Nvidia Reveals Kayla Platform for Developers of GPU-Accelerated ARM Applications

Nvidia Announces Second-Generation CUDA on ARM Platform

by Anton Shilov
03/20/2013 | 11:50 PM

At the GTC 2013, Nvidia Corp. introduced its code-named Kayla development platform that weds the company’s Tegra mobile application processor with Kepler graphics and compute processor. The Kayla is aimed at software developers, who plan to develop GPGPU-accelerated applications for ARM-based systems.

 

“Kayla offers a sneak-peek at the capabilities that will be unleashed by Logan, Nvidia’s next-generation Tegra mobile processor. What’s amazing is that Logan will be the size of a dime, whereas Kayla is now the size of a tablet PC,” said Jen-Hsun Huang, chief executive officer of Nvidia.

Nvidia’s next-gen Tegra 5, code-named Logan and due in 2014, will feature ARM Cortex general-purpose processing units as well as Kepler graphics processing unit with support for GPU computing, Direct3D, OpenGL 4.3 and so on. With a graphics processor capable of general-purