Tseng Labs ET 4000W32i 1Mb video corruption

[Alex_Baker]

I bought this card a while ago for a decent price, but I may know why.

This is a Tseng labs ET 4000W32i VGA card with 1Mb of video memory, chip number is ET4000AX. As you can see from the attached pictures there are corrupted groups of pixels that flash rapidly. Here is a video; https://youtu.be/GgTMg_G7jSE

Perhaps some people here know what sorts of things would cause this. I have tried re-seating all socketed ICs, and thoroughly inspecting the board for damage with no results. The monitor is a 4ish year old cheap LCD PC monitor. Interestingly, moving the player back and forth in that game sort of "cleans" the corrupted spots.

The test system is an intel 386sx with 4Mb of memory, I have a soundblaster AWE64 installed, and storage/floppy is handled by the "Chips" chipset on the motherboard. This motherboard is from an industrial computer, so the processor is on a card that slots into a passive 16b ISA backplane. I was not able to find any specific information on this vga card, all I know is that it had the CompuAdd branding on it.

[Alex_Baker]

more photos.

[amyk]

Since it has socketed VRAM, you can try moving them around to see if it's a problem with the socket or the chip.

[fzabkar]

I notice that the soldered DRAMs are 80nsec parts whereas the socketed DRAMs are 100nsec parts. Their date codes are similar, though. This would suggest that the card was produced with 1MB and 512KB options, and the additional DRAMs were probably installed by the factory.

https://www.dosdays.co.uk/media/micron/MT4C4256.pdf
http://www.bitsavers.org/components/fujitsu/_dataBooks/1990_Fujitsu_Dynamic_RAM_Products.pdf

It might be an idea to replace the slower DRAMs, but the cost would probably be uneconomical.

You might try to experiment with JP2 near the ISA fingers.

The other thing you could try is to select one of the high resolution screen modes using DMODE.EXE (a Tseng Labs tool). This might enable you to see stuck bits in the display RAM.

https://ia804501.us.archive.org/28/items/tseng-labs-et4000-vga-manual/TsengLabsET4000VGA.pdf

[Alex_Baker]

Do you know where I might be able to find Dmode.exe? I found a couple of disk images on archive.org for drivers/utilities but it did not include dmode. It did include Vdiag, which I ran and saw no issues, except for this modern monitor not supporting 3 of the tests that it did.

It might be an idea to replace the slower DRAMs, but the cost would probably be uneconomical.

You might try to experiment with JP2 near the ISA fingers.

I tried removing the socketed DRAMs all together so that it would act like a 512k card, it made no difference.

I determined that JP2 is connected to IRQ2 on the ISA slot. I was told that IRQ2 can cause compatibility issues with sound cards, but switching the jumper or removing the soundblaster entirely had no effect.

before I go too far down a rabbit hole, is it possible for this type of corrupted pixels to be caused by bad game files?

[fzabkar]

Remove the socketed DRAMs and piggy-back them onto the soldered ones. This sometimes enables good RAMs to mask the bad ones.

I checked my old ET4000 driver diskette, but there were no EXEs. Sorry.

[HwAoRrDk]

Do you know where I might be able to find Dmode.exe?

You can find it here:
http://ftp.oldskool.org/pub/drivers/TSENG/ET4000/dos/UTILITY/

[Alex_Baker]

Do you know where I might be able to find Dmode.exe?

You can find it here:
http://ftp.oldskool.org/pub/drivers/TSENG/ET4000/dos/UTILITY/

Awesome! Thanks!

Remove the socketed DRAMs and piggy-back them onto the soldered ones. This sometimes enables good RAMs to mask the bad ones.

I checked my old ET4000 driver diskette, but there were no EXEs. Sorry.

The manual on archive.org that you linked actually calls it Dmode.com, not .exe, sorry. Though apparantly a .exe version does exist, que HwAoRrDk.

[coromonadalix]

btw    a dot.com file can be launched  like an exe ?



and yes this link as an dmode.exe

http://ftp.oldskool.org/pub/drivers/TSENG/ET4000/dos/UTILITY/

[Alex_Baker]

Dmode was a fail, attempting to run it only gives this message; "The DMODE.EXE only used in ET4000/W32 video controller". So Dmode is not compatible with the W32i.

[coromonadalix]

https://dosdays.co.uk/topics/Manufacturers/tsenglabs_downloads.php  ?

[Alex_Baker]

update, under the DOS_other directory on oldskool.org there was another Dmode.exe that works.

