No teasers -Nvidia meeting notes E3 2002

ben6

Regular
Arrived at the Nvidia booth around 3:50 in the afternoon , after a tiring day of walking around the E3 show floor. Upon arrival talked briefly with Derek Perez and Carrie Cowan, PR people for Nvidia and awaited Brian Burke's conclusion of his previous appointment.

Nvidia's booth was a interesting appointment only area with computers and Xboxes all over the general area.

The following games were on display: UT 2003, Project Gotham Racing, Counterstrike, Condition Zero , EarthWeb, Tomb Raider 4 and Sea Dogs 2.

Brian arrived from his prior appointment and we headed off to see if we could find a conference room. We found a corner to talk (all the rooms were occupied) and here's the main things we talked about

1. Ti4200 was released recently. With a MSRP of around $179 , it brings Geforce4 features and great performance to the mainstream market. (expect a Visiontek 128MB 4200 review from me very soon)
2. Nvidia owns around 2-3% of the chipset market
3. Nvidia is working on all kinds of 3d chips while a PDA chip was not specifically mentioned it's a logical next step for Nvidia, in the future, core logic and the successor to the Nforce
4. Nvidia is interested in a P4 bus license of course . (Public statements by Jen-Hsun Huang the CEO state that the main obstacle to this is the high licensing fees asked for by Intel and competing with Intel's own chipsets)
5. We talked in general terms about the future. AGP 8x and DirectX 9 are likely to be the big things pushed this fall.
6. Brian called the Parhelia a interesting product with interesting new features . He questioned the ability of Matrox to mass-produce the chip at high clock frequencies as it's .15um and over 80 million transistors (Matrox is targeting 220mhz as of this writing fyi) .

7. Brian made a further comment that's interesting: Problem with Parhelia, is that it's still a DX8 part. This fall , as you know Microsoft will release DirectX9 and you're either DX9 or you're not this fall.

One could infer some things from this comment, but Brian didn't specifically mention if their next product is DX9 or not.

At this point, I had some technical questions and Brian went looking for someone to answer them . He found Tony Tamasi, Nvidia's Director of Product Development .

8. Talked a bit about anisotropic filtering. Tony said "ATI only anisos about 10% of the given screen at a time, while Nvidia does the whole scene. Different methods and our image quality with aniso is superior" . He didn't call ATI's method rip-mapping btw but similar to it.

9 N-Patches Nvidia will likely include N-Patches as a subset of some other HOS implementation, much like they did with EMBM and Pixel shaders.

10 FSAA- They want to not lose speed with it enabled. Will likely increase number of samples in the future

11. Tony made a interesting quote " Last year we did a demo of the Final Fantasy Movie. The things that were cut down were the resolution and pixel shader effects. By the end of this year we're likely to be able to render the movie with all of the pixel shader effects"

In conclusion, while my meeting with Nvidia didn't go too far into the future, it did get me more than a little excited about it. It was interesting and different to talk to Brian in person for more than the few minutes we usually chat or email about. Nvidia is one of the biggest driving forces in computer graphics today, if not THE biggest.

Nvidia now has around 45% total graphics marketshare in the PC market (including integrated, desktop, mobile) . That's an amazing growth for this company that only 9 years ago was fighting for it's survival with the release of NV1.
 
It was more like 7 years ago ;) ...but still...it seems like it was almost yesterday when I was involved in discussions surrounding the pros and cons of the Diamond Edge 3D (NV-1), vs. the Creative Labs 3D Blaster (3D Labs Gaming Glint) :eek:
 
ben6: Thanks for these notes... your meeting notes have been excelent.

( I hope I have something at least as interesting as these, to tell about from Assembly 2002. ) ;)
 
ben6 said:
7. Brian made a further comment that's interesting: Problem with Parhelia, is that it's still a DX8 part. This fall , as you know Microsoft will release DirectX9 and you're either DX9 or you're not this fall.

As long as NV30 does not offer EVERY new DX9 feature, it is as less a DX9 part as parhelia. PS2.0 aren't important anyway during the life time of the first generation DX9 products.

