root42, one additional question: do I get the right feeling that the card is rendering graphics very slow? Is the PC really slow? Do you happen to know what frequency the ISA bus is running in your machine?

For instance, on my 386 DX / 33 MHz, I set the ISA bus speed to 11 MHz and Landmark Speed reports around 4488 chr/ms with this video card in 80 x 25 text-mode. Furthermore, graphics seem to be rendered faster on the 386.

When you have some time, can you please check what LM reports on text-mode video speed?

I hope you already set the 0WS jumper.

I am asking just out of curiosity. This is by no means related to DRAM timing but just a quick observation while watching your video.

Cheers,

ALEKS wrote on 2021-07-10, 11:01:

root42, one additional question: do I get the right feeling that the card is rendering graphics very slow? Is the PC really slow? Do you happen to know what frequency the ISA bus is running in your machine?

For instance, on my 386 DX / 33 MHz, I set the ISA bus speed to 11 MHz and Landmark Speed reports around 4488 chr/ms with this video card in 80 x 25 text-mode. Furthermore, graphics seem to be rendered faster on the 386.

When you have some time, can you please check what LM reports on text-mode video speed?

Hm, it is a 486DX/33 with MR BIOS. I didn't change anything in the BIOS, so it reports itself 8.3 MHz. It is fast enough for my taste... I am more intrigued with the W32 part and that the card will run AT ALL, not that it runs super fast. I can play around with the ISA bus at some point, but first I want to get it stable.

ALEKS wrote on 2021-07-10, 11:01:

I hope you already set the 0WS jumper.

I did not yet set the 0WS jumper, but will also experiment with that once the card becomes stable enough.

I will pick up my card in the next weeks and try to fix it as much as possible. However I was wondering: Can we replace the 35ns RAM with 50ns and then it will be better...? The ET4000W32I demands 50ns FPM, and the 41c16257-50K seems to be compatible. The timings are different to the 35ns variant:

If there is a chance that it will help (even with different termination resistors) I would try to source some of the 50K variant.

@ALEKS: are there any news on the memory issues? I read on your project page (https://alexandrugroza.ro/microelectronics/sy … ller/index.html) that you want to design an interposed module to fix the delay of the signals. Is there anything presentable already? I would love to pick up this project again!

Hi root42,

I was busy with other stuff (both daily routine and other microelectronics projects) and haven't had time to check on the VOGONS forum.
Yes, I finally reprised working on the ISA video card. The goal is to make it more stable and easier to be replicated by hobbyists.

I don't have anything ready yet but I have very good results from an active memory driver circuit that I was working on. It is going to be an all-SMD design on two small interposer PCBs that will be pluggable directly in place of the old dampening resistors.
I just took the time to build a second graphics card because the first one is actually in use in my main retrocomputer. And since I have all the parts and initial revision PCBs (the ones with the RAMDAC adapter) I just assembled another test subject.
This time I went with sockets for easier improved memory driver interface.

The LA probes adds some capacitive loading on the memory refresh lines which introduces some real-time visual artifacts. But that is as expected. I can filter those out "mentally", while regarding the VGA CRT output.

The logic analyzer shows that the /D* output signals are timed as it should, even though their /* reference input signals are not that perfectly shaped in time.

As I was saying, despite the unshielded wire forest, I already obtained very good results. No flicker and no random pixels igniting on the screen anymore. The colors are very bright and accurate as well.

However it will take some more time until I can release the updated memory driver circuits. I just want to make sure everything works as expected in the first place.

Cheers,
A.

Hey Aleks,

that looks awesome! I am looking forward to the interposer boards! I will have some manufactured and send them out to the people who ordered the original kits with me, once you are done with them!

Really looking forward to having a perfect homebrew ISA accelerator!

does this design support 24bit colour?

what about GUI acceleration in Windows 3.1?

maxtherabbit wrote on 2021-11-07, 13:34:

does this design support 24bit colour?

what about GUI acceleration in Windows 3.1?

you forgot to ask if it can run crysis in 4k 🤣

weedeewee wrote on 2021-11-07, 14:16:

maxtherabbit wrote on 2021-11-07, 13:34:

does this design support 24bit colour?

what about GUI acceleration in Windows 3.1?

you forgot to ask if it can run crysis in 4k 🤣

🤣 I wasn't trolling though, these are genuine questions

24 bit ISA ET4k cards were a thing (Diamond Speedstar24)

maxtherabbit wrote on 2021-11-07, 14:35:

weedeewee wrote on 2021-11-07, 14:16:

maxtherabbit wrote on 2021-11-07, 13:34:

does this design support 24bit colour?

what about GUI acceleration in Windows 3.1?

you forgot to ask if it can run crysis in 4k 🤣

🤣 I wasn't trolling though, these are genuine questions

24 bit ISA ET4k cards were a thing (Diamond Speedstar24)

well... the 24 bit should be ramdac specific so... ATT24C490-80 from the datasheet... yes, though this one has no gamma corrected true color modes ? though it's not really 24bit only 18 apparently. oh well.
the windows 3.1 gui acceleration would be chipset and driver specific. If there's a card out there with the same chipset which has drivers for win3.1 with gui acceleration, then the answer for this should also be yes.

