ATI Meeting

ATI

We had to split up the FiringSquad team for the Tuesday meetings, due to scheduling conflicts. I took the 10AM ATI meeting, Brandon had VIA and Kenn was still recovering from the late night Voodoo 4/5 Comdex update he didn't finish until early in the morning.

The ATI booth was on the exhibition floor of the Las Vegas Convention Center. There were several demonstration units running on various ATI cards along with a nice glass (or plexiglass) cabinet with all the ATI cards on display.

Umm, yeah.

The ATI meeting situation was very similar to the Nvidia and Matrox meetings. There was nothing really new to talk about. We had already tested the Rage Fury MAXX card about a week before, and ATI definitely wasn't going to talk about their next generation 0.18 T&L part. Since we already knew everything about the Rage Fury MAXX and we couldn't talk about future products (our bread and butter), we had to find something else to discuss.

Curious about the power of the Rage Fury 128's hardware DVD, I asked what was the slowest CPU that could support full DVD playback at 30 frames per second. The answer? A Pentium II 233! That's very impressive considering that a few days earlier we were testing the Savage2000's DVD playback, and it was giving us slight hiccups on a Celeron 300A overclocked to 450MHz. Granted, the Savage card came bundled with the Zoran SoftDVD player, which is arguably, one of the worst software DVD players out there, but full playback on a PII 233 is still pretty impressive.

MAXX latency?

Next, I moved onto the Rage Fury MAXX framerate latency issue. Right after we put up our Rage Fury MAXX hands-on Preview, some readers had written in asking about the latency issues associated with ATI's Alternate Frame Rendering (AFR) technology. If the card was rendering two frames at a time, doesn't that mean one frame is always late? Others had raised this same question the day before, and ATI still hasn't formed an official response, but unofficially, one ATI engineer told us that the latency would be no worse than enabling triple buffering. Yes, there is a slight delay, but with double the framerate, the latency should be negligible. Hopefully, we'll receive an official response sometime in the near future.

The latency isn't a problem with only 2 chips, but it would become an issue with more than two chips. ATI freely admits that 3dfx's multi-chip solution is far more scalable than AFR. The VSA-1000 multi-chip architecture can scale up to 32 processors, but AFR technology can only handle two chips. ATI argues that once you move beyond a two chip solution you enter into the tiny super high-end market. This makes sense. How many people are actually going to pony up $599 for the quad-chip Voodoo5 6000?

How's business?

We ended the meeting talking about ATI's market share status. With NVIDIA and 3dfx trying to take a piece of ATI's OEM pie, we were curious about ATI's market share. Here's the 3D Graphics Market Share from Mercury Research 1999:

ATI 38%
• S3 17%
• Nvidia 12%
• Matrox 12%
• 3DFX 8%
• Intel 6%
• SiS 5%
• Others 4%

ATI's is still in first, more than doubling the next closest competitor, S3. Okay, time for the next meeting! Let's see what Brandon discovered at the Aureal, Turtle Beach, and Yamaha.