Thank you DEAT.

Hello
I have 2 Diamond Stealth video 2500 promotion at24 pci card with bios v1.03.
I tried with modded 2052 and 2072 drivers, with Windows98SE
Both card after installind driver and reboot after windows 98 splashscreen give an interesting pixelated picture.
What was wrong?
With which driver can i use this card?
Or windows98 is the problem?

Thank you

It looks like the drivers clocked the memory way too high. You can adjust it with third-party tools like Powerstrip.
But AT24 does not belong in this thread about 3D cards.

Putas wrote on 2025-02-10, 18:33:

It looks like the drivers clocked the memory way too high. You can adjust it with third-party tools like Powerstrip.
But AT24 does not belong in this thread about 3D cards.

I only here find any information of drivers, and here was hte modded driver, what i tried:
Re: Worst fastest early 3D cards questions

Re: Worst fastest early 3D cards questions

Drivers may work with an AT25 and aT3D card but the AT24 wasn't really the same and even if working they were different projects; even the AT25 might simply be a rebrand of the aT3D or probably a lighter aT3D revisited project to be sold as cheaper 2D only accelerator or somewhere in the middle.

Yes, sorry i was confused.
I think alliance at24/25/at3d is the same chip.
I read elsewhere about Diamond stealth video 2500 is a dx5 card, but with original driver no direct3d

Hello. My Alliance AT3D (Supergrace VP-128) only works with 2072 driver. The 2073 supplied by DEAT and Putas gives me this screen each time I start a Direct3D Application. I already tried clocking the card down to 50Mhz (driver default is 60mhz in registry). And I tried setting windows resolution to 640x480 16bit (same as d3d Application). What else could I try?

I've been testing a lot of cards lately that were mentioned in this thread, and I have come up with an idea for a project. I don't really have time to do it right now, but I would like to if I can figure out how to get reliable data and arrange it in a useful way.

Basically, I would like to set up a system or two from 1996-1998, and then compare these early 3D accelerators with regard to Windows GUI acceleration and output quality, as well as 3D accelerated performance at low resolutions. I would also like to include examples of the fastest and slowest versions of the cards, since I have found that performance can vary by an almost ridiculous amount with some of these cards. The goal is to determine what kind of graphics settings and resolutions (for desktop or for games) will provide the best performance for a given card.

Believe it or not, the companies that made these cards actually intended for them to be used for games and they spent a lot of resources to do this... so if we find them worthless maybe we're just using them for the wrong thing.

So, it would be performance focused testing, catered to the capabilities of each early 3D accelerator, designed to make sense for people who play games now, rather than aiming for the performance standards of 1997.

It's a bit hard to describe, but... I have noticed from recent discussions that a lot of people have really low standards for frame rates... to the point that any comments about what is good\bad playable\unplayable are kind of meaningless to anyone that expects a responsive gaming experience. Like, I'm not going to play a game at 1024x768 at 17fps average. I am just not going to do that. I am going to turn the resolution down until it runs smooth enough to not hinder gameplay or input responsiveness.

Let's face it, if we want to play games from 1996-1998 at 1600x1200 with anti aliasing and absolutely flawless performance (meaning, FPS = refresh rate or higher) we can just run the games on newer hardware from the early 2000s. So, if you want a somewhat period correct machine, what do you do? Just run at 1024x768 or 800x600 and consider everything slower than a Voodoo 3 or TNT 2 Pro worthless? Why not embrace the pixels and keep lowering the resolution? Doing so opens up a whole bunch of totally serviceable 3D accelerators that are just too weak to handle even 640x480 smoothly.

Part of what inspired me to do this was attempting to run a fairly light 3D game at 640x480 on some crappy low end card and ending up with a slide show, and then switching to 400x300 and having a buttery smooth experience with a surprisingly appealing pixelated look that feels totally okay for a retro PC. Some cards actually had severe graphical corruption at 640x480 because they did not handle running out of video memory well, which gives the impression that they are unusable garbage. 2MB S3 Virge DX cards had this issue... and yet, going to 400x300 made them look good and run great.

Anyway, long winded post to say... I want to do a poor man's period correct late-90s gaming PC test, with a focus on performance and responsiveness. Eventually.

EDIT: Also, the amazing work done at Vintage 3D is very relevant to this, but my testing and results would be more general per-card while hopefully giving a picture of what each card is good at in comparison to the rest. Like, if you've got a pile of dirt cheap cards that you can't even give away, you'll know how they compare in a practical sense for the things they're actually good at.

I think that is the right direction and hope the vision can be fulfilled. Sounds like years of work.

As Virge's own library shows, most devs would not be bothered with higher framerates enough to support lower resolutions. Search for outliers, but better get used to that for such a project.

I have run Rage2 with 2MB and on 3D it cant go higher than 400x300, Rage2 4MB cards can go higher, but Rage2 cant do 640x480 or 800x600 well.

G-Police is at 15bit vs 16bit colors
and Turok 1 screenshots at Rage2
Re: ATi RagePro drivers and software

Some early low end 3D cards here: https://read.cash/@Geri/low-end-3d-cards-from … he-90s-2d4beee8

I am pondering Rage II at the moment, got an unused P200 classic and thinking about a "best of mid 1996" style build. Pending finding suitable parts in the stash. Not gonna be one I'll spend money on.