8. Talked a bit about anisotropic filtering. Tony said "ATI only anisos about 10% of the given screen at a time, while Nvidia does the whole scene.

This statement intentionally missleading and completly bullshit.
 
ben6 said:
11. Tony made a interesting quote " Last year we did a demo of the Final Fantasy Movie. The things that were cut down were the resolution and pixel shader effects. By the end of this year we're likely to be able to render the movie with all of the pixel shader effects"

What exactly does this "demo" consist of? No freakin' way are they rendering anything close to the movie in realtime. Gazillions of polygons.
 
The NVIDIA FF Movie demo was of two characters (the bald guy and the woman) talking to each other in an interior room. I saw it at an NVIDIA demo a few months ago, and it looked quite good. Not as good as the real movie, but more than good enough for a video game cut scene.

I'm sure they intentionally chose an easy scene to emulate.
 
For the final rendering of the movie, I think a higher subdivision level of the geometry was used than in the Nvidia demo. The shaders were also greatly simplified, but other than that it was the original scene from the movie in real time and did look freaking good. Other scenes would clearly not have been possible to do (more characters, settings, special effects, etc.), so I don't htink there's a way that they could really "render the movie with all of the pixel shader effects" by the end of this year. Even if they are getting closer, I doubt they could handle all the scenes, let alone all the complex shaders...
 
...and the resolution (2K?), and the motion blur, and depth of field, and the AA level, and the texture filtering etc?

Still looks very nice though. :)

Regards / ushac
 
and the resolution (2K?), and the motion blur, and depth of field, and the AA level, and the texture filtering etc?

heh, motion blur? depth of field? I'm sure if 3dfx kept going the way they had been (can't get more from one chip? ok just add another!) they might have been able to do motion blurring and depth of field on the level that was required (might as well throw FSAA into that mix as well, heh..) to achieve what they did in FF..

after all, they already had given Quantum 3D some boards with WAY more power than a Voodoo 5 6K... I think it was on the order of 16 VSA-100s? can't remember anymore..

I really wonder if nVidia is going to try to go full DX9.. if so, that means they'll either need .13um, or perhaps a multichip design ain't so bad after all?? heh. nah.. still they ARE really hitting the manufacturing process' wall pretty hard.. before they didn't really stress them with such high transistor counts, it was all about speed instead. hmm.. time for some of that magnetic technology to come to PC's!
 
My guess concerning AA and NV30 is that it should at least get 4xRGMS under conditions almost for free or alternatively 2xRGMS+32/tap aniso.
 
Nappe1 said:
( I hope I have something at least as interesting as these, to tell about from Assembly 2002. ) ;)

Damn, they still throw the Assembly party in Finland??!??! Hell, I didn't even know that scene is still alive.
 
My guess concerning AA and NV30 is that it should at least get 4xRGMS under conditions almost for free or alternatively 2xRGMS+32/tap aniso

At which point an ATI representative can make a completely factually correct claim that the NV30 only AA's 10% of the screen at any given time. :)
 
Sharkfood said:
My guess concerning AA and NV30 is that it should at least get 4xRGMS under conditions almost for free or alternatively 2xRGMS+32/tap aniso

At which point an ATI representative can make a completely factually correct claim that the NV30 only AA's 10% of the screen at any given time. :)

Naturally ;) And the NV PR guy could easily turn around and claim the same for ATI's aniso algorithm. A perpetuum mobile is not supposed to stop is it?
 
ushac said:
Ailuros,

Really? What basis do you have for that?

Regards / ushac

If you consider either 2xRGMS or Quincunx on current NV hardware close to being "free", then why should it be so hard to up the ante this time one notch higher?

I'm fairly certain that NV30 will support up to 128-tap adaptive anisotropy this time, just don't ask me if it will be exposed in drivers out front. Although it is just an assumption does it sound really that exaggerated to consider 32tap to have minimal performance loss this time around?

DX9 compliance is nice to have but there have to be several other usable selling points for next generation hardware.
 
Back
Top