My ET4000/W32I 2MB (ISA) uses Magic RAMDAC (90MHz) : MU9C4910V-90DC (9526B/TA605X)
I added vBIOS of my card in attachment (chip used : M27C256B DIP28, not sure how usefull that will be - but I hope it's complete).
VRAM used : V53C16256HK50 (x4).

Nice project!

Maybe already mentioned before, but does this card have scrolling problems in Commander Keen 3 etc.? I know some ET4000 cards that do.

This project looks really interesting, it's so cool to see all of these new "retro" ISA/VLB video card designs lately.

I have one small suggestion to the author, in case he is planning to work on a new version of the PCB to solve the memory issues mentioned above: it would be a good idea to shift the position of the VGA connector so that it matches with one of the ISA brackets readily available at Digi-Key or Mouser. I believe the current position, which is common on some Tseng ISA cards, is not "standard" and therefore it could be difficult to source proper metal brackets for those who intend to build one of these cards. Not sure if the routing of the traces would be negatively affected by this, it's just a suggestion that I think could add value to the project in the long term.

In any case, nice job! Looks like a tremendous amount of effort.

Are there any differences in vesa bank mode not Lfb mode with the card using interleaved memeory as compared to one not using interleaved memory?
I use Flight Simulator Dos 5.1 which uses 640x400 256color VESA 1.2 banked mode. Sytem memory is set for EMS. The micro-labs site doesnt report a score
increase in Dos of 2 but does give a performance boost from 9-15 with interleave on the same chip though. I understand that vram in does is not a good thing
but dram is. I understand the interleave is meant to simulate vram which says its good for LFB. Curious if any increase on this card for banked frame buffer?

"Apparently the interleaving only works when running accelerated video modes (like in Windows) and has no effect on DOS performance.On an 805 that's okay though, because as I understand it 805 DOS performance is already top notch. I'm really not sure how S3 DRAM controllers stack up against Tseng in terms of speed. Benchmarks say one thing, but real world results seem to indicate they should be about the same." Old post by Anonymous Coward. If he says so I will take his word for it as he is quite encyclopedic on early 386/486 systems.

FYI, I have some 60ns FPM DRAMs on order which I will replace on my board. Matt from DOSReloaded forum has replaced his DRAMs with 70ns FPS and it fixed all the visual artifacts. TSOP44 FPM DRAMs are VERY uncommon and so hard to come by, but replacing the too fast ones with slower ones will probably fix the issues! I will report back once I got my ICs.
Matt also tried out 60ns EDO DRAMs because they are much more common in this package, but the EDOs are also too fast and will give the same artifacts.

No , EDO DRAM is not too fast, it doesn't work with ET4000, which their memory controller works only with FPM memory well.
EDO was my mistake. I thought, it is a FPM DRAM. Datasheet told me: EDO DRAM

I have a dead fujitsu SCSI HDD, i have parted him out and saw: MB814260-70 with TSSOP-44 case -> Oh OK, i can check it on my "Reinfall-ET4000". (german "letdown-ET4000" )
Pictures is much better (of coruse, i have only two 70ns FPM DRAM) and i have disabled second MB memory (second MB = 35ns DRAM)-> artifact is gone, with 33 Ohm Resistor in RAS/CAS line (!)

My ET4000W32i Chip came from compaq corpse and i have with 41c16257-35 extreme artifact. If i press my finger on address line and pictures got a bit better.
It means: my ET4000W32i from compaq has other charge/quality as our ET4000-Chip.

regards
matt

Oh, hi Matt! Anyway, I just wanted to say here that I will replace the 35ns FPM DRAM on my board eventually (and test out your suggestion) as well, as discussed over in the other forum, so people here are also in the loop. 😀

The 60ns FPM DRAMs have arrived! Man those are hard to get! Next time better use a different package! 😁

Will have to see when I have the time and patience to do the replacement!

Hi root42,

Unfortunately I had to pause the electronics hobby for some time as I am caught with other responsibilities. I put on hold all my projects (including the new ISA-PCI system backplane that I was working on for my next retrocomputing experiments).
I haven't had time to read the forum or even send the DRAM drivers to fabrication (although I have the "airwire" circuits apparently working good, in one of my retro PCs -- however I have to make sure the actual design works as expected on the PCBs).

But I am glad to see activity on the project! I haven't read the previous pages but I get that you want to try lower speed DRAM chips. That might actually be a solution and they might work without hassles. The original TSENG datasheet requests for 45-50 ns DRAM ICs, as far as I remember.

Just a hint, you can use a desoldering alloy from Chipquick (I forgot what was its designation) and you can desolder SMD packages, literally within seconds of heating the chip pads with a simple soldering iron.

Cheers,