interesting Sigs...hints?

[PXR5]

as I understand it, Frankymail, it's not that there's not enough colors to display; it's accurracy in calculations, things like applying filters in Photoshop, or blending lights in Quake 3--operations like these can cause pixelation/banding even with VCQ, and higher precision makes for less artifacts. Not absolutely sure though, I'm just now trying to get my head around color work after years as a b&w photographer ( see my website for some of the results)

--e

edit: lol, oops, I guess several people decided to
check out my site--& we exceeded Angelfire's bandwidth limits.
Sorry, time to upgrade.
Which reminds me to warn modem users, the images will be slow-loading, as I went for high quality .jpgs. Worth the wait for the difference IMO if you're using a Matrox.
Access may be intermittent. ---e

[Novdid]

You have some really nice pics over at your site.

Keep up the good work!!!

[PXR5]

why thank you--all works in progress, as I learn more.

reminds me though--anyone know where I can find the old "powered by Matrox™ logo? I need to get that on the site, but can't find it anywhere.

---e

[Maggi]

Originally posted by frankymail
Dark gray isn't the same color as light gray, so 32-bits colors really is 4 billions colors to me; using alpha information does create a different color. Take an apple and at look at it; then, go outside and look at that same apple in the sunlight; it's color ain't the same is it?... Anyway, how could it be better than 32-bits colors? IMO, it already produces reality-like picture quality. Can anyone of you guys can see pixelisation in 32-bits colors that is caused by color inaccuraties? Not me... It's like using a Ferrari in the city: pretty useless You think you colors could be more vivid? Stop using a nVidia card and buy yourself a better monitor

Hey Franky,

here's some homework for you



Cheers,
Maggi

[Novdid]

I remember that thread.

Fantastic illustration of your point there, Maggi!!!

[Maggi]

Thanx for the flowers, Novdid ...

Btw, picking up the above mentioned example of 32bpp vs. 40bpp

24bpp color information is equaling about 16.9 million unique colors, right ?

now 30bpp is something like 1073 million unique colors and solely the amount of shades above 1000 million would be already trippling our actual colordepth ...

[Tempest]

Anyway, how could it be better than 32-bits colors? IMO, it already produces reality-like picture quality. Can anyone of you guys can see pixelisation in 32-bits colors that is caused by color inaccuraties? Not me...

Most women can actually see a wider and more accurate spectrum of colours than us men.

And then there are some ppl that see better than others: http://www.redherring.com/mag/issue8...mutant-86.html

[Parhelia Pro256]

Powered by Matrox

[PXR5]

thanks Parhelia Pro!
and Maggi for the link

---e

[frankymail]

Just look at it like this: let's say you're playing a game at 1600x1200@32bits color with dithering ENABLED (and maybe 4x SS FSAA in the future? ) ... Don't you guys agree that there's almost not banding noticable? I don't know about you, but IMO, taking a %25 hit on performances (I presume, because textures would have to be %25 bigger too if they had 40-bits color depth... which does take a toll on memory bandwidth) to improve something I find already pretty close to perfection isn't worth it...

However, If they want to implement 40, 64 or 128 bits for 3d internal rendering only or 2D purposes, I have absolutely no objections, though 128-bits colors might be too much... it's about 4 billions shades of each three colors and alpha :P

I just want to be practical here: I don't mind having a graphic card that produces a pad pixel or two here and there (I'm strictly speaking about 3D here) if that one is pretty fast, features filled (*useful ones*) and reasonably priced.

Anyway, what I would really like to see is some humhum- leaked Parhelia specs -hum post by Ant or others... maybe even a picture of the card? Though that one could cause some problems...

See ya later.

(Do you imagine me working on easter? Well, neither did I until this morning )

[Wombat]

It makes sense to do 10-bits internally. All those passes and blending, that's a lot of rounding error.

It's just like high school chemistry: If you start with 6 significant digits, do your calculations as accurately as possible, but round the answer to 6 digits, too.

[atko]

frankymail you said before, that you are an engineer... so - I think - you know that if you have more bits, then you can represent more digits. If you have more digits, then you will be able to compute more accurate numbers (especially if you have to divide or multiply those numbers).

In 3D graphics the higher precision color depth is very useful, when you have to deal with transparent/transcluent textures on each other. For example the smoke of different things (eg. flying rockets or burning things) is usually made by using transcluent textures on each other. If you use a lots of textures on each other, and you compute the final image with low internal color depth, you will lose a lot of detail because every operation won't be so accurate, because of rounding or truncating the numbers. For example 1.25 powered by 2 is 1.5625. If you round this 1.25 up, you will get 1.3 powered by 2, which is 1.69 (1.7, if rounded). If you truncate this 1.25, you will get 1.2 powered by 2, which is 1.44 (1.4 if truncated). And this is why (afaik) we need higher internal color precision in our forthcoming graphics hardwares...

But I think Parhelia won't have higher color precision than 32 bits (8 bits per channel).

[frankymail]

I have no problem with 40-bits rendering or more, as long as it's done internally, so that there is no a big performance hit. If you start using textures with more that a 32-bits color depth, you seriously start to joepardize available bandwidth. But I repeat, I'm in favor of higher-than-32-bits 2D and 3D (internal) rendering.

BTW, thanks for the multiplication lesson. Maybe tomorrow, you can explain the square root to me, and next significant numbers


C Ya

[atko]

Originally posted by frankymail
BTW, thanks for the multiplication lesson. Maybe tomorrow, you can explain the square root to me, and next significant numbers

For your request, Sir!