply 103 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

That would work if I could enter the BIOS and set the drives, but I can't.

Anyway, I probed the ISA slots, scope shows all is good. Wrote a program to read/write all the 8 bits through RAM in the RTC, all OK. No stuck data lines, at least not the low byte ones.

Then I removed the ROM on Realtek and probed the socket pins, and what do you know, the /OE is toggling when the /MEMR signal arrives, so the chip is detecting correct address 0xc0000. In fact the address bits on the socket follow ISA slot and I can see the increments. If I pull the ISA slot data pin with resistor to ground I get a bit change on the card. The program works, the read works, so why is there 0xff on the bus.

Well, under the ROM there is a '245 and it's controlled by the Realtek chip to allow reads and writes. And it's permanently disabled. Nothing gets through in any direction. So as a last resort I put the Trident in the slot - and test card shows the ROM data now. So Trident ROM read works but apparently the card itself does not. It could be the ALE signal though. But why doesn't Realtek work? I suspect the card needs some register write to enable it first. So a modern BIOS will do it while probing hardware but these old ones don't do that. One more reason to hate this piece of sh... The card still works in a 386 system, I checked.

You said it's unlikely that I have 2 cards that don't work, but I'm pretty sure now I'm just unlucky with this combo and that is exactly my problem. I see a Trident 8900C for sale, nice clean card and I can live with the price but it's missing RAM. I have two more working DRAM chips so I can at least use the VGA modes, so I think I'll buy it.

Reply 104 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t
Deunan wrote:

That would work if I could enter the BIOS and set the drives, but I can't.

Maybe I can tell you which keys to press to get into BIOS set up the floppy?

I can send you one of mine cards for shipping only from GER, but its missing the slot bracket.

EDIT: Also found a RTG3105. Dont know if its working, will test later...

Reply 105 of 187, by Deksor

User metadata
Rank l33t
Rank
l33t

Some other crazy hack you could do is to replace the HD146818P with a DALLAS rtc chip like on mine, find another motherboard that supports DALLAS chips that have a AMI bios and then set the settings you want, and finally put the DALLAS chip back in the suntac mobo.

You think this is crazy and won't work ? Well when I brought the "new" dallas chip in my suntac mobo, there was still some data left inside (this chip was probably salvaged from another mobo. The previous owner modded it to have a CR2032) and some settings were still present when the bios fired up !

Trying to identify old hardware ? Visit The retro web - Project's thread The Retro Web project - a stason.org/TH99 alternative

Reply 106 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t

Hmm seems to be something wrong with my Realtek but it works...
Came without RAM and I took one half from the Trident...

Afraid you need another explanation.

IMG_4780r.jpg
Filename
IMG_4780r.jpg
File size
1.75 MiB
Views
1434 views
File license
Fair use/fair dealing exception
IMG_4781r.jpg
Filename
IMG_4781r.jpg
File size
1.73 MiB
Views
1434 views
File license
Fair use/fair dealing exception

Also was interested and tried the "manual" VGA initialization, works!

DSC_8556.JPG
Filename
DSC_8556.JPG
File size
1.66 MiB
Views
1427 views
File license
Fair use/fair dealing exception
DSC_8558.JPG
Filename
DSC_8558.JPG
File size
1003.43 KiB
Views
1427 views
File license
Fair use/fair dealing exception

Reply 107 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie
Deksor wrote:

Some other crazy hack you could do is to replace the HD146818P with a DALLAS rtc chip like on mine, find another motherboard that supports DALLAS chips that have a AMI bios and then set the settings you want, and finally put the DALLAS chip back in the suntac mobo.

No need for a Dallas - I have one but it has dead battery and needs drilling, too much work. I can just add a coin battery and a diode to a Motorola chip. But the problem is I have no other 286 boards, Suntac or otherwise. Though maybe PCem can produce suitable NVRAM "dump".

Predator99 wrote:

Hmm seems to be something wrong with my Realtek but it works...

Oh, it's surprising you got Realtek to boot with AMI BIOS. But perhaps it's because you already had a working card there before? Does it also work with CMOS battery removed, when settings are lost? If not, does the Trident 8900C work from cold boot when settings are lost?
Also, I don't think the font corruption is a coincidence - if you can test this card in another (non-Suntac) mobo you will most likely find there is no corruption there.

