Nvidia unveiled a few interesting products and technologies at the GPU Technology Conference on Tuesday, but its decision to spill the beans on upcoming Tegra chips was probably the most significant announcement. Here is why.
Tegra 4 was delayed after Nvidia was forced to respin the chip. The teething problems are nothing unusual when transitioning to a new architecture on a new manufacturing process. Nvidia doesn’t use Intel’s tick-tock approach, so it was clear that the transition would be risky - and it was.
Something went wrong and Nvidia was forced to delay the Tegra 4 by at least a quarter. It then went into damage control mode and showed off the Tegra 4i, with a reference phone design in tow. The Tegra 4i is based on the old A9 architecture, but like the Tegra 4 it is a 28nm chip and it has beefier graphics. It also has an integrated LTE modem, which means it should get more phone design wins.
The only problem is that the T4i won’t ship until the last few weeks of 2013, so Nvidia’s decision to release the full spec was rather unusual. Apparently that wasn’t enough, so now Nvidia is talking about next generation Logan and Parker chips, which will probably end up as the Tegra 5 and Tegra 6 respectively.
They sound impressive to say the least. Logan, or Tegra 5, will feature vastly improved graphics, based on Nvidia’s Kepler GPU, which is used in current-gen Nvidia discrete graphics. It will also support CUDA, Nvidia’s programming model for GPUs that should allow the graphics cores to be used for something a bit more productive than Angry Birds. Parker, the next-next-generation Tegra, looks even better. It will be able to use 64-bit ARMv8 cores and it will be based on Nvidia’s Project Denver processor initiative. It will also feature Maxwell GPU cores, doubling its graphics performance per watt over the previous generation.
However, there is a very good reason why the specs sound so impressive - the chips are far from ready. Tegra 5 is expected to launch in early 2014, while Tegra 6 is coming at some point in 2015, and that’s if everything goes as planned, with no respins and delays whatsoever. The competition, however, won’t sit idly by and we should see similar chips from Qualcomm, Samsung and Apple.
Nvidia now appears to be leveraging its GPU design prowess, which is a good thing. Although Nvidia is usually associated with snappy graphics chips, the first three generations of its Tegra SoCs did not offer world beating graphics. In fact they were routinely outperformed by other high end ARM chips, namely Apple’s A-series parts with huge Imagination Technologies graphics cores. Things might be about to change in fourth generation chips and if all goes well, Nvidia might even grab the GPU lead with subsequent generations.
Tegra 4 will probably be Nvidia’s last chip without integrated LTE, and subsequent chips, starting with the Tegra 4i should feature LTE on board. This is another bit of good news for Nvidia, as it couldn’t remain competitive without LTE. Samsung is facing similar issues with its new Exynos 5 Octa chip, which it likes to call an octa-core although it really isn’t one. It lacks LTE and therefore Samsung has to tap Qualcomm Snapdragon 600 chips for the Galaxy S IV. Although it likes to talk about its “revolutionary” big.LITTLE Octa chip, in reality most carriers and consumers will probably go for the LTE enabled Galaxy S IV, which is based on Qualcomm’s chip. Now that’s not something Samsung likes to talk about, for obvious reasons.