NVIDIA Morgan Stanley Webcast: Sony, G80 and 80/65nm

[Arun]

Summary / Partial Transcript Of The Following Webcast:​

NVIDIA Corporation at Morgan Stanley Semiconductor & Systems Conference (Replay)
03/07/06 at 8:45 a.m. PT

I personally was VERY interested by the Sony and G80-related tidbits! The emphasis on 80/65/55nm is really interesting too. Anyway, read on! And sorry about the less than perfect formating; that's the problem of taking notes in notepad...

- Misc ramblings about how cool NVIDIA is and how much it is one of the analyst's favorites companies.
- Focus (has been for a while) on the intersection between Consumer Electronics and Computing
- ASPs have increased over the course of the last 10 years
- The dynamic (and thus ASPs) has the opportunity to stay the same.

- Organized in 4 business units.
-- GPUs are our largest business unit.
- Associated Brand: "GeForce", one of the best known technology brands in the world.
-- MCPs are our fastest growing business unit in terms of dollars.
- Grew 100% last year. 300M Revenue annually. On good track to improve 100% again.
- Associated Brand: nForce. Voted number 1 brand for corelogic, and not Intel!
-- Professionnal Solutions Business
- Sell stuff to content PROVIDERS for other people using our GPUs. But also for films!
- Associated Brand: Quadro. Made its way into basically ALL GPU-related workstations.
-- Handheld Business, with 3G Cellphones as target last year, 3GSM this year too.
- Customers: Sony Ericsson with their Walkman Phone, Motorola with their 3G Razor.
- Target: Smart phones, portable game/content players (PSP/IPOD/GAMEBOY-like).
- Very fast growing business, higher in relative terms but lower in absolute terms than MCPs thus.

- For the 4 businesses: PCs, Game Consoles, Handset, Car Navigation Systems, etc.
- Don't do "basic" (unhearable) CE devices, not sufficiently complex to fit in our expertise.
- Market of $20-30B, expansion with multimedia => growth opportunities.
- About possibility of becoming a $10B company: "Most people thought it'd be pretty tough to be a $500M company".
- Don't think in those concepts. Company can become very large if you work on sufficiently value-added things
- And you need to have a management team and a culture that can sustain that growth. Not much more complicated than that.
- Benefit of our business: people love to work for us just because they can help building amazing things.

- [bunch of gibberish about our position being better this year, SLI being pwnage, etc. etc.]
- Last year, our MCP business grew 100% exclusively on nForce4 Discrete Core Logic.
- So there is a LOT, LOT of room with integrated core logic!
- Both of those designs are going to get OEM design wins we were really weak on last year.
- Weak OEM position, especially so with the US OEMs. Coming into this year, we are at 20-25% (GPU&MCP)
- Expecting to much more than double, or possibly triple, their marketshare in OEMs as we leave our
- That's for desktop, notebook, etc. - was first to market. So all that is ramping right now through spring
- So expecting IGP, New OEM Design Wins, etc.; continuing progressive growth.
- Expecting good Q1 & Q2 thanks to share gains etc. => seasonality will be countered [partially?]

- Desktop and Notebook markets are radically different, because of DIFFERENT CYCLES mostly.
- Desktop PC is more modular => The desktop cycle lags the notebook cycle by 3 to 4 to 5 months.
- So you've got to release earlier, or you don't get the market.
- Also limited by physical space and power/heat properties. So it has to focus on performance-per-watt.
- Performance-Per-Watt has doubled [between GF6 & GF7]
- HD Decoding requires 20-25x more processing power.
- So CPUs are likely to struggle, especially so if you care about power/heat
- As such, that's a further advantage for the notebook market.
- Overall, we think it's gonna be a "pretty good year for us in notebooks".

- My sense is that our US OEM Notebook share is 10%, OEM Global Notebook share is around 20+%.
- Hoping we can "triple that, double to triple that" as we leave the year.

