Do you remember that tiny piece of leak about NVIDIA’s Tegra 2 3D? Well, today, the full slide has been published and we can now see that Tegra 3 will have four Cortex A9 general computing cores, and a GPU that NVIDIA announces as three times faster than Tegra 2 – that matches our previous guess. That’s got to be really scary for the competition for a couple of reasons: 1/Tegra 2 is already the fastest SoC (system on a chip) in production. 2/The graphics processor does accelerate Flash as well, so NVIDIA is poised to offer an excellent “full web” experience. That said, there’s one piece of information that most people don’t talk about much:
(low-level, drivers)software written for Tegra 2 will work on Tegra 3 from day 1. That means that all the work done by handset makers and NVIDIA itself can be leveraged to bring up Tegra 3 – this is *huge*. Interestingly, the “sampling*” date is 4Q2010, which means that partners already have the chips, so brace for the demos at Mobile World Congress.
*Sampling means that NVIDIA sends chips in limited quantities to partners and developers so that the whole eco-system can get ready. This is most useful for everyone as developers can see how their code reacts on the real thing, and they can provide feedback to NVIDIA in case something goes very wrong. Most of the time, kinks are worked out in the drivers to avoid are-spin of the chip that would take at least one month or more.