First big game I know that required VBE extension was SimCity 2000 (1993), It shipped with 1.2 VBE extension for tons of vendors including Cirrus (CL-GD5420 CL-GD5422 CL-GD5424 CL-GD5426 CL-GD5428 CL-GD5429).

Linear mode (LFB) is VBE 2.0 released in December 1994. Ken Silverman made NOLFB utility to help run games taking advantage of VBE 2.0 LFB under Windows XP where LFB is broken - this gives us a nice list of games that can use it NOLFB - Disables LFB enabling DOS games to fallback to VESA 1.2 modes
all that list plus Quake and FastDoom.

feipoa I found your Re: Photon Torpedo with greenish display and Photon Torpedo 564-TEK is a 1:1 copy of STB Nitro 64 ISA https://www.vgamuseum.info/images/pepino/cirr … nitro_isa_f.jpg Much cleaner routing, 4-layer like Diamond but RGB Video always on top layer all the way to video connector with everything else kept away from analog video tracks. Im starting to wonder why did I always believe Diamond made good quality VGA cards 😮

Btw both GD5434 based Diamond SpeedStar 64 and STB Nitro 64 ISA/Photon Torpedo 564-TEK wire to MEMR/MEMWR meaning working VBE 2.0 linear mode.

Ironically Newly made ISA CL-GD5434 https://github.com/Leshak-pcb/Pine-CL-GD5434 doesnt.

Reading CL-GD5434 Layout guide they werent so sure about isolated ground islands anymore and tried the "its magic" line
"There is a certain amount of art involved in the exact positioning and size of the cuts in the ground plane and the power plane. Some experimentation may be required to obtain satisfactory results"

then later say "The power plane and ground plane cuts must follow each other. It is critical that an isolated ground or power plane not overlay a noisy digital power or ground plane." while attached images show exactly that - DAC/IREF isolated ground crossing VDD :] A bit of a clown show.

Fancy detective work. I forgot about that Photon Torpedo vs. Diamond SpeedStar 64 posting. I was able to take better photos in 2016 :) You can clearly see the vertical lines when you open up the Diamond image in that older Vogons link.

It also sounds like I should hunt down a STB Nitro 64 BIOS for my Photon Torpedo card so that the card is more authentic -those two PCB layouts look identical.

Seems to me that the best Cirrus Logic-based ISA graphics card to replicate is the STB Nitro64. Anyone have its VGABIOS and who would be willing to upload it?

in the Re: Newly made ISA CL-GD5434 The Serpent Rider also says "I checked STB Nitro CL-GD5440 based on this design http://www.vgamuseum.info/images/vlask/cirrus … l-gd5446stb.jpg - no lines." so maybe STB employed good layout guy while everybody else copied reference design.

I know that some games could be forced to not use LFB e.g. by noLFB util (and some programs may require LFB by design) but the idea was to enable LFB for faster VRAM access when HW allows it by linear mapping.

kixs wrote on 2021-02-28, 16:19:

Here is the BIOS of my Nitro64 ISA card saved via NSSI.

OK, so there's the Nitro 64 ISA VGABIOS!

keropi wrote on 2020-11-06, 11:06:

I just finished soldering my card - it works nice
I do get those vertical lines with either BIOS (diamond/generic)
It would be nice to know the source of it....

kixs wrote on 2020-11-06, 11:22:

Many cards have these vertical lines on LCDs... I've tested several S3 VLB cards these past days...

Since this other guy's homemade GD5434 card also has issues with vertical banding, seems like it is related to the PCB design. I'm probably not going to invest in this project since the stripes exist. I can get another Diamond GD5434 for $100 which has banding already; I don't need a 3rd one. An asset to the community would be an STB Nitro 64 ISA replica. Once you see the clarity on an LCD for yourself, you want nothing else.

If Madao's GD5429 card, or rasz's variant, doesn't exhibit the vertical banding, I'll probably assemble one. Looking forward to some anecdotes here.

Dont manufacture pcb with my changes! It was meant strictly for educational purposes and only touched layout. There are still things worth taking care of, for example design reference asks for decoupling on every VDD pin, but the project replaced 2 caps with 10uF tantalums which are not so good at high frequencies. Whole MEMR/MEMWR thing. Personally I would get rid of second MB as it does nothing for speed nor high resolutions, remove VESA connector, and order second pcb version with split grounds like Cirrus requested for comparison.

Don't you need 2 MB to do 1024x768x65k ? or is that only with interlaced mode? I normally run 1024x768x65k in Win3.1 on ISA 486 systems.

2MB has no use in DOS. But it's a must for Windows. One is more resolution choice, the other is speed improvement.

