I had cards with specific ISA capable ATI64 chips (afaik only ones with external DAC) in my auction watchlist and they pop up rarely at $50-100. Too rare and expensive for a donor, would be easier and cheaper to build FPGA card emulating it from scratch.

I think that writing SVGA and some 2D accelerator implementation in FPGA/VHDL would take some hundreds manhours of development. I even don't have such skills. I will check what version of Mach64 PCI I exactly have. But if I remember well it was a single chip full-integrated VGA...

Hm, no luck, mine is ATI Mach64 264VT2, according to table at http://en.wikipedia.org/wiki/List_of_AMD_grap … rocessing_units
it's PCI-only. It's needed to have Mach 64 CX or GX, only these supports ISA, VLB, PCI.

I bought one GX for cheap, to be donors for my VLB card if it ever were to need spare parts for whatever reason. Perhaps the chips can go to a well designed ISA card too...

Should be. Maybe it would be possible to design some combo ISA/VLB card where the VLB tail would be dettachable via some connector (to prevent collision on ISA-only MB). 😀

ISA/VLB Combo Card is possible. But Layout should being designed to "flip-card" (Like ELSA Winner XHR 1000 with S3 928 and EISA/ISA )

Back to topics.
Critisim about ground and 2 layer layout hears a bit desctrutive. It was a bet and work time on card is only 1 3/4 days. Here is some neglect. Ground pane with needle is auto-generated by KiCad, but this doesn't make me worry.

I have a newbie question: What's the motivation for creating an ISA card based on the GD-5429? Are they too expensive to find in the 2 MB version these days? I see some 1 MB cards on eBay for about $50. What's the cost of all the components needed to assemble one of these cards?

Why was the GD-5429 selected over the GD-5434? Is there any benefit? I saw some GD-5434 cards selling for about $100 a few months back. Cheapest now is $125.

Can the GD-5434 be used in place of the GD-5429?

rasz_pl: Where do we find your updated gerbers for this project?

Concerning the Mach64, I think a VLB replica with 4 MB would make a really nice card for use in Windows. Even the 2 MB cards have become unaffordable.

100 usd for a GD5434 ISA:.. no chance in germany..
50 usd for a GD5428/9 ISA in germany/europe.. not easy to find.

Mach64.. i wouldn't make it, with two reason: no chip aviable and mach64GX/CX PCI is not very common as like Mach64CT/VT2

I see some IC prices online:

CL-GD5429: $14.92 USD
CL-GD5434: $22.34 USD

Can I use the 5434 on Madao's PCB?

No, GD5434 has difference by footprint (QFP-208 vs QFP160 by GD542x )

Of coruse, Cirrus Logic Alpine familiy (543x) has 32bit direct VLB Interface and 64bit Memory access.
542x has 16bit VLB inteface and 32bit memory interface

OK, thanks. Is there any benefit in using the GD5429 over the GD5434, aside from the $7 cost difference in the chip?

Madao wrote on 2024-12-12, 07:34:

Critisim about ground and 2 layer layout hears a bit desctrutive.

I repeat I am not pooping on your hard work! My feedback Is meant to be constructive. Yes your PCB works. Picking up some additional insight and new tricks will let you design even better ones in the future.

Madao wrote on 2024-12-12, 07:34:

but this doesn't make me worry.

Imo
"some 542x chip has vertical noise/stripe on TFT. (It is not pcb issuses)"
"some Chip has noisy VGA output (stripe on TFT display)"
is caused and can be mitigated by PCB design. Stripe means either jitter or more likely interference.

My second try https://github.com/raszpl/CL-GD542X_ISA

RayeR wrote on 2024-12-12, 03:37:

Hm, no luck, mine is ATI Mach64 264VT2
It's needed to have Mach 64 CX or GX, only these supports ISA, VLB, PCI.

Yes, only mach64 with external DAC are ISA/VLB capable 🙁 Rare and expensive.

rasz_pl wrote on 2024-12-16, 00:47:

My second try https://github.com/raszpl/CL-GD542X_ISA

Do you have any interest in altering the PCB for the GD5434 chip?

rasz_pl wrote on 2024-12-16, 00:47:

Imo "some 542x chip has vertical noise/stripe on TFT. (It is not pcb issuses)" "some Chip has noisy VGA output (stripe on TFT di […]
Show full quote

Imo
"some 542x chip has vertical noise/stripe on TFT. (It is not pcb issuses)"
"some Chip has noisy VGA output (stripe on TFT display)"
is caused and can be mitigated by PCB design. Stripe means either jitter or more likely interference.

I'm not sure we my "stripes" are the stripes being referenced here, but I have a Photon Torpedo GD5434 and a Diamond GD5434. The Diamond has some striping or banding at particular display modes. If I put the Diamond's BIOS into the Photon Torpedo PCB, those stripes go away. Hints and PCB design differences?

It can be explained that both vBIOSes use some way different timings, e.g. due to different DRAM speed and it may cause that some specific frequency is amplified by PCB layout less than other frequency and all that lead to interference and visible stripes. Good layout with solid GND should make it less sensitive to varisous frequencies and should behave more consistently...

