Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : ATI Technologies in 1997 (T.ATY) -- Ignore unavailable to you. Want to Upgrade?

To: Sleeperz who wrote (2680)	2/5/1999 8:18:00 AM
From: NTT	Respond to of 5927

>The Rage proven what where? It is not even on the PC market yet. The >Rage Fury has been delayed several times already. I'm talking about the Rage128 chip, not necessarily the Fury board. My conclusions come from the tests performed by reviewers given the early prototype boards as well as the now available PC OEM board as well as the 50,000 or so users (Apple estimates) who have the Rage128 in their Apple Yosemite machines. I'm talking about results from REAL shipping boards. >The Rage does even have antriscopic filtering which helps to improve its speed and >slows down on Trilinear. I think you misunderstand here. Anisotropic filtering slows down all chipsets. Yes, the R128 should probably have had it there for the OEM checklist, but NONE of the existing chipsets shipping with it give decent performance with it turned on. The Rage128 is not magically faster because it does not support this feature. Since none of the other chipsets ever use it, they may as well not support it either. There is no penalty incurred for having the feature turned off. None of the industry benchmarks turn this feature on while testing, so it's not really of any consequence. Maybe next year they will when chips can do single cycle anisotropic, but for now it's just a checklist item that is not fast enough for practical real world use on any existing chipset. Also I'm not sure why you say it slows down in Trilinear. It is single pass trilinear just like the TNT. None of the existing chipsets will do single cycle trilinear when multitexturing, including the TNT. This was an illusion created by TNT drivers that did not actually filter the pixels from the two mip maps. Instead it just chose one or the other to render. This method is not true trilinear and hence only takes the hit of bilinear. All chipsets that I know of can be enabled to use this mode, but the result is not as nice as true trilinear. Also, you say the press release says the S3 does TRUE 32 bit 3D rendering. Hmm, and what are the TNT and Rage128 doing? Fake 32 bit rendering? As far as DVD decoding goes, they only support motion compensation. Even on a PII/450, this can still take about 50-60% CPU utilization. On very high bitrate sections of DVD playback, they can even drop frames by going over 100% CPU utilization! This has been shown to be true on a TNT. The Rage128 with its added IDCT engine typically only uses 20% CPU utilization even on very slow CPU's since it handles most of the decode in hardware. So even though Savage4 says it will do DVD playback, it will still occasionally drop frames while playing back and won't support any of the higher resolution HDTV modes. >So your saying the RAGE 128 doesn't fit into the High performance >category either since the S4 offers similiar performance. If the Savage4 was available today, then yes it would be high performance. But when it is released 6 months from now all of the shipping chipsets will be at least double the speed. Hence it will no longer be high performance. I'm basing this on expected ship dates for the TNT2, Voodoo3 etc. >A Rage Pro XL costs $25 the same as the Savage 4 GT. Where do >you get the Savage 4 costs 4 times more?? The RageIIc is ATI's low end part and yes, it costs 1/4 of the Savage4GT. >Well the Savage 4 looks to cost less than the Rage Fury whenever it >comes out. And as I pointed out before the Savage 4 is offering >just good performance at a reasonable price. The key here is not to look at it as if it was available today. Given that the board won't ship for 6 months, there will be new chipsets released that will make the Savage4 look like a VGA card. Would you pay maybe $20 less for a Savage board that gives you half the performance of a TNT2? I don't think so. Mind you I'm making all my performance claims based on the fact that S3 has not released any fill rates for this board. As I mentioned before, if this chip is actually competitive with TNT2, V3 and others in terms of fill rate, it will be a decent chip. With S3 not publishing any numbers for it, it appears to me that it probably maxes out at 250Mp.