It seems Tegra 4 hasn’t been received extremely well by manufacturers so far. Toshiba is the first of the major smartphone and tablet manufacturers to use it, although we know that Vizio is also going to use it in that 2560×1600 tablet of theirs. So why aren’t manufacturers scrambling to adopt Tegra 4?
So far Acer and Asus are avoiding Tegra 4, even though they’ve used their previous Tegra 2 and Tegra 3 chips, but I think this time they will go either for Qualcomm (most likely) or even Samsung, who should be more willing to sell their Exynos 5 Octa chips now that Apple is cutting their ties with them, and trying to stop making chips in Samsung’s foundries.
This has likely happened because Tegra 4 failed to impress most of the manufacturers. Performance wise I don’t think it’s going to be too bad, although that remains to be seen. So far the Tegra 4 demos still looked like the most impressive ones to me. But will that be enough?
Virtually every other GPU maker from ARM (Mali) to Vivante, Qualcomm and Imagination are supporting OpenGL ES 3.0, which is the new standard in mobile graphics (think of it like the transition from DirectX9 to DirectX10), and OpenCL (for GPGPU computing). But Tegra 4 doesn’t support any of that, and is instead still using the old Tegra 2 architecture, which I believe was based on the old Geforce 6 series architecture.
There have been some rumors that Tegra 4 would use the latest Kepler architecture but unfortunately that didn’t turn out to be true, and I think, or at least I hope, that the reason for that is because they want to “save” the architecture change for next year when Nvidia should also unveil a new custom CPU core (like Qualcomm’s Krait) based on the ARMv8 64-bit architecture, and a new GPU architecture, called Maxwell.
They are planning to use that new ARM-based CPU and GPU for higher-end machines, like servers and super-computers, but it could also be scaled down for mobile devices, which would really streamline their product line-up – everything based on ARM (well I’m sure they’ll continue to sell the x86 GPU’s, too, for now), and arriving in devices “from super-phones to super-computers”, which is Nvidia’s motto these days.
But while this seems to make a lot of sense cost-wise and business-wise, it’s still a huge blow for this year’s mobile chip, which will probably fall even further behind the competition than Tegra 3 did, by the way things are looking for them right now, and they’ll have to compete on price, just like they did with Tegra 3 last year.
Their Icera 500 soft-modem is, however, looking very good, being able to support any 3G or LTE frequency on the fly, just with software tweaks and updates, which none of the other players in the modem market seem to have right now, but it may take a while before we see that in devices, even in Tegra 4-based devices.