[fzabkar]

Use a hex editor to create a file consisting of several thousand repetitions of 0x55 0xAA, and then display it as follows:

   c:\type 55AA.bin

Hopefully this will identify any bad bits.

You may need to experiment with different screen modes to test all of RAM.

FWIW, there is a TLITEST.EXE file in this package:

https://dosdays.co.uk/media/drivers/graphics/TSENG/Tseng_Labs_ET4000W32PCI_Win3.x.zip

[Alex_Baker]

I ran that test on about a dozen different graphics modes but I did not get any results. The only thing that I noticed was that some of the higher resolution/column modes were fuzzy or noisy, the picture below is 800x600 @256 color. The screen I am using is a modern LCD.

edit: the image is rotated 90deg CC from how it actually is, I don't know why this website does that.

[fzabkar]

I find it strange that the chip at 2C is facing in the opposite direction to all the others.

[coromonadalix]

good catch,   is it socketed correctly or 180 degree ??


all the others ic's  are in the same direction ??

[aeg]

That's the RAMDAC, if it was in backwards, there wouldn't be any video output.

[Alex_Baker]

At this point I think I need to look into some more comprehensive graphics stress tests, or another DOS game.

So far I have only seen this problem show up in that one DOS game, so I want to make 100% sure that this is not or is not caused by the game before going down the hardware rabbit hole.

[amyk]

Try Windows 3.11? I believe there are drivers for that card.

[Alex_Baker]

Try Windows 3.11? I believe there are drivers for that card.

I was kind of thinking that, I have install disks and I have been wanting to try windows 3.11 anyway.

[Alex_Baker]

I installed windows 3.1 this morning, I had said I was going to install 3.11, but my disks are actually 3.1 so I just went with it. Unfortunately the graphics problem showed up again during setup.

It slowly got worse as setup continued.

Do we know what graphics mode setup runs in? we know for sure that this card has issues in that mode I guess.

[coromonadalix]

normally vga 640x480  ??  no ?

not sure if its related to brightness and contrast ...  over saturation 

maybe the ram dac got funky ?

[Alex_Baker]

I am leaning more and more towards a memory issue. Mainly because when you look at this picture:

All the correct pixels are there, but some of them are in the wrong place, as if an upper memory address line was flaky. Address line being either a bad mem chip or the "GPU" reading from the wrong location in memory. I will see if any replacement memory chips are available anywhere, just in case it comes to that.

normally vga 640x480  ??  no ?

not sure if its related to brightness and contrast ...  over saturation 

maybe the ram dac got funky ?

I feel like the ramdac is ok, all the colors looked fine in the vidtest program that I ran, even the 256 color mode.

[fzabkar]

What if you were to use a DOS TSR program like Snarf (or Screen Thief) to capture the screen output? Would the resulting file be a copy of the actual content of the card's RAM?

Snarf - Tool to capture full-screen DOS graphics :
https://www.hddoracle.com/viewtopic.php?p=15796#p15796

[Alex_Baker]

What if you were to use a DOS TSR program like Snarf (or Screen Thief) to capture the screen output? Would the resulting file be a copy of the actual content of the card's RAM?

Snarf - Tool to capture full-screen DOS graphics :
https://www.hddoracle.com/viewtopic.php?p=15796#p15796

Interesting Idea I will try that.

on a different note,
This afternoon I decided to socket the other 4 Dram chips. That allowed me to swap them out as I please. I tried removing the 4 extra chips (socketed ones from the factory) and replacing the others one by one with a different chip. Unfortunately I am still at square 1, I saw no changes swapping out chips, so perhaps the ram ICs are ok and the problem is elsewhere.

[Alex_Baker]

I was not able to get snarf to work, kept getting and error 1, but I did get screen thief to work.

Here is what I got.

[fzabkar]

To me, that means that the RAMDAC is not at fault. The problem must be happening when data is written to DRAM.

FWIW, these old cards had a "wait state" jumper. Perhaps one of those hard-wired, unpopulated jumpers at the LHS of the PCB selects this option (JP5 and JP6).

https://stason.org/TULARC/pc/graphics-cards/U-Z/UNIDENTIFIED-XVGA-ET-4000.html

Is JP2 meant to be open, or must you select either 1-2 or 2-3? It might be an idea to trace the pins to the ISA slot to see where they go.

[Alex_Baker]

JP2 is connected to IRQ2 on the ISA slot, left is open right is closed. Other than that there are no other jumpers on this thing.

