Re:Why 3dfx 1 cards had such a washed out color output?
by hmstx on Tuesday November 17, 2026 @02:30PM
(#50536884)
On low precision in darker colors being a function of gamma...
There were issues in Quake/Quake 2 with bright areas not being quite right, so some measures were taken to try and achieve some form of overbright lighting [1]. Especially in Q1 where the light baking did not perform bounces/radiosity, there were a lot of black spots with adjacent highlights for dramatic effect, and you could see some banding there with the OpenGL renderers, which is more apparent with lower texture settings and pumped gamma.
But also, as a side effect of this overbright ability, lighting got flattened even more in GL [1]:
>The original quake (software) engine used overbright lighting, which means the lightmap brightness can go up to 200%. So, light.exe creates BSP files with lightmaps that go up to 200%.
>In glquake […] there is no overbright lighting, so every part of the lightmap that goes above 100% is flattened to equal 100%. This is why glquake looks just fine in darker rooms, but in bright areas the lighting looks very flat.
>[Some engines] restore the overbright feature of the software renderer, so that lighting hot-spots will actually be as bright as intended.
A similar effect was pursued further in Quake 3 with r_overbrightbits/r_mapoverbrightbits, which fakes a higher dynamic range through futzing with the hardware gamma ramp and bit-shifting/normalizing color channels [2], achieving highlights that pop a lot more... at the expense of introducing even more banding in dark spots.
He mentioned it in passing in a .plan in 2000 that touches on higher precision buffers [3]:
>Q3 hacks the gamma tables to sacrifice a bit of precision to get a 0.0 to 2.0 range, but I wanted more than that for even primitive rendering techniques.
Eventually, in his .plan dated january 29th, 2003 [4] he appears convinced,
>The future is in floating point framebuffers. One of the most noticeable thing this will get you without fundamental algorithm changes is the ability to use a correct display gamma ramp without destroying the dark color precision.
There were issues in Quake/Quake 2 with bright areas not being quite right, so some measures were taken to try and achieve some form of overbright lighting [1]. Especially in Q1 where the light baking did not perform bounces/radiosity, there were a lot of black spots with adjacent highlights for dramatic effect, and you could see some banding there with the OpenGL renderers, which is more apparent with lower texture settings and pumped gamma.
But also, as a side effect of this overbright ability, lighting got flattened even more in GL [1]:
>The original quake (software) engine used overbright lighting, which means the lightmap brightness can go up to 200%. So, light.exe creates BSP files with lightmaps that go up to 200%.
>In glquake […] there is no overbright lighting, so every part of the lightmap that goes above 100% is flattened to equal 100%. This is why glquake looks just fine in darker rooms, but in bright areas the lighting looks very flat.
>[Some engines] restore the overbright feature of the software renderer, so that lighting hot-spots will actually be as bright as intended.
A similar effect was pursued further in Quake 3 with r_overbrightbits/r_mapoverbrightbits, which fakes a higher dynamic range through futzing with the hardware gamma ramp and bit-shifting/normalizing color channels [2], achieving highlights that pop a lot more... at the expense of introducing even more banding in dark spots.
He mentioned it in passing in a .plan in 2000 that touches on higher precision buffers [3]:
>Q3 hacks the gamma tables to sacrifice a bit of precision to get a 0.0 to 2.0 range, but I wanted more than that for even primitive rendering techniques.
Eventually, in his .plan dated january 29th, 2003 [4] he appears convinced,
>The future is in floating point framebuffers. One of the most noticeable thing this will get you without fundamental algorithm changes is the ability to use a correct display gamma ramp without destroying the dark color precision.
[1] https://www.quaddicted.com/engines/software_vs_glquake
[2] You can probably read the C/OpenGL code better :)
[3] https://github.com/ESWAT/john-carmack-plan-archive/blob/mast...
[4] https://github.com/ESWAT/john-carmack-plan-archive/blob/mast...