- [Vista: misc. ramblings about GPU importance]
- [Vista: the interface is really nice, according to Jen Hsun Huang]
- Vista has the ability to have applets taking advantage of the GPU [so it'll help over time too for adoption.]
- Any application that wants programmable access to the GPU now has it. Potential big usage shift.
- [misc. other Vista-related rambling and generalities about ASPs thus remaining stable]

- BluRay vs HD-DVD: "It's really hard to say". But PS3 BluRay => Highest Volume Media By The End Of This Year.
- Almost all movie studios have backed Bluray, some back both though. So the content will be there.
- Jen Hsun's Reminder: "The Matrix DVD was the PS2 Killer App". If Sony backs BR properly, it's the "program's conclusion".

- "So I think your RSX full production silicon is ready" - "Yeh."
- "Sony is manufacturing RSX in their own fabs, as well as Toshiba fabs"
- "Our focus now is on diesize reduction and bring up more fabs"
- In terms of our economic relationship with Sony, it has 3 parts:
--- NRA: "We are signed up to build the RSX for Sony, to port RSX to multiple fabs to increase the capacity."
- Had the think in terms of being able to build 30M units a year. Needed to potentially repeat the PS2 success.
- Tons of different fabs needed to reach that capacity, so it's a lot of work.
- For us, we're also signed up to do cost reduction to new processes.
-- Also working on making the RSX more prevelant at Sony. RSX+CELL = Digital Architecture for Sony.
- So they want to use that architecture in all kinds of things, among which HD Video-related.
- Look at PS2: Two huge chips back one, One small chip that's hardly costing anything now.
- So, have to "reimplement the RSX" within the next 10 years into all those kinds of digital devices.
- Considered to be a 50% Margins Business.
--- License Element of the PS3: Extra fixed licensing revenue every quarter for 3 years.
- Been recognizing that for a couple of quarters now.
- That applies to the PS3, but also the use of RSX Technology in "all of these other platforms".
--- Royalty Component, as they ship our technology in each of these platforms.
- That element of our business hasn't started yet.
- But lead time ahead of when you might see boxes in the market. [Uttar: Reminder: They said June/July for Royalties]
--- System is structured so that licensing revenue is likely to stop when royalty revenue peaks.
- "In the front of the life of the PS3, it's all NRA and it's all fixed license fees"
- "So that's why we're seeing the benefits of Sony and PS3 pretty substantially this year"
- "The PSP is really one the last 3D devices that will be built inhouse by anybody on the planet"
- "Nobody has invested beyond Texture Mapping [that is, besides the industry-only companies, thus Intel kinda]"
- "Not one vertically integrated company has invested in internal programmable shading architecture - ever."
- It was a huge discontinuity in investment, so from his pov, it doesn't make sense without volumes like theirs.
- "Our technology is pretty open market, and we're delighted to sell to anybody who wants to buy it".
- So it seems illogical to want to reinvent it.

- "Well we hope that... uhm... yeah, we would like that... uh.... we would be ecstatic to work on
future generation game consoles and, uhm, you know, whether it's be... uh... uhm... gameboy or PSP, the
volumes are just so enormous... and... hum... and it's always fun to have in your pocket something you built."

- "As you know, the foundries did a great job with 90 - and I think the reason for that is that 130 was so hard,
90 is, uhm, 90 is, uhm, I don't think it's easy but it seems like a wall in the park for them. So, we are
ramping up 90 pretty aggressively... Our 7900, the new high-end is 90, the 7600 which is the follow-up to our
really really successful 6600 is a 90... and, uhm, 7300 is a 90. We have integrated core logic going on in the 90
- "And very quickly we'll be doing a shrink in the 80 for cost improvement reasons, but basically it's the
same process. Most of our new desgins [unhearable] in the pipeline are 65 so, uhm, we are investing in
65 now and 55 very shortly after that."