Reply 108 of 187, by Deksor

User metadata
Rank l33t
Rank
l33t

Well I don't think that Dallas chip was used in a 286 before, probably a pentium due to the datecode on it.

Trying to identify old hardware ? Visit The retro web - Project's thread The Retro Web project - a stason.org/TH99 alternative

Reply 109 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t
Deunan wrote:

Oh, it's surprising you got Realtek to boot with AMI BIOS. But perhaps it's because you already had a working card there before? Does it also work with CMOS battery removed, when settings are lost? If not, does the Trident 8900C work from cold boot when settings are lost?
Also, I don't think the font corruption is a coincidence - if you can test this card in another (non-Suntac) mobo you will most likely find there is no corruption there.

Currently using this board:
download/file.php?id=58225&mode=view
Didnt connect a battery and it forget its settings every day. I boot with XT-IDE thereofre no need to enter any data.
The Realtek now works also on a cold boot on the 1st attempt. But display errors still there. They are only visible on certain colors. 320x200 games run without problem.

Reply 110 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

Few more hours of tests and I know why the cards don't work. BIOS detects wrong video ROM checksum. It's wrong because all reads from 0xc0000-0xc5fff are OK, but at 0xc6000 and above it's a glichfest. The card ROM is 0x8000 long - so the upper 0x2000 bytes are gone. Unfortunately on Tridents there is some code in that area, not just fonts, so it has to be correct. I tried mod the card bios to pretend it's smaller and now it passes the checksum test but hangs, and I was not able to fully patch this code out.

So a mono card would work because it has no BIOS, and most EGAs have only 16k so would work as well. But I'm hoping the 8900C will work too, since it doesn't need ALE signal. I tried to soft-mod that signal out on the 8900D with capton tape but it doesn't seem to cure the problem. Also, this mobo is seriously picky about the card and slots - so far the Trident works only when jumpered to 8-bit and inserted into 8-bit slot (just like manual says, slot 1 or 7). Any other combination and I can't read anything form the card at all.

This is a really interesting problem, now I've patched out BIOS self-test and I'm modding that. But having to swap two chips every time is annoying, I wonder if getting an ISA Ethernet card with empty ROM socket would help here.

Predator99 wrote:

I boot with XT-IDE thereofre no need to enter any data.

Can you try without that? You don't need to boot into OS, just to see if the VGA gets initialized with invalid CMOS data. Also, which slot are you using?

In the meantime I'm going to try and test all 36 DRAM chips in external tester.

Reply 111 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t

Yes, works also without XT-IDE on ethernet card. Thats clear because this option ROM is intiliaized immediately before boot, just like a SCSI disk. It doesnt rely on CMOS settings but autodetects the HDD.

Just tested the Realtek in the outer 8-bit slot (Slot 1), Slot 5 and Slot 6.
Slot 2 didnt work but this may be a corrosion problem on this board....

Interesting issue with the cut at 0xc6000....but if this is the reason it should work what I described above with Video-init from DOS.
You should find a way to boot to DOS. Do you have a floppy? If you tell me the BIOS whcih is working I can tell you how to get into BIOS and set up the drive...

Thats my Realtk BIOS, doesnt seem to be Code after 0x6000.

Filename
REALTEK.rar
File size
16.96 KiB
Downloads
61 downloads
File license
Fair use/fair dealing exception

Reply 112 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

Ok, thanks for testing.

Realtek doesn't seem to work on this mobo at all. Trident sort of does in 8-bit slot but not really due to the glitching issue.
There is no need to boot to DOS to init the card, I can just mod the BIOS to copy the video ROM to RAM, patch it and run from there. That doesn't work either - and I suspect the glitching problem is probably affecting VRAM as well. I tested this method (with the ROM filled with garbage on purpose) on PCem and it works there. The code patches are good.

In fact this mobo has other strange quirks. The BIOS is actually 2 parts: the init/boot and a testing suite. This testing part is just a BIOS extension located at 0xf0000 - so I figured I'll overwrite it with the video ROM and run it from there. Doesn't work either. In fact, even though the BIOS is executing code from 0xf8000 and it works, trying to read this part of memory or copy it to ram hangs the system, or makes it go crazy. But perhaps this is some sort of side effect of the way I'm injecting my code, this is still early system init so maybe some stuff doesn't properly work yet.

