[Dave Baumann]

SFGate has an interview from the San Francisco Chronicle with NVIDIA's CEO Jen-Hsun Huang.

The interview is mainly of a financial nature after the recent poor quarter end results, however there's a few interesting details in there. Heres a snip:

"If you go back to where Nvidia first started, there were something like 50 graphics companies and they made 2-D graphics processors. Many of them were sufficiently large companies. . . . And yet today, all of those companies are gone and there's really only two companies that hold the majority of the (market) share.

The reason for that is because the barriers to entry, or the cost of building these devices, is so high that very few companies can afford to build them. If I tell you NV30 takes 300 world-class engineers to build, and I ask you a secondary question of how many companies in the world have 300 world- class engineers in 3-D graphics, your answer would be probably two."

Read the full interview here.

[Chalnoth]

I thought this little snippet was of particular interest:

Our NV30 will represent the majority of our product shipments by the end of next year.

Firstly, I would imagine that he doesn't mean the NV30 specifically, but all NV3x parts.

Next, it seems that the only way this could be possible would be through the shipment of a low-end NV3x part. Here's to hoping!

[Snap]

Originally Posted by Chalnoth

Firstly, I would imagine that he doesn't mean the NV30 specifically, but all NV3x parts.

Then why say "NV30" specifically? I think you are reaching there. We already know that NV30 is the only .13um tapeout they are doing at the moment and I don't think other elements of the NV3x range will be on any other processes - they'll need the smaller micron processes to offset the cost of the lower parts.

However, its kind funny how he's stating here how NV30 is going to be a significant part of their shipments but in the last CC he was downplaying the importance of the ultra high end parts because of their low volume. So, which is it Huang?

[Chalnoth]

Originally Posted by Snap

Then why say "NV30" specifically? I think you are reaching there.

Don't forget that he said by the end of next year. By the end of next year, how many NV30's do you really believe nVidia will be selling? They've never sold much of any high-end product a full year after release.

If the past is any clue, there will be at least two more high-end NV3x parts released before the end of next year. In order to make NV3x into the majority of nVidia's shipments, however, nVidia will absolutely need to also release at least one low-end part.

[OpenGL guy]

Originally Posted by Chalnoth

I thought this little snippet was of particular interest:

Firstly, I would imagine that he doesn't mean the NV30 specifically, but all NV3x parts.

Next, it seems that the only way this could be possible would be through the shipment of a low-end NV3x part. Here's to hoping!

Someone should have asked why these never happened with the GeForce 1/2/3/4 and why he suddenly thinks it will happen now.

[Chalnoth]

Well, I *think* that the NV1x parts (Specifically GeForce2MX and GF4MX) currently make up most of nVidia's shipments. That statement seems to indicate that nVidia may skip a mainstream NV2x part altogether, meaning most, if not all, of the new chips out by nVidia next year will probably be based on the NV3x core. I don't see any other way for nVidia to ship mostly NV3x parts by the end of the year.

I am kind of curious, however, as to what made nVidia change its mind on the low-end front. They really put out too many NV1x cards, if you ask me. I wonder what made them (apparently...we won't know for sure 'till next year, I suppose) change their mind? Perhaps developers got really upset that nVidia was holding the lowest common denominator back? I'm not sure hardware boards like this have had much effect...but perhaps JC's comments on the GF4 MX did?

Or maybe, just maybe, nVidia never felt that the NV2x core really had that much more to offer, and that they would hold off on increasing the lowest common denominator (available from them, anyway) until the much more flexible NV3x cores became available?