- NV50-related...
- "Increasing flexibility of the programmability, enabling the artists to express themselves in a free way."
- "Our next generation product will just take [effects] to a brand new level"
- "Our next generation product is the combination of 3 years of heavy-duty work. We started architecting it about 4 years ago, and, you know, my best calculations have this investing $250M into it already, and by the time it launches as well as the entire product family, we will have invested about $500M in R&D."
- "It is a spectacular computing machine and, uhm, we can't wait to show it to you later this year".
- "And this year, along with Vista, is going to introduce a very important new API, it's called DX10"
- "And DX10 is just a giant leap forward in unifying the way people program graphics. Instead of, you know,"
"vertices, and shaders and textures, it's unified in a very elegant way. And it's unified in a way that
"makes it possible for us to abstract [GPU] programming to the next level."
- "Our DX10 [part] is nearly finished now, and uhm, it'll be rolled out this year sometime."
- "And, uhm, if I have my way, our next generation GPU will be the first DX10 GPU in the world."

Enjoy!


Uttar

 

[Tahir2]

Thanks Uttar.

NVIDIA are offering rebates in the UK for their top end models. I tihnk this is a first for NVIDIA.

 

[Geo]

Awesome job, Uttar. Green love left in your cp when I can hit you again.

"The PSP is really one the last 3D devices that will be built inhouse by anybody on the planet"

That might be the most significant prediction in the whole thing. I think many of us have been leaning that way, but it's interesting to hear Jen-Hsun say it too.

Note on "NV50" --Jen-Hsun did NOT use that term. That was the interviewer. I've officially switched to "G80" and will brook no "NV50" not from the lips of an actual NV representative!

 

[Bobbler]

He sure says "Uhm" a lot.

Good work Uttar!

Some interesting stuff in there.

 

[Geo]

"Our next generation product is the combination of 3 years of heavy-duty work. We started architecting it about 4 years ago, and, you know, my best calculations have this investing $250M into it already, and by the time it launches as well as the entire product family, we will have invested about $500M in R&D."

We all know that Jen-Hsun *loves* to talk about how much they spend on R&D. But has he made such a bold statement on investment on anything since NV30? Now, for the moment, let's forget about how that one turned out. . .I'm just after do we take this as a statement of a very significant (USC?) departure for G80 architecturally. . .or just Jen-Hsun bragging on his R&D budget?

 

[trinibwoy]

Now that's the kind of gung-ho talk I like to hear!! I'm really interested in how they plan to bring all this innovation in a DX10 world where the focus seems to be on convergence.

- NV50-related...
- "Increasing flexibility of the programmability, enabling the artists to express themselves in a free way."
- "Our next generation product will just take [effects] to a brand new level"
- "Our next generation product is the combination of 3 years of heavy-duty work. We started architecting it about 4 years ago, and, you know, my best calculations have this investing $250M into it already, and by the time it launches as well as the entire product family, we will have invested about $500M in R&D."
- "It is a spectacular computing machine and, uhm, we can't wait to show it to you later this year".
- "And this year, along with Vista, is going to introduce a very important new API, it's called DX10"
- "And DX10 is just a giant leap forward in unifying the way people program graphics. Instead of, you know,"
"vertices, and shaders and textures, it's unified in a very elegant way. And it's unified in a way that
"makes it possible for us to abstract [GPU] programming to the next level."
- "Our DX10 [part] is nearly finished now, and uhm, it'll be rolled out this year sometime."
- "And, uhm, if I have my way, our next generation GPU will be the first DX10 GPU in the world."

 

[borntosoul]

hmmm ill be really suprised to see g80 on 65nm late this year. maybe he was talkiing about the mid end parts but i doubt it.


edit: oh thanks Uttar, great catch.

 

[Pete]

Uttar, you're cranking out great posts faster than I can rep you. Please, don't stop.

 

[Geo]

Uttar, you're cranking out great posts faster than I can rep you. Please, don't stop.

I've been bullying him into it in IRC. Actually, I told him the secret mojo to my rep success is cc summaries. People can tell, and appreciate, when you take a bullet for the team.

