next-gen iPad unveiling this week, Apple flashed up a slide claiming that the iPad 2 was 2x as fast as Nvidia's Tegra 3, while the new iPad would be 4x more powerful than Team Green's best tablet. Nvidia's response, delivered by spokesperson Ken Brown, boils down to "Hey guys, it's flattering to be compared to you, but how about a little data on which tests you ran and how you crunched the numbers?"
Nvidia is absolutely right to call Apple out on the meaningless nature of such a comparison, and the company is likely feeling a bit dogpiled given that TI was waving unverified
webpage benchmarks around less than two weeks ago. Both claimed sets of results are equally invalid, but there's a difference -- Apple's percentages may be bogus, but even the iPad 2's GPU is significantly
faster than Tegra 3's -- depending on which tests you pick and what resolution you run them at. The tussle between Apple and Nvidia may seem new, but it has its roots in history.
An Old Rivalry Reborn -
For a brief period of time after the demise of 3dfx, it looked as though the two-way race between Nvidia and ATI might include a third player: PowerVR. The company marketed a line of video cards under the "Kyro" brand. Kyro's claim to fame was its highly efficient rendering approach; the card's used tile-based rendering to dramatically reduce overdraw (meaning the amount of time the GPU spends drawing and shading pixels that are never displayed). Although its GPUs were much less powerful than the competition's on paper, the Kyro and Kyro II were surprisingly effective in real-world benchmarks.
The Hercules Prophet II - PowerVR Driven
Unfortunately for PowerVR, the market was already moving towards Nvidia/ATI-style hardware transform and lighting (something Kyro lacked). Teams Red and Green both adopted features that improved rendering efficiency, available RAM bandwidth continued to leap upwards, and PowerVR's approach to rendering was left behind...until the smartphone boom.
The Imagination Technologies' GPUs built into the iPad 2 and iPad..whatever both utilize tile-based rendering. In some ways, 2012 is a repeat of 2001 -- memory bandwidth is at an absolute premium because adding more bandwidth has a direct impact on power consumption. The GPU inside Tegra 2 and Tegra 3 is a traditional chip, which means its subject to significant overdraw. Adding to the problem is the fact that Tegra 3 islimited to a single 32-bit memory channel.How much this limits T3 performance is unclear, but higher resolutions would be more prone to bottlenecks than lower ones.
Given that the iPad...Whatever uses a resolution of 2048x1536, it'd scarcely be difficult for Apple to choke Nvidia's GPU just by forcing it to run at that resolution. The Transformer Prime's 1280x800 resolution works out to 960,000 pixels -- the next-generation iPad displays 3,145,728 pixels. That's more than 3x as many; those of you who've been around awhile may recall the days when memory bandwidth limitations left GPUs slamming into brick walls at higher resolutions.
Apple's comparisons are bogus, but the T3 bandwidth issue they indirectly point to may not be. Tegra 3 came at an inconvenient time for NV as far as what ARM designs were ready to go; it's largely a stop-gap until the Cortex-A15 and 28nm manufacturing are more mature. At present, Asus is the only company shipping Tegra 3 hardware, though that should change next quarter. LG is expected to launch a quad-core Optimus 4X phone in Q2, while Fujitsu, ZTE, Tianyu, and HTC showed off devices at Mobile World Congress.
-- Article updated to reference other Tegra 3 phones announced at MWC.