These are simply notes from the analyst meeting held June 16th at the company HQ. I listened via webcast. [Comments in square brackets are thoughts.] For our published research and models please login to the website and visit the library.
Notes:
The standard computer today is far from sexy.
They industry is at an inflection point that we will remember going through for years to come.
Believes that a new architecture is required and the GPU has an important role to play. Future systems will include both a sequential processor and a parallel processor co-processing along with it.
In order to meet the processing needs of rich internet applications, a processing architecture like the above is what is needed.
Growth comes in three slices: mobile/embedded (TEGRA), workstations (TESLA) and consumer PC (GEFORCE / ION) and they have TAMs of $7B, $5B and $26B (consumer only) respectively. Today they are said to be $0B, $1B and $4B.
It’s not unnatural to believe that GPU spend should be equal to or greater than CPU spend. [Interesting point, not the consensus view.] Given how important visualization is the power focused on that part of the problem should be equal or greater to sequential computation resources.
Suggests that today people are looking at actual performance and not powerpoint slides of what chips will do and that Tegra is doing very very well.
The only way there can be mainstream market shifts in such an established industry area you need giant volume and major changes in overall performance. [Says that CPU "hit the wall" but I don't think that's true at all.]
Moving into CUDA discussion which enabled them to put GPGPU functionality “on the back of” the standard GPU “workhorse.” 120M units out there that can run CUDA today by downloading and installing software. But can the speed improvements be delivered?
As a new market and way of doing things it has taken a ton of work to “make this architecture stick” and they think they have achieved it. [So what about alternative technology providers for the CUDA wave if there is in fact one coming?]
Tesla:CUDA as Quadra:GeForce ?
Power efficiency is an opportunity area where GPGPU computing can be attractive. Many tasks like image procesing, video transcoding are handled well by the GPU. General like this functionality is more integrated into the OS and accessible with Snowleopard and Windows 7.
Silly to have a quadcore x86 CPU processing video if you are watching a movie or processing audio if you are listening to music. The GPU is much better and more efficient at that.
Power management and dissappating power is very expensive versus eliminating it.
Lots of discussion of programmable shading and CG and that it would lead to more general things like CUDA. HPC and digital film creation are some of the few that don’t want to use Windows or MacOS. Describes application functionality like automatic image recognition and tagging as some of the “magic” that the GPU architecture will enable.
For games it’s not about how things look but what you can do with the game in terms of experience and more dimensions. The PhysX engine is key. Runs on a CPU but runs on a GPU faster. 3D vision is another way to bring more richness to the experience.
The Nvidia approach does result in a higher cost product versus ATI. Maintains that their gross margin problem has more to do with their $1B in excess inventory rather than structural cost disadvantages.
They are trying to do something dramatic and are expanding the industry. It means that there can be short-term pain and investments to be made. CEO “really believes” in what they are doing. [This part must be scary to investors.] Leadership of the company is not taking pay. [Really?]
The market opportunity will roll out as applications become CUDA-enabled. One should see the emerging companies summit for where Nvidia may invest. Bought a stake in the company that become Google Earth. Lots of joint development going on with large companies. Mindful of negative synergies from acquisitions. Tegra was built from the ground up around an industry standard processor. [I wonder if the Intel/Wind acquisition has any impact on Nvidia?]
Tegra will be focused on the mobile/embedded opportunities that demand HD video where there are no other acceptable solutions. [Have to see about that.]
Claims that they are focused on things which are too hard for others to do, will result in major design wins and generate major increases in TAM. These are the continual tests that are run against the company business plans.
Mike Rayfield, GM of Tegra – “First HD Mobile Processor”
80% of the top 100 websites have Flash built in. [?] Hulu visitors average 30 minutes a visit. Average YouTube watcher spends 28 minutes a day online. Going HD. Even FB has lots of images, video and animation. Video minutes on FB up 13x YoY.
Looking at Tegra complete its like a stick of gum. 8 processors. 2 ARM, GPU, 2D engin, HDV encoder, HDV decoder, audio, image sensor. Means that ARM processor is running at 5% when doing 720p video. [Cisco backs up the notion that video content is going to be 2/3 of what we are processing in the future.]
42 design wins so far. About 18 are smart phones. “Household names, household carriers.” Working with 27 different manufacturers right now. 25m unit opportunity in 2010. First launches are end of 2009. Working with 27 different wireless carriers. Many love this little Foxconn notebook. Transfer price to carriers will be about $200 which is perfect for them. [Are networks the gating factor here?]