Depends on what you call stripes, but with only bios chip swap might be as simple as another bios programming different frequencies which your particular LCD screen is able to sync easier.

Horizontally flickering image from video card?
http://techmind.org/lcd/phasexplan.html

From your second URL, I suppose it would be (a) Clock/Pitch incorrect, but no matter what I adjust on my LCD, the vertical striping doesn't go away when I'm using the Diamond SpeedStar 64 (GD5434). If I use that same Diamond SpeedStar 64 VGABIOS in my Photon Torpedo (GD5434), the image looks perfect on my three different branded LCDs.

Here's some photos of the banding/striping from my LCD. It is difficult to get an image of and this is the best I could do.

It is easiest to see the vertical banding on the desktop computer flying around in 3Dbench. You will need to zoom into that photo:

as well as in the checkered floor, top right of the image:

Correction, CL has only a couple fixed video clock frequencies so thats not it. It could be different phase (porch size etc) or video ram speed setup?
Diamond always had good reputation. I looked at some Diamond cards and for example Diamond SpeedStar PRO ISA (https://www.vgamuseum.info/index.php/cpu/item … rrus-logic-5426) is 2 layer made very closely to recommended layout (not sure if thats great as they copied split grounds). Analog RGB goes straight to VGA connector on top layer so thats good.
Diamond SpeedStar 64 https://theretroweb.com/expansioncards/s/diam … peedstar-64-isa on the other hand uses 4 layer PCB so should be better, but somehow someone decided to push Analog RGB thru 3 VIAs going under/over/under, added useless thin guard traces and pulled ground noodles between analog video tracks. Another not so great thing I noticed is they use thermal relief vias for ground stitching, you only use those if you intend to solder. They dont couple heat to the larger mass so soldering process is more controllable, but the downside is they make holes in ground plane. Diamond kept doing this up to 1998 😮, finally cards from 1999 use normal vias.

>If I put the Diamond's BIOS into the Photon Torpedo PCB, those stripes go away. Hints and PCB design differences?

Wait, I think I understood that the other way around 😀 What happens if you put Photon bios in Diamond? still striping?
Hard to tell from the photo due to moire, but second picture it almost looks like vertical strips are walking back? I would try another LCD, this might just be monitor deficiency.

I only have so many LCD's. In my previous post, I noted that 3 differently branded LCD's all exhibited the same symptoms.

No, the vertical bars aren't walking, they are stationary.

Almost all ISA graphics cards have some vertical strips evident in DOOM, 3DBench, and pcpbench. I thought it was normal until I tested the Photon Torpedo (w/Diamond BIOS) and, I think, the Mach64 ISA.

I don't recall why I'm not using the Torpedo BIOS in the Torpedo graphics card. I strictly use the Diamond BIOS in the Torpedo. I just tried the Torpedo BIOS in the Diamond, but screen stayed black, and no POST beep.

Maybe Torpedo bios is programmed with higher ram clocks and crashes card during initialization? Dont know if that would explain not POSTing at all.

Madao Have you experimented with MEMR/MEMW SMEMR/SMEMWR? Documentation states connecting to MEMR/MEMW will enable linear mode. Almost all CL ISA cards have it connected to SMEMR/SMEMWR like in eval design, so is your pcb. BUT there are exceptions:
https://www.vgamuseum.info/index.php/cpu/item … logic-cl-gd5422 MicroMax PCBs with clear "CL-GDS42X EVAL BOARD 1.0" print and
https://www.vgamuseum.info/index.php/cpu/item … logic-cl-gd5424 AVGA3B (same eval PCB) have a footprint for jumpers just above C27 letting you select between those two options.
and the greatest weirdest find of them all https://www.vgamuseum.info/index.php/cpu/item … logic-cl-gd5420 ATIF-VGA, Cirrus Logic based ATI card?!?! 😮 MEMR/MEMW wired in from extended portion of the slot.

I dont quite get why would anyone wouldnt want to wire it to MEMR/MEMWR? Especially considering

"An adapter board based on this schematic will not function in an 8-bit ISA connector; the
schematic shown is for an implementation of the CL-GD542X for a 16-bit-only ISA bus In-
terface. The adapter board is limited to a 16-bit ISA interface because SA17, SA18, and
SA19 (from the 8-bit connector) are not used and are replaced by LA17, LA18, and LA19
(from the 16-bit connector)."

and Chip pinout clearly was designed for MEMR/MEMWR, going for SMEMR/SMEMWR forces going all the way to the other side with those two signals.

All the cards are wired using LA17-19 so wont work in 8bit only slot anyway. MEMR/MEMWR sounds like a no brainer, you dont lose anything, you gain linear mapping in 15-16MB hole in VESA modes.

Hm, interestnig about the linear memory mapping. Was this a specific CL chips feature or more generic also for other chips? I saw this option of memory hole in various SETUPs many times but didn't know what it would be good for, never had a SVGA that would utilize it.
So then it would be possible to write a TSR that would implement basic VBE 2.0 extensions which will return this fixed LFB address on specific function call and properly initialize requested video mode to use this LFB...