Nvidia has given an update about the roadmap for its Tegra processor at the GPU Technology Conference in San Jose, California. Tegra 4 will still be followed by Logan (Tegra 5) as planned with a Kepler GPU and support for CUDA and OpenGL 4.3, but “Stark” has been replaced by “Parker” (Tegra 6) which will be the first 64-Bit Tegra processor based on Denver CPU, Maxwell GPU and make use of Finfet transistors.
Logan will be available in 2014, and Parker should be available in 2015 with 100 times more performance than Tegra 2. With this kind of performance, the separation line between desktop and mobile processors will be gone.
Nvidia also unveiled Kayla (“Logan’s girlfriend”), a development platform for CUDA and OpenGL based on Tegra 3 quad-core ARM processor and a Kepler GPU connected via a PCI express slot.
Jen-Hsun Huang (above) showcased Kayla performance by running real-time ray tracing at GTC using CUDA 5, OpenGL, and PhysX processing at the same time in Linux (Ubuntu). That’s something that was only possible with GPU for PC such as GeForce and Tesla up to now. Kayla development board should be available in Q2 2013.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
4 Replies to “Nvidia Updates its Tegra Roadmap with Parker 64-Bit ARM SoC, Unveils Kayla CUDA Development Platform”
I can’t help but wonder about this. Will it be the same Kepler architecture as on the desktop side ? Because that would mean the same Nouveau drivers could work.
I understand Kayla demo is running the same test programs as for the PC platform, and those demos were run in Ubuntu, so at least they’ve already got the drivers for Linux. I can only assume they made Kayla to have a platform development before Logan silicon is ready early next year, so that software working on Kayla will be ready at the same time Logan launches.
“That’s something that was only possible with GPU for PC such as GeForce and Tesla up to now. ”
Except that other vendors have acceleration platforms too. nvidia seem to live in a microsoftian warm blanket and seem to think they’re the only company doing anything.
Not a single mention of OpenCL anywhere – so we know where nvidia stand on that at least. So this will be a complete pain on mobile platforms since google will presumably use their renderscript junk to abstract away from either when they could’ve taken some real leadership.