One of my main criticisms of Nvidia's (NVDA) mobile strategy (and indeed its future direction) is that it is currently not playing to its strengths. While Intel (INTC) has gotten away with its 5-year old Atom core in its latest phone/tablet "Atom" chips since they happen to be very competitive with the best ARM (ARMH) designs, Nvidia has not been so fortunate on the graphics side. Despite being the world leader in everything graphics outside of a mobile phone, it is a mediocre player at best in the graphics department of its mobile SoCs.
Imagination Technologies, a small U.K. based IP licensing firm (similar business model to ARM), has been absolutely nailing it with mobile graphics. After ending up as one of the losers of the desktop graphics card revolution (Nvidia and ATI emerged the victors), it went off to carve out a niche in low power graphics. Sure enough, as smartphones and tablets became important, Imagination's IP has been in quite high demand by high-profile players such as Apple (AAPL) and Intel (both of whom own significant equity stakes in the company).
ARM, too, is producing mobile graphics designs that are quite capable, and will likely be leveraged heavily by licensees in future ARM-based mobile products. While Imagination's technology is still superior, ARM's will improve, too, which could seriously impact Imagination's licensing revenue (and could prove to be a boost for ARM, since graphics licenses are separate from the CPU core and/or instruction set licenses).
So, back to Nvidia. I suspect the reason that Nvidia's graphics solutions (at least on the hardware side) are not quite up to snuff is simply because of the relative nascence of the firm's mobile division as well as time-to-market concerns. The design found in Nvidia's "Tegra 3" chip is based on a 2004-era graphics processor known as the "NV40." I can see why the firm used this design:
- Time To Market - The NV40 is not particularly complex, so it must have been one of the easier designs to re-work to fit into a mobile processor with fairly swift time-to-market.
- Pre-Kepler GPUs Unsuitable - Nvidia make a U-turn with its "Kepler" products. While "Fermi" and its predecessor were generally power-hogs that tried to brute-force their way into performance (AMD (AMD), at the time, had very svelte, efficient designs that made their products quite desirable over the Nvidia products), Kepler represents a fundamental shift in Nvidia's designs.
This focus on more efficient designs (and a bifurcation of ultra-high end designs for HPC/Workstation and the mainstream chips without all the GPU-compute overhead built into the die) was likely due to Nvidia's desire not only to win back mobile PC discrete graphics market share, but to have a design that was suitable for use in a future generation "Tegra."
I firmly believe that Nvidia's experience as a high-end graphics chip developer will be invaluable to winning the Android tablet market. Even with lackluster hardware, Nvidia has leveraged its strong software abilities (the main advantage Nvidia has over AMD in the high end) to get unique special effects and features built into games for Android. Coupled with potent hardware, Nvidia's products could have an excellent value proposition in mobile with class-leading mobile graphics.
CES 2013 may be where Nvidia unveils its next generation Tegra codenamed "Wayne." If the graphics chip is all-new and competitive with Imagination's offerings, then I expect a slew of brand-new design wins for the "Tegra" line. Until then, Nvidia's "Tegra" will probably still sell well in lower cost devices like the Nexus 7, but will fall short of the potential market-dominance that the firm could achieve.
Additional disclosure: I am short ARMH