NV30 - "Record Number of Tapeouts"

[Dave Baumann]

From a CC summary:

NV30 will be a important discontinuity in 3D graphics. Made at TSMC at 0.13u. Significantly higher engineering costs. Record number of new chip tapeouts.

That sounds like they are trying to positive spin that, but a 'record number of tapeouts' doesn't sound desireable to me. Ichy has mentioned before that NV30 was late and this could certianly explain that.

 

[Trawler]

Didn't GF3 require somewhere in the league of 5 tapeouts?

 

[nggalai]

LOL !

ah, good one. Don't you just LOVE spin doctors?

ta,
.rb

 

[RussSchultz]

"Record number of new chip tapeouts"

Sounds like there was a record number of new chips that were taped out, not that a new chip took a record number of tapeouts.

A company does not freely discuss how many revisions it takes to get to production, and they're certainly not going to brag about it if it takes too many.

 

[Randell]

unless they are in to damage limitation. Lets face it that one statement can be spinned pro or anti nVidia. Because its .13, record number could indicate efficiency,

The only thing that matters is the feature set, availabiltity & cost of final product

 

[ben6]

Will try and get some clarity today . If I can , I'll post Nvidia's answers.... Anyone like my little summaries of meetings?

 

[Dave Baumann]

A company does not freely discuss how many revisions it takes to get to production, and they're certainly not going to brag about it if it takes too many.

If its eating into significant Engineering budgets causing them to increase the budget, which appears to be the case in this area given the quote, then they have a duty to explain why to their shareholders.

I doubt it would be multiple products since they are talking about the internal designation of one product (i.e. I doubt they are talking about both a high end an low end product - we've heard that NV18 is being done separately). This doesn't preclude 'NV30' consisting of more than one chip though.

 

[RussSchultz]

I doubt it would be multiple products since they are talking about the internal designation of one product (i.e. I doubt they are talking about both a high end an low end product - we've heard that NV18 is being done separately).

I would suggest you listen with your own ears to the conference call before offering your analysis.

At 63:30 (approximately), the gentleman talking on the conference call is stating that engineering and tapeout expenses are dramatically higher this quarter for various reasons. First, the chip must be in .13 (speaking specificially about the next generation. Here he states it will be a big leap, and it will come about this fall) Then, he goes on to add that "a record number of tapeouts for a variety of products to refresh new features and new initiatives in nearly every segment of our business..." will add to the engineering expense for the 3rd quarter.

You can listen to this conference call yourself on

http://www.corporate-ir.net/ireye/ir_site.zhtml?ticker=NVDA&script=1100

 

[Nappe1]

Will try and get some clarity today . If I can , I'll post Nvidia's answers.... Anyone like my little summaries of meetings?

keep going buddy... you are doing excelent job.

 

[Dave Baumann]

"a record number of tapeouts for a variety of products to refresh new features and new initiatives in nearly every segment of our business..."

Ahhhh. that would suggest they may be transitioning their entire product line to .13um ASAP. This is a little odd as they stated the same for the GeForce 4 line - it sounds like this may be a little shortlived lineup. I'd also guess that this means the next Crush may well end up with an integrated NV18 giving it integrated DX8 functionality.

 

[Typedef Enum]

Ben,

Try and bait nVidia into comparing NV30 to Parhelia...I know you can do it

 

[Dave Baumann]

Try and bait nVidia into comparing NV30 to Parhelia...I know you can do it

I don't think there will be much to compare - Parhelia is really competitive to GF4.

 

[RussSchultz]

"a record number of tapeouts for a variety of products to refresh new features and new initiatives in nearly every segment of our business..."

Ahhhh. that would suggest they may be transitioning their entire product line to .13um ASAP. This is a little odd as they stated the same for the GeForce 4 line - it sounds like this may be a little shortlived lineup. I'd also guess that this means the next Crush may well end up with an integrated NV18 giving it integrated DX8 functionality.

I don't think that what he said suggests that at all. When he was talking about .13u, he was talking specifically about the next generation project. He had some commentary which sounded like justification of why they chose .13u (the next gen was only possible in .13u). This justification may be required because .13u is risky, or that mask costs are pretty astronomical. (for example, .13u full mask tapeouts are upwards of $1M US. .18u full mask tapeouts are about $300k US. .25u are about $200k US.)

He then mentioned that the other variety of tapeouts would also add to the engineering expenses. He did not say one way or the other that they were .13u, but there was a definate division in his speech and tone that broke the two ideas into two separate items. I would not have percieved them to be related.

 

[Dave Baumann]

He did not say one way or the other that they were .13u, but there was a definate division in his speech and tone that broke the two ideas into two separate items.

I don’t think there would be as large a product transition if it wasn’t to go to .13um. If a low end DX8 product could have been viable on .15um then they would have gone for it before IMO; if NV18 is a DX8 product then I would guess it could only be cheap enough, ongoing, by going to .13um – others have said this before. They may be opting for large tapeout cost now for a longer sustained product cycle.

 

[RussSchultz]

I guess I'm confused.

He obviously stated the next GPU was going to be 0.13u

He also stated that they were taping out multiple products for refresh and new initiatives.

He didn't say whether those refresh products or other new initiatives were 0.13.

 

[incurable]

... if NV18 is a DX8 product ...

Maybe I'm totally out of line with nVidia naming conventions but doesn't the codename NV18 suggest this is still a DX7 (NV1x) part, probably - like NV17 - using a modified NV15 (GF2GTS) rendering core?

Ciao

incurable

 

[mat]

i heard that NV18 will be a NV17 + AGP 8x + Firewire Port (to competete against ATI AllInWonder8500 DV)

(dont know if its true)