 

[dizietsma]

Nice work Uttar.

The talk of G80 later this year, which I assume is the same part as the DX10 part they talk about makes for some interesting speculation on timsescales.

As mentioned above does " most of our designs in the pipeline " mean that G80 will be 65nm or will it be like last time and be initially on 80nm for 8800 and then 65 for the 8900 refresh ? I'd be conservitive and say out on 80nm...

I wonder how quickly the 80nm process will come out for 7600/7900 now as well ? I'm thinking very soon means less 1Q or less, so say May/June. If there is going to be a bit of a price war now between Ati and nvidia the sooner the better of course (and making dual or quad cheaper of course ). Certainly I'd be wary about picking up a new 7900 at the moment as these could be very cheap on 80nm. I think 80nm could knock 25% off the cost for nvidia and so they could decrease price by 15% and still improve margins. Maybe we will see a 7900GT at $249 and the dual card at $500.

 

[Megadrive1988]

very interesting Utter, thank you.

I remember at the NV30 ~ GeForce FX launch event in early 2003, Jen-Hsun or Kirk or one of the Nvidia higher-ups said that they were working on the NV40 and NV50 generations. that was 3 years ago, so they've been working on NV50 since 2002.

 

[Ben-Nice]

- NV50-related...

- "Our next generation product is the combination of 3 years of heavy-duty work. We started architecting it about 4 years ago, and, you know, my best calculations have this investing $250M into it already, and by the time it launches as well as the entire product family, we will have invested about $500M in R&D."
- "It is a spectacular computing machine and, uhm, we can't wait to show it to you later this year".


What machine is this? They can't be talking about the PS3 can they?

 

[Geo]

My impression was G80+Vista.

 

[Maintank]

If there is going to be a bit of a price war now between Ati and nvidia the sooner the better of course (and making dual or quad cheaper of course ).

I think with the performance numbers and the die size of the 7900, Nividia is planning on a price war in the near future. The die sizes of the competing chips and the small performance advantage the X1900XTX has makes me think Nvidia is going to compete on price.

 

[Geo]

On the one hand they are well positioned for it. On the other, their guidance/predictions re signficant increase in already-record margins limits their ability to do so to some degree. Doesn't mean they couldn't if they felt it necessary, but it makes it unlikely they are planning a scorched-earth campaign on pricing at this point. But certainly between "no price war" and "scorched-earth" they have some room there --definitley more room than ATI does anyway.

 

[Sxotty]

If they lose marketshare to ATI I am pretty sure they will pursue a price war rather than slip away. Investors might be upset, but their stock it high anyway right now.

I think it wil be a good test about whether there is informal collusion between ati and Nv. If a price war doesn't erupt. The thing is I kind of feel it might already have started, afterall launching the 7600gt at <$200 is mighty fine.

 

[Tahir2]

NVIDIA 7600GT is positioned in the same bracket the 6600GT was.

For that reason its stated price is normal but it has meant ATI has needed to reduce the price of the X1600 range.

 

[Megadrive1988]

- NV50-related...

- "Our next generation product is the combination of 3 years of heavy-duty work. We started architecting it about 4 years ago, and, you know, my best calculations have this investing $250M into it already, and by the time it launches as well as the entire product family, we will have invested about $500M in R&D."
- "It is a spectacular computing machine and, uhm, we can't wait to show it to you later this year".


What machine is this? They can't be talking about the PS3 can they?

no not PS3. the PS3 does not use NV50 or a derivative of NV50. they're talking about the NV50 PC GPU, or the NV50 GPU within a top-end PC with Windows Vista.

 

[basanti]

someone told me Nvidia got a bad deal from Sony as they will stop recieving royalties once the console sales start to peak? is that true?

 

[Razor1]

The deal with Sony is supposed to be even more then the PS3, so I don't think so, also even for plans for the next PS.

Well this interview clarifies that the g80 will be unified?

PS utter stop posting so many good posts, can't give you rep points that quickly