SI
SI
discoversearch

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 : AMD:News, Press Releases and Information Only! -- Ignore unavailable to you. Want to Upgrade?


To: Brian Hutcheson who wrote (2918)12/16/1997 11:11:00 AM
From: greg nus  Read Replies (1) | Respond to of 6843
 
Brian, Bottom Fishing Other techs that have made the turn after bottoming...AMAT^...MU^...DMIC^...PS. The three Fools seem to have taken a FOOL'S Holiday. Web Mistress reports she didn't boot the sickman but she must have him tied up and gagged. Same story goes for make me slow. My kind of Mistress!



To: Brian Hutcheson who wrote (2918)12/16/1997 6:20:00 PM
From: Petz  Read Replies (2) | Respond to of 6843
 
Summary of Interview with Atiq Raza (AMD Chief Technology Officer), BOOT magazine, January '98
Front Cover: AMD's multimedia assassin says don't believe Intel's hype
Subtitle: License to KILL

on the 3D instruction set...
Its an extension of single instruction/multiple data, which is required to deal with things like picture composition forom its elemental triangular building blocks, the creation of lighing effects, the creation of 3D effects and textures in a picture.
But quite frequently the composition of the picture is concurrent and interlaced with the physics effect. Because if it is any kind of reality simulation, there is physics behind it that allows that reality to be simulated. That physics runs in the processor. If the geometry was running outside the microprocessor, it'd have the overhead of leaving the processor and coming back in the processor as the physics and creation of the picture were being simultaneously done. At the same time, there's a video effect. The changing of that picture constantly, that video effect could be running in MMX. In today's approach, there's a lot of overhead in doing it, and in our approach that is eliminated."

(translation: no context switching overhead, rather, parallel operation between regular instructions, MMX instructions and 3D extensions)

BOOT: And this is built into the processor. Does this mean that people won't need a 3D accelerator?

Raza: Well, there are several kinds of 3D accelerators. Some accelerate geometry. Some accelerate rendering or set up. We still need the rendering and the set-up function because that is the process of the virtual dots in the frame buffer. That process is most appropriate as close to the frame buffer as possible, and that's where the accelerators are going to excel.

BOOT: What kind of 3D performance can we really expect from the K6-3D?

Raza: Take a Pentium II today -- which is a pretty damn good system -- and compare it with a K6-3D, you will see performance improvement between 50% and as much as four to six times with the AMD part.

BOOT: Six times? Strictly processor level?

Raza: Strictly the processor level.

BOOT: What about frame rates in Quake? What exactly will a stock K6-3D spit out?

Raza: I do not have an exact number. My expectation is that it would perform as well as 3Dfx, potentially better.

BOOT: So you expect the K6-3D to do 50 fps?

Raza: Or better. (note, a review of the P2-300 Compaq Presario 4850 in the same issue shows a frame rate of 24.5 using integrated ATI Rage Pro AGP graphics.)

BOOT: Alright, we'll hold you to that. So can we expect games coming out in 1998 to support the K6-3D directly?

Raza: Yes. We are being rather aggressive about making sure their graphics algorithms used in Immediate Mode are optimized.
...
Many of these companies are also responsible for defining our product. This product wasn't defined in a vacuum. The 1st time we defined it, we went to the game guys -- they threw us out! So we started all over again and took all the things they said were wrong, incorporated them, then looked around again. It took about three rounds before they started saying, "We like this. We want this."


...About yields
...I can't give information that is not shared across the industry.

BOOT: Sure you can! We've heard that one out of every four chips produced by AMD is actually workable. True?

Raza: No, we're doing a whole lot better than that.

...ROADMAP
BOOT: Lets talk about your roadmap for 1998. What are the new chip's delivery dates?

Raza: Expect K6-3D in volume in the first half of the year. Expect a wider smapling in early second quarter, late first quarter of the year. You should also expect a new version of the K6-3D with on-chip Level 2 cache (K6+3D), which will be a one-chip alternative to the entire Intel module, allowing the bus to run extremely fast.

OTHER STUFF
K7 will be running upwards of 500 MHz, backside bus, Alpha bus, competitive with Intel Willamette and superior to Intel Katmai (which is superior to Deschutes which is superior to Pentium II)
Will start move to 0.18 micron in 1999
Will use copper interconnect in 1999

I recommend finding BOOT on your newstand, I didn't see any of this on their web site yet. There are also good articles about x86 compatibility, IA-64
Petz