One of the biggest concerns about NVIDIA (NVDA +0.16%) has been the viability of its Tegra business. Many criticisms have been offered – some very much justified and some maybe a bit unfair. Nevertheless, it's tough to ignore that NVIDIA's Tegra revenue during the current fiscal year was down dramatically from the last year. However, the coming year should be very exciting for Tegra.
Project Logan looks like a real winner
NVIDIA announced a chip known as Tegra K1, formerly codenamed Project Logan. Interestingly, this was announced in two flavors. The first is a 192-core NVIDIA Kepler SMX paired with 4+1 ARM (ARMH +0.00%) Cortex A15r3 processor cores. This should be the world's fastest mobile system-on-chip as far as graphics is concerned and it should be right up there with the leaders on CPU performance.
More interestingly, NVIDIA talked about the second variant of Project Logan which sports fundamentally the same system-on-chip but with the 4+1 Cortex A15r3 cores replaced with two of NVIDIA's custom "Project Denver" cores. These are ARMv8 compatible (so ARM still gets royalties), but instead of four comparatively weak cores, NVIDIA is plopping down two very fast processor cores.
This is the right design decision (as Apple and Intel have proven with A7 and "Haswell," respectively) as the vast majority of consumer software can only really take advantage of one or two cores. It was a gutsy move for NVIDIA to forego marketing hype for sound technical judgment, but this will ultimately prove to be the "right" way to go.
NVIDIA's position in the mobile world is clear – gaming focused
NVIDIA seems to be doing something very interesting with Tegra. Instead of going after the mass market of tablets, it seems to be betting on a niche in which it is uniquely positioned to play – anything and everything that needs monstrous graphics power. Now, that's not to say that Qualcomm and Intel won't compete aggressively here, but companies like Qualcomm and Intel (INTC +0.78%) need much higher volumes to move the needle than NVIDIA does.
This means that the focus for both Qualcomm and Intel across their product lines will be much more on higher volume, lower cost parts. Intel, in particular, has a manufacturing lead that allows it to be the "low cost, high volume" producer at good margins. While NVIDIA is obviously going to try to gain as much business as possible, it is likely that Tegra K1 will be a fairly expensive chip (given its performance level), meaning that it'll be quite high margin and probably find its way into more "premium" Android tablets while others focus on the lower end/higher volume devices – where visual computing isn't as important.
Foolish bottom line
NVIDIA looks to have done a spectacular job with Tegra K1, going all-out to build a graphics-performance focused part on TSMC's 28 nanometer process. It will be exciting to see how Tegra K1 fares next year – it'll certainly do much better than Tegra 4 and, frankly, it's probably the most compelling mobile system-on-chip currently announced today. It's a great time indeed to be an NVIDIA shareholder.
