Nvidia Posts Strong Earnings Thanks To Kepler and Tegra 3
The other major factor was the announcement of Google's Nexus 7. Nvidia's Tegra 3 is at the heart of that device, and it gave the company's tablet chipset a major boost. Unlike Tegra 2, which staked out a strong position in both tablets and phones, Tegra 3 hasn't been widely adopted outside of tablets, and the Android tablet market hasn't exactly dethroned the iPad. Google's Nexus 7 has reportedly been selling extremely well, boosting Nvidia's consumer product revenue by over fifty million dollars.
Nvidia's Professional Solutions Business (PSG), which includes Quadro and Tesla, was the one down side, revenue in this market fell slightly, to just under $200 million. Nvidia states that this was due to the combination of weak European sales and possibly the delay of Intel's Romley chipset. It expects sales to pick up again in Q3.
So How's Tesla Doing?
One point that Nvidia doesn't specifically address is how strong Tesla sales have been. The company has three business segments -- consumer GPUs, Professional Solutions (Quadro and Tesla), and Consumer Products Business (Tegra). Wall Street hasn't paid much attention to Tesla's performance over the past four years, but with Intel's Xeon Phi shipping this year, we suspect that's going to change. Nvidia has been fighting to create a GPGPU business around its own products -- is it working?
By all appearances, not very well. We base that impression on a consideration of the company's revenue, broken out by segment. Figures are given for the first half of the year going back through 2010. The roughly 7% decline in GPU revenue since 2010 has been more than offset by Tegra's strong growth; revenue from consumer products has increased four-fold in just two years. That's a significant achievement for Nvidia and it speaks well of the company's efforts in the mobile sphere.
Professional revenue, in contrast, has held steady for the past two years. That doesn't mean Nvidia hasn't sold any Tesla parts, but it suggests that Tesla's impact on revenue has been marginal at best. If the company is selling Tesla hardware, it may well be selling it to the same companies that would normally have bought Quadros. It's always been apparent that Tesla would be a long-term endeavor, there's no quick way to drive adoption of an emerging standard, particularly when it requires the adoption of new programming techniques.
The problem for Nvidia is that its honeymoon is just about over. The upcoming K20, due in October, is reportedly targeting between 1.5-1.8TFLOPS of double-precision performance, compared to Knights Corners' estimated 1TFLOPS. Even if K20 has a marked real-world advantage, however, Intel's software compilers, performance analysis tools, and code optimization capabilities are second to none. Xeon Phi's launch will focus a great many eyeballs on Tesla's performance and capabilities -- and NV has to be ready.