I was doing research, trying to answer the question "Which was the first GPU to support 24-bit color". I know all color since 1992 is 24-bit, even in games like Doom. I mean 16 million simultaneous colors on the screen, not just 256 unique colors from a 24-bit palette.

I started digging, and naturally I came across the ATI Mach 32. Later I find out that RIVA TNT also "added" support for truecolor. So I'm left wondering, is 24-bit color support some ancient technology that was forgotten after 1992 and rediscovered in the year 1998? Or are they talking about something different?

I have two guesses, but I'd love to know the real explanation:

  1. Truecolor support in RIVA TNT meant it's hardware accelerated, as in the sprites are stored in the VRAM, as opposed to the Mach 32, where the VRAM is just a frame buffer so acceleration would be considered software.
  2. Nvidia meant 32-bit color texture, not even talking about frame buffer pixel depth.

Anyone know what both Nvidia and ATI really meant?

  • 1
    \$\begingroup\$ Simple answer : the Mach 32 isn't a GPU. To define a GPU, you need to talk about having a graphics pipeline with stages such as input assembly, vertex shader, primitive assembly, fragment shader, etc, and the GPU implements all or part of that pipeline in hardware. The Mach 32 does none of that. None. \$\endgroup\$ Sep 13, 2020 at 11:06
  • \$\begingroup\$ Interesting. Okay, but in that case the PS1 GPU was the first official GPU in 1994. It did have a 24-bit color mode, so to reiterate my question, why did they add truecolor support in 94, forgot about it, and reintroduced it in 98? \$\endgroup\$ Sep 13, 2020 at 12:39
  • \$\begingroup\$ Let's not get into extended discussion in comments, but I suggest you go back and re-read the Wikipedia article you linked - adding support for 32-bit colour is by comparison to 3dfx, which in common with other early consumer 3D accelerators only had 16-bit colour. There was no "forgetting", and to be honest you're splitting hairs over semantics in a Wikipedia article here. \$\endgroup\$ Sep 13, 2020 at 16:37

1 Answer 1


The Mach32 chip uses 24bit color, with 32 bits alignment, and it is not intended for 3D graphics. For what I read, using Mach32 cards required a resident software in DOS that slowed DOS down (I'm guessing this is true of every card).

Cards using the Mach32 chip add VRAM. I believe the chip does not have a texture sprite cache. I would like to see an schematic of the chip.

On the other hand RIVA TNT is capable of fully accelerated 3D graphics with 32bit color pipeline. Also, unlike Mach32 card, RIVA TNT supports DirectX 5.0 and OpenGL 1.2.

More about on Mach32:

More about on Riva TNT:

  • \$\begingroup\$ Thank you for your answer! I'm not sure I understand what "fully accelerated 3D graphics with 32bit color pipeline" means though. I mean a frame buffer is all that's required to blit pixels onto the screen. The pixels themselves aren't 2D or 3D. Are you saying the first 3D accelerators weren't able to produce a 24-bit color pixel on screen? Maybe this 32-bit support refers to precision of calculations and not color depth? For instance it's one thing to multiply floats and another to use doubles, the would all fit in the [0, 1] range but doubles have a greater precision. \$\endgroup\$ Sep 13, 2020 at 7:15
  • 1
    \$\begingroup\$ @BorisRusev What I mean with 32bit color pipeline, is that it loads 32bit color textures, does 32bit color operations, and outputs 32bit color. Were there 3D accelerators prior to that one that were not able to produce 32bit color? Yes, they would use 16bit color. It might be worth noting that the RIVA TNT depth buffer and stencil buffer had less precision. \$\endgroup\$
    – Theraot
    Sep 13, 2020 at 7:30

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .