To: Chip Anderson who wrote (12359 ) 5/8/1999 11:17:00 PM From: SBHX Read Replies (1) | Respond to of 16960
First, I apologize if you thought I was attacking you. I was just chatting, and I kinda admired Hook ever since he made that wicked comment about MS's Direct3D guys from way back. :-) I was wrong about the 250MHz V3. That was a rumour that I heard from fall comdex '98 hall conversations by some game developer. That's why I asked this thread to see if it is true. I stopped believing it when corrected, but it was exciting for a while to think that graphics chips are going for MHz at the same rate as CPU. There were certain elements of truth in higher clocks since the extremely rare [and hard to find ;-)] vastly overclocked 200/175 MHz TNT-2 was reviewed by Tom Pabst.sharkyextreme.co.kr But on closer reading, it does appear that GT is defending V3's 16bpp by claiming that : 0. 32bpp (and renderman) is clearly better quality, but is slower. 1. framerate is more important than video quality 2. TDFX's 16bpp is really 'like' 22bpp. 3. When 32bpp rendering gets a lot faster, 3Dfx would support 32bpp. Various sites have reported that 32bpp frame rates on the TNT-2 and r128 are quite respectable, so I assumed that GT is looking at putting 32bit support in. Anyway, I would be very surprised that V4 did not have some truecolor 24/32bit (or higher) mode, but you're right, I am just guessing and do not have any proof. I'm just speculating that since these new 32bit color modes are running fast on competitor cards, TDFX, being the bleeding-edge technology leader that it is, will not sit idly by and will respond accordingly. :-) I could have sworn I heard GT talk about the 16bpp artifacts in multipass rendering (>2 textures) but since I can't find the article, perhaps I am mistaken. The reasoning for 16bpp artifacts goes like this : 1. Q3A uses lightmaps instead of lighting to the extent that there are up to 3,4,5 passes of textures for most pixels on the screen. 2. When there are 2 or 3 passes, a pixel previously rendered at RBB555 is read back and merged with the new RGB555 color generated in the new pass. Hence each pass will introduce more round off errors. With only 5 bits of info for each R,G,B, it doesn't take many passes for the roundoff errors to become noticeable. 3. This effect is not a problem for games that do not merge previously rendered pixels with the new color. (So lay off the alpha!) 4. This effect is also hard to see if there are lots of motion and frame rate is high, but do show up on still frames (if the game does multipass >2 textures). 5. Even with 8bits each of R,G,B, if you multipass it enough, the errors WILL become noticeable, but it takes a lot longer. I posted heavily in the ATY thread because I was heavily exposed and I also felt the stock was severely undervalued at 12. I am also guilty of being somewhat pedantic and long-winded and too chatty. :-) I don't mind being wrong sometimes. But I do mind (a lot) if I lose money. :-) I bought TDFX because the PE was very low, and it looked like a good buy. I hold more TDFX than SIII BTW, so my exposure here is not in the realm of just 'play-money'. Happy investing. PS1. Please be gentle on that gift of wicked acid prose you posses, save it for real enemies who want to do you harm. :-) PS2. In hindsight, I should have sold at 20, but I don't mind too much since I am still $6 ahead. Here's hoping.