So on one hand I'm happy this wasn't a trivial fix, because I'd loose interest in this mobo by now. This way I keep thinking how to hack it. On the other I don't want to spend too much buying more stuff to test it properly.

Reply 113 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t
Deunan wrote:

Ok, thanks for testing.

In fact this mobo has other strange quirks. The BIOS is actually 2 parts: the init/boot and a testing suite. This testing part is just a BIOS extension located at 0xf0000

Yes, this applies to the AMI diagnostic ROM which is 64 kb in size.

Did you manage to get the Award ROM running in PCem? It is only 32kb in size. The 64 kb AMI image runs as "AMI 286 clone" as well as "Award 286 clone".
I tried everything with the Award image. Read direct from the ICs, used NSSI image, filled the 1st 32kb with 00 or FF, ... nothing. PCem doesnt start with it. Maybe you have an idea whats wrong here?

Reply 114 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

It doesn't work for me either. But PCem has it's own quirks, so that doesn't surprise me. I have an OPTi 495SLC BIOS that only works with a specific CPU in PCem but has no such issues on real HW - here it boot with both 386DX and 486DLC. Emulators will always have some limitations.

It has been a pretty productive day for me, I found one bad DRAM chip so now I'm down to 512k on this mobo. In the process I had to update my tester a bit since I noticed it's not very good at finding some fault patterns in 1-bit memories. I really should make a PCB for this at some point, it's a wire mess now and difficult to use. I found some sellers with cheap GM71C1000 chips so I might as well upgrade bank 0 to 2MB.

And I also realized (or rather, I knew about it but forgot) that Atmel 28C256 has slightly different pinout than 27C256. So that explains some of the problems I got accessing 0xf0000 area while BIOS at 0xf8000 works OK. A14 line is in the wrong place. I guess I need to get more of these modern 27C512 replacements like SST27SF512 or W27E512. These are fully pin-compatible. All the 5V FLASH chips I have are 32-pins.

So this project is on hold until I get the 8900C Trident and/or the other parts. If that doesn't work then I'll need to get a cheap Hercules card and find a way to display that on a modern LCD monitor. If OSSC can output 50Hz properly then maybe my FreeSync Iiyama can actually show that. One 286 board, so many new things to learn.

Reply 115 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t
Deunan wrote:

Also, I don't think the font corruption is a coincidence - if you can test this card in another (non-Suntac) mobo you will most likely find there is no corruption there.

Tested the Realtek 3105 in a 486 board. Same display error in yellow and blue letters, but overall the picture looks sharper with better black level.

Deunan wrote:

It doesn't work for me either. But PCem has it's own quirks, so that doesn't surprise me. I have an OPTi 495SLC BIOS that only works with a specific CPU in PCem but has no such issues on real HW - here it boot with both 386DX and 486DLC. Emulators will always have some limitations.

It doesnt make sense for me as both BIOS run on the same real hardware. Unfortunateley I dont see any debugging screen in PCem to see where it stops. They should add a CPU screen for debugging...

Reply 116 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie
Predator99 wrote:

Tested the Realtek 3105 in a 486 board. Same display error in yellow and blue letters, but overall the picture looks sharper with better black level.

Hm, curious. But then it only confirms what I think about these cards. POS. I liked Realtek affordable Ethernet cards. I love their low-noise audio chips. But these VGAs should be burned to ash and forgotten. The card has poor DAC with a lot of internal noise and no HF filters of any kind on the output. And it requires both +12V to work, and -12V to actually show something on the screen. On top of it, Realtek card is slower than some 8-bit ones I've seen.
The only reason I still keep mine is to test new mobos with it. If something catches fire, I'd rather write off the Realtek than a nice Trident D or better VGA.

Predator99 wrote:

It doesnt make sense for me as both BIOS run on the same real hardware. Unfortunateley I dont see any debugging screen in PCem to see where it stops. They should add a CPU screen for debugging...

A monitor for port 80 writes would help as well. I haven't read all the source code but I suspect the emulator doesn't fully emulate the hardware, or makes some assumptions about it, and this works on AMI BIOS but AWARD perhaps does things differently enough to break things. It's to be expected, most of this stuff was reverse engineered by watching the original code execute. This is very time consuming and a lot of trial and error, and guesswork, is involved. There simply isn't anyone with free time to do it for AWARD BIOS.