[fzabkar]

JP2 is connected to IRQ2 on the ISA slot, left is open right is closed. Other than that there are no other jumpers on this thing.

At the top left corner of the card are 2 unpopulated jumpers. I suspect that the 0WS / 1WS jumper has been preset to 0WS by a hardwired trace. I'm thinking that the trace could be cut and the alternative pin selected by installing your own header.

[Alex_Baker]

You are right. The one labeled JP5 is connected to one of the select pins for IC 5C, which is a "dot frequency generator" for graphics applications, probably don't need to mess with that. I can't tell where JP6 goes.

There are also two more unpopulated jumpers underneath the main IC labeled JP 7 and JP 9, it's impossible to guess what they may be.

I guess I can try adding jumpers and testing different combos. But I don't know how the jumpers could be my problem, shouldn't the card work as it is?

Perhaps another clue is the fact that this card looks like it spent some time bouncing around in a bin of other electronics given the scratches and bent lead-tails on the back. Maybe the old owner gave up on it.

[fzabkar]

Perhaps it's time to give up and try a different card? I mostly used Trident cards during those days.

IIRC, the only other thing that was configurable was shadowing of VGA BIOS in RAM. I think Chips & Tech chipsets had this feature.  I can't see how this would change anything in your case, though.

Edit:

One more idea. Does your motherboard have a turbo mode? Perhaps you could reduce the speed, just in case the card is too slow for the CPU.

Edit#2:

Here is the ET4000 datasheet:

https://mirrorservice.org/sites/www.bitsavers.org/components/tsengLabs/Tseng_Labs_ET4000_Graphics_Controller_1990.pdf

[coromonadalix]

man a 33 year  datasheet      getting old here

[Alex_Baker]

I did end up installing some jumpers and testing different combination, none of which had any effect on the corruption.

I may just need to cut my losses on this thing, which is annoying because it was cheap for an ET4000, but not that cheap for an ISA VGA card.

[Alex_Baker]

Well now I feel stupid, the Tseng card works now, let me explain.

I decided I would search through the bios for anything that could help me. I found settings for ROM shadowing and some various other video related things, no differences. But I found a setting called "Hidden Refresh" under advanced chipset settings, as it turns out disabling whatever that is solves all my woes. If anybody can explain what hidden means I am all ears, but for now I am just going to take what I can get.

Should I have looked through the bios before? probably.

[nctnico]

It could be a soldering issue on the Tseng labs chip itself.

[coromonadalix]

from vogons ??

never heard of that one ?

Hidden refresh is a method to do a refresh cycle as part of normal access cycle to a DRAM chip rather than needing a dedicated refresh cycle which will be slower, it helps to increase memory bandwidth. All 72pin SIMMs support hidden refresh and most 30pin SIMMs should too, at least 1MB and bigger ones.
Slow refresh is extending the time period between refresh operations to reduce bandwidth needed on refreshing the memory to increase performance. Data corruption can happen if the memory cannot handle it, you will have to experimentally verify that things are stable.

[rsjsouza]

I have been silently following this thread and I was going to suggest to wash your board (residue and other issues can sometimes affect funcrionality), but thanks for sharing the solution. I have seen many BIOS options over the years but I don't remember that one.

Enjoy your "new" board

[mfro]

I guess this doesn't help much with your problem, but it appears to me the card is *not* an ET4000 W32i but a plain ET4000 (as it has the ET4000AX chip).

[HwAoRrDk]

I guess this doesn't help much with your problem, but it appears to me the card is *not* an ET4000 W32i but a plain ET4000 (as it has the ET4000AX chip).

Oh yeah, that's true. I didn't notice that. A W32i will have a chip marked "ET4000/W32I".

Is OP sure they're not just using the wrong drivers?

[Alex_Baker]

I guess this doesn't help much with your problem, but it appears to me the card is *not* an ET4000 W32i but a plain ET4000 (as it has the ET4000AX chip).

Oh yeah, that's true. I didn't notice that. A W32i will have a chip marked "ET4000/W32I".

Is OP sure they're not just using the wrong drivers?

I was starting to wonder about that, it must be a problem with how the PC benchmark program I was using reports it, which is a W32i, but I don't know how it would actually tell.

[coromonadalix]

had doubts with the VG-5500 markings ??  even the fcc id markings ??

[Alex_Baker]

had doubts with the VG-5500 markings ??  even the fcc id markings ??

As big as compuadd was at one point (according to Wikipedia) I got no results looking up those markings, so all I had to go off of was the incorrect reporting of the benchmark software.