As a partner of [company name deleted - as it was recently formed], I attended the Analysts’ Day at nVidia on June 16, 2009.
1. nVidia is convinced that the future belongs to a new architecture in which both a CPU and a GPU share the tasks that they are more appropriately designed to perform. Jen-Hsun Huang likened the use of a CPU to perform all functions to the use of a Bugatti Veyron’s 1001 HP engine for city commutes [generates a lot of heat].
The CPU/GPU model is one that he reiterated some forty times. The CPU/GPU co-processing model has been embraced by both Microsoft [for Win-7 - DirectX] and Apple [for OSX-Snow Leopard - OpenCL]. They (NVDA) had a bunch of impressive demos.
Comment: Demos are demos. Even if you are “driving”, they are designed to feel good.
2. Extolled the virtues of Atom+ION. The benchmark numbers were impressive. To make a long story short, Atom (by itself) cannot run HD 1080p video at forty plus fps. Atom+ION can! Video transfer and video conversion [from HD to iPod] were 5x faster with Atom+ION [when compared to an Atom - which was unassisted]. In other words, this re-iterates the CPU+GPU model for all future computing (per nVidia).
3. Tegra is designed to deliver the highest performance at the lowest power. The nVidia philosophy is to start at zero power and go upwards from there. Tegra shuts off anything that is not working, and “divides the work” among multiple CPU’s. These CPUs are strategically chosen after a review of the types of jobs that can/will be run on the computer. Tegra has eight CPU’s [all of them have an ARM core]. The eight are:
a. 2xARM [CPU]
c. 2D Engine
d. HD Video Encode
e. HD Video Decode
f. Audio – Poeral Player
g. Image Sensor
4. NVDA already has 42 Tegra Design Wins. Of the 42, 18 are SmartPhones+MID [sans modem] subsidized by wireless carriers. The 42 design wins are at 27 unique manufacturers.
5. 27 wireless carriers have been engaged worldwide.
6. Opportunity is 25M smart-phones in 2010. OEM/ODM paired with carriers. First launches end of this year.
Comment: That is a very optimistic number.
7. Tesla and Tegra will drive future growth.
8. Migration to 40nm will further improve margins.
9. On the near-term, inventories have been whittled down from 150 days to 64 days.
Comment: Good for new product launch, and profit margins.
10. NVDA claims that Tesla is a desktop supercomputer [so was the Power Mac when it debuted], and this opens up a huge market that wasn’t previously served by nVidia.
Comments: This could possibly blind-side Intel. nVidia has about a two year edge when it comes to “middle-ware”
11. To improve gross margins, NVDA wants to:
a. Drive sales UP.
b. Increase ASP by transitioning product mix to higher ASP products.
c. Decrease COGS by eliminating waste, and by transitioning to 40nm [complete transition will take about eighteen months].
d. Moderate OPEX.
1. Nothing we did not expect.
2. NVDA seems to have a good read of the future of computing, but almost all of the future is partly based on the assumption that Intel will not counter nVidia’s version of parallel computing with one of their own. In fact, Intel has significant experience deploying supercomputing centers.
3. Most of this will play out in the end of 2009 or early 2010.
© Bapcha’s Stocks, 2009.