Carriers are excited about this netbook/MID opportunity to drive their business. 100 of them were at Computex where they don’t typically go. Carriers want to brand the devices. (Wistron, Inventec, Compal and others.)
Average TV watcher still spends 4 hours/day versus the 30 minutes for YouTube or Hulu. [This could take a bite out of Windows use at the low end.] Users typically have lots of tabs open and many will have video running.
Able to show amazing and real video at under 1w. (!) [Early days but we will have to see what the Qualcomm Snapdragon looks like.] Working closely with Adobe to be able to accelerate Flash. (Not just Lite but also 10.)
The Mobile TAM is under $4B now but will reach $12B by 2013 (smart phones, portable players, MID, automotive have design wins today and others will join the mix.)
Tegra 650 is the product today. T2 is 4x and due 1H2010 adn T3 is 10x and due out 1H2011. Same power envelope (1/2 watt.) [Can they do it?] Maintains that Intel is 2 years behind and before they catch up the 4x and 10x products will be out.
“Dehydration of the PC” strategy is a great negative tagline for Intel.
Suggests they can be mid-upper 40’s in GM and “not be a drag” on corporate margins.
Competing versus QCOM which has modem around which they can wrap additional transistors to gain share of the smartphone. Suggests that pace of change in multimedia and modems are different such that it makes no sense to lock them together.
A slower market does help incumbents which can slow transition to new technologies like Nvidia. Basically says that how much revenue falls into the current year is hard to say but by Q1 of next year the momentum will be very clear. Said that the ASP range is $18-30 depending on functionality.
Devices now are dominated by the display which means there is “plenty of room” for chips. Supports Windows CE, Windows Mobile and Android. Have designs going in all those. [What about Palm WebOS, iPhone and Symbian?]
Analysts seem to feel that they company is involved with mostly “non-Tier 1″ partners so would like to see that change.
Talking about $50-100M in revenue for the “back half of the year” and could be into the following year depending on the carriers. Division has 500 people so you can do the math to figure out what the breakeven is. [Noting a little testiness in talking about numbers.]
Company is starting a magazine? What does the mix look like in three years? 1/2 Tegra and the other 1/4 (?) servers and workstations. GeForce will be the rest. In 20 years computers will be everywhere, they will be very small and they will all be connected to the web.
Analysts are very nervous over costs at this company. That’s clear. IPTV will be a huge driver. It’s all going 1080p HD. Not too many computers today can do that, let alone a $300 MID.
On the x86 they want to avoid it like the plague because of power consumption. It’s too much more than ARM. When you start thinking in terms of billions of devices, 10 watts is too much.
Where are all the software developers going? It’s not desktop. It’s mobile. iPhone, RIM, FLASH, etc.
Went through the gorey details of the 10 point collapse of margins due to inventory and price corrections. The professional industry remains weak. Auto industry is suffering and they tend to buy lots of workstations.
Suggesting that they will get into the digital video streaming market? Needs a 30x price/performance improvement. [I wonder if Akamai is looking at this?]
Refers to an “enormous amount of middleware software” to support these processors. Investing heavily in tools to help build software for the Nvidia platform and education programs at universities.
Enabling a new architecture is a big deal, lots of cost. Outcome is uncertain. Look at Sony with Cell. That cost a fortune and basically is a failure. Company has 1000 hardware engineers out of 5600 but lots of software. The development technology organization is over 200 people strong and around the world.
“ION success would be 2-3x what it is today if not for the Intel FUD tactics. The lawsuit is about a future product and only confuses the market into thinking it is related to current products.”
The goal is to put a GPU into the computer. The chipset business is about the glue that makes that possible/easy to do. It’s all just I/O.
Doesn’t a combined CPU/GPU vision favor the competition? CEO questions old views of market segmentation around user experience. [Very interesting point, worth more thought.]
[Are we really talking about old data-intensive versus compute-intensive here?]
Nahalem is a great CPU. Atom is a great CPU. Very unique versus what Nvidia does. The complement is what matters. Making it too similar defeats the purpose. Larrabee x86 cores don’t make sense in this context.
Claims that there is a 10-15% premium in the marketplace for Nvidia. Price wars hurt players in the channel if they are violent.
Integrated graphics have been integrated for a long time. It will set it’s own limit on what is possible. Would never make sense to do gaming or anything intensive on that integrated graphics platform.
Discussion of USB and SATA in chipsets and why anyone would care about where it sits. Seems to confuse most analysts. The goal is to have a CPU and GPU in every computer.
Is Nvidia behind on DX11? Response is nobody has shipped it yet. Aspires to be Apple in terms of talking about products that are released and available now, not in the future.
Don’t consider corporate an opportunity. (!) [They are not getting it on RealVR in the enterprise.]
The real fear finally voiced by analysts is that Intel will kill them. Because you can see the value that the Nvidia chip brings it makes it impossible for Intel to defeat them by integrating a commodity function like I/O.
Suggests that Intel is basically saying that they will give $30 to you if you use Atom without Nvidia. But for many markets that still won’t do. Those are the Nvidia customers. The cost of an MP3 player is basically zero but people still flock to Apple because for $49 or more it’s great.
When you turn CUDA on in a video editing environment you will be amazed at what is possible.
Drew Henry – GM Desktop GPUs
Discusses the fact that the two main OS vendors are moving to embedded support for the GPU. You can write to OpenCL, DirectX, CUDA or FORTRAN.
ION is a new brand. It’s aimed at the small, fully-functional part of the market. Specifically designed for this market and $299 price points. GeForce is aimed at the higher end of the market. Gaming, creative, workstations, etc.
Running 1080p Blue Ray full speed on an Atom PC as a demo. To copy to a sony video walkman it must be copied but also converted. Both the Atom/Ion and Atom only systems are running Windows 7. The video transcoding is built into the OS which is very nice. The GPU system can do a 2 minute clip in just over a minute the non-GPU system takes 15 minutes. The real difference is about 5x if you let them run. So a one hour video converts in 38 mintues on an ION and 3.3 hours on an Atom.
And there are over 100 million GeForce GPU’s out there now that can all use this ability. Moves into discussion about PhysX. Integrated into major game engines, is cross platform and licensed by most of the major developers. [No discussion on Havok.] See Darkest of Days for examples of the effects. See also Dark Void, U-Wars, Terminator Salvation as well. Giving a demo with the GTX 285 (high end stuff.)
758 titles said to be supporting SLI today which will show up when people move to Windows 7. 3D Vision being demoed. Several hundred games can work out of the box in 3D. Still expensive so it’s not mass market. Working on a low end 3D glasses approach to help more consumers get the experience.
The GPU is becoming the de facto standard for processing video. Parners are taking Nvidia technology into this market for editing, enhancement, conversion, upscaling, HD and Blu-ray.
Demonstration of VReveal which was developed by the CIA to clean up video. Now this can be used for that huge inventory of VHS, cellphone video and all that content to be uploaded and used.
Showcasing some devices like the Lenovo S12 notebook. Turns an Atom into a fully-functional PC. Sims3 sold 1.4m units in 7 days. Over 100m units sold to date. ION makes it playable on a netbook. [Lots of the same points being made here for 20-30 minutes or so.]
[Should do a blog post on Intel "crimes against humanity."]
Intel and AMD are supporting Havok. What does that mean?
Claims that PhysX and Havok are totally different animals. (?)
Basically the fact that Windows 7 and Mac OS XI makes the GPU automatically a full player in the computing environment is the key for the market shift.
Good question on what level to write to in terms of being able to have the OS optimize performance.
Andy Keane – GM Tesla Computing
Cray, Connection Machines, BBN, Maspar, etc.
Innovation in the HPC market is gone. Mostly x86 now, some GlueGene.
Where is that next factor of 10 or 100?
Want to do supercomputing? Go down to Fry’s, get the $49 card and download the software.
Key industries are Seismic, supercomputing, universities, defense and finance (300m, 200m, 150m, 250m and 230m TAM for each one.)
“World’s Highest Performance 1U server.” New category.
Two areas of Tesla. Personal Supercomputer and Tesla co-processing cluster. 300 installations of Tesla clusters. Can democratize the top end of the supercomputing pyramid. 100’s will spend 10M, 100,000’s will spend $50k-$1M but millions can spend less than $5K for a PSC.
David White, CFO
Reviewed the charts regarding the impact of the inventory correction and the impacts on margin. Company put out a target model getting back to 13-15% operating margins over time.
Ending Q & A
Working down inventory and of course there is a mix shift but why are wafer starts going up? Answer is that Q1 was almost none so Q2 has to be up sharply.
How constrained is 40nm? Believes they have the terms with TSMC to get what they need.
GPU attach rates are going to go up sharply with the new OS releases later this year. ION is a good entry level system but the discrete chips add much more processing power.
The business works better when channel inventories are lean. Windows 7 should help drive demand in the second half. Dell Quicksilver has two discrete GPU untits with SLI and a 9400 on the motherboard. Win7 and Nahalem should be a good combination and a viable new enterprise platform.
{ 1 trackback }
Comments on this entry are closed.