Second megabyte on CL-GD5426/'28/'29 unlocks 256 colors in 1280 x 1024 (interlaced), nothing else.
In CL-GD5434 2MB not only unlocks maybe useful 32K 1024 X 768@75, 256 1280 x 1024@75, 16M 800 x 600@60 but also doubles data bus width probably also making blitter faster.

rasz_pl wrote on 2024-12-16, 07:16:

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

I thought they were talking about "jailbars" that show up on LCD's but not on tubes--

Re: Let's improve video output quality of VGA ISA/VLB cards

Granted this is a pretty long thread and I have serious covid brain at the moment, but ... is anyone actually building these? I'd love a new VLB GPU in my 486 rig!

douglar wrote on 2024-12-18, 19:32:

I thought they were talking about "jailbars" that show up on LCD's but not on tubes--
Re: Let's improve video output quality of VGA ISA/VLB cards

LoL, a third time STB card using Cirrus Logic produces perfect picture!
Jail bars wont be caused by RGB routing/filtering. Those affect blurriness/sharpness of image.
Stable vertical jail bars at pixel boundaries fixable by swapping DAC chip exclude PLL/monitor sampling frequency as the source of problems. PLL issues would result in walking irregular interferences, not perfect stripes.

What is left and what weedeewee pointed out almost immediately in your thread is DAC iref. Stripes look like IRef slightly dropping every second pixel in 320x200, maybe corresponding with main VGA chip fetching data and momentarily drawing more power?

All STB cards use Q1 LM334Z with sot23 CR1 BAV-something diode, 220(or 150 depending on the card) and 20 ohms divider and nothing else. Perhaps crucially they
- dont link IREF with caps to main chip VCC like reference design suggests
- looks like STB didnt buy the divided ground plane BS and DAC iref circuit is build over solid common ground
- main chip is properly decoupled with 100nf caps all over the place

Most other Cirrus Logic cards use Q1 2N3904 + LT431 (your Diamond) or also LM334 with diode, but stick to split grounds from reference design and omit some decoupling caps due to routing constrains.

If you fancy playing with it you could put a scope probe in x10 mode on pin1 of Q1 on the STB card and pin 3 Q1 on Diamond one, both connect to chip pin 78 IREF. Crucial for scope ground to be as short as possible, preferably using spring clip https://electronics.stackexchange.com/questio … e-ground-spring, closest ground is on chip pin 80 right on the corner. You can also look at chip VCCs on pins 1 26 58 81 107 121 140 and DAC VCC 72 85. If iref theory is correct you should be able to see regular drops on pin 78 and maybe 72 85. Bars on your screenshot show up every two pixels and are around 1/4 of a pixel wide so those bumps should be repeating at around 6MHz and last ~20ns (40-50MHz)?

EDIT: wait a minute, that 40-50MHz frequency I estimate from screenshot looks funnily familiar. CL Memory clock is one of

37.585 MHz
41.165 MHz
44.744 MHz
50.114 MHz
only CL-GD5429:
57.273 MHz
60.852 MHz

so it might just be insufficient VCC decoupling after all causing momentary DAC supply dropout.

My working theory based on seeing differences between Trident cards on the same PCB with only revision difference of chipset and same DAC, is that due to seeing faint bars on the card that was dimmer on same monitor/motherboard, that it was something to do with drive strength of the signal. Some people have mentioned bridging inductors in the output.

That might make whole screen brighter, even bad impedance matching (it would take 1.7 mm wide RGB tracks for perfect 75 ohm impedance) cant explain regular sharp vertical strips.

Hi, all.

I'm modified highspeed rom to make it little bit faster. Could you please try it?

Hi Madao. I just wanted to order the PCB for S3 Trio64V+ VLB. I'm using gerbers, which are on your github. Unfortunately I've found some issue with Silk Screen on Bottom. It seems the text, which suppose to be placed on Top under the Video BIOS chip, is loated on Bottom and is mixed with the other text. Is there any chance to have this updated? I would like to avoid modifying your work by deleting this portion of gerbers, especially that this text is with your signature!

zami555 wrote on 2025-03-11, 19:45:

Hi Madao. I just wanted to order the PCB for S3 Trio64V+ VLB. I'm using gerbers, which are on your github. Unfortunately I've found some issue with Silk Screen on Bottom. It seems the text, which suppose to be placed on Top under the Video BIOS chip, is loated on Bottom and is mixed with the other text. Is there any chance to have this updated? I would like to avoid modifying your work by deleting this portion of gerbers, especially that this text is with your signature!

Cleaned gerber file is uploaded.

I haven't regonizied this typo, because chinese pcb manufactur has removed one of both string.

Thanks
matt

Yftul wrote on 2025-02-25, 07:25:

Hi, all.

I'm modified highspeed rom to make it little bit faster. Could you please try it?

Is this rom with set "Waitstate zero " ?

Madao wrote on 2025-03-11, 20:08:

Cleaned gerber file is uploaded.

Wow, that was pretty fast, thank you very much 😀 I've double checked it and looks good now.

Madao wrote on 2025-03-11, 20:09:

Is this rom with set "Waitstate zero " ?

Hi, Madao..

I used the original rom and reduced some constants.