Reply 117 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

Another busy evening but the case is almost closed. I finally know what's wrong, so now I just have to fix it.

I put ATI Graphics Solution card into this mobo and it hanged the system with AWARD BIOS during video init. So I probed the pins and realized the card is never letting go of the I/O CH RDY signal, so basically it enters infinite wait state. It happened as the BIOS was trying read MDA memory at 0xb8000. I put a piece of capton tape over this contact and the BIOS goes futher but still no video output, and CGA does not have any ROM to init. It's all in the BIOS. So clearly the BIOS is not detecting the card at all.
After checking all the control signals again, the pull-ups and chip connection, it suddenly occured to me that RESET DRV is sort-of high (around 3V) and its active-high signal. So I shorted it to ground with a wire and... it all works now. The ATI card, the Tridents as well.

Long story short, B2 pin on ISA bus has broken connection to the '001 chip and is just floating. The voltage depends on what, if any, cards are in the slots. And apparently I somehow missed that pin when testing the ISA slot connections. Or maybe it's a microcrack and was still sort of conducting when I tested with the meter.

I still don't know where the copper has cracked but I'll look for it when I have a bit more free time, and there's a better light outside. Worst case scenario I can always add a wire, it's a not a signal that changes state except on actual mobo reset. Maybe all this info will help someone else fix their mobo.

UPDATE: Found the microcrack, right next to the ISA slots on one pin of a simple RF filter (wire through a ferrite core) that keeps this signal clean - as it goes all way around the mobo. I tried to make a photo but that just wasn't possible without a microscope with camera adapter. The crack would make/break contact as the mobo flexed - meter showed completly open and some 50 ohms depending how I tentioned the PCB. So it's possible it tested OK-ish when I checked the mobo by itself but went bad each time a card was inserted.

Reply 118 of 187, by Predator99

User metadata
Rank l33t
Rank
l33t

Congratulations Deunan! Would also be happy to see a photo of your complete board.

In the mean time I got another one. Looking at the corrosion I was thinking its never going to power on:

SuntacPeacockDirty.jpg
Filename
SuntacPeacockDirty.jpg
File size
1.7 MiB
Views
1265 views
File license
Fair use/fair dealing exception

Its Peacock branded, also with AMI Bios

Filename
Suntac Peacock.rar
File size
1.75 MiB
Downloads
67 downloads
File license
Fair use/fair dealing exception

After some cleaning...

SuntacPeacockc.jpg
Filename
SuntacPeacockc.jpg
File size
1.74 MiB
Views
1265 views
File license
Fair use/fair dealing exception
SuntacPeacockcr.jpg
Filename
SuntacPeacockcr.jpg
File size
1.75 MiB
Views
1265 views
File license
Fair use/fair dealing exception

....I gave it a try. And after replacing the keyboard controller and filling bank 0 with RAM it get as far as "20 1A" on the Post card! A message is displayed on the VGA:

CMOS INOPERATIONAL
SYSTEM HALTED

I assume its related to the mc146818, but the IC itself looks OK...

Reply 119 of 187, by Deunan

User metadata
Rank Oldbie
Rank
Oldbie

I will make a photo once I get new DRAM chips. There's been a shipping error so it'll take a while still. In the meantime I got a cheap 287-10 and it works fine - I've even back-ported my 386-only testing program to 286. Fun! Also I've discovered that NSSI will not run with 512kB system RAM, even though there's more free DOS memory than the minimum in the TXT file.

Still waiting for 2 more memory chips for the Trident C, I had 4 originally but one is cracked and dead. The card works with just 2 as 256k but I assume in 8-bit mode so is slower than it could be. Curiously the 8900D doesn't work still, it now tries to init (and sometimes the screen turns all blue) but hangs. So either I was right about the ALE signal after all or the BIOS requires 386 (unlikely?).

As for you new mobo, I can see A LOT of corrosion near the ISA slots, the '245, keyboard socket and the RTC. There's no way data signals D0-D7 are still working on this thing.
Considering the 008 chip state as well I'd say this is parts donor board at this point. I mean, it coudl be repaired, but is it really worth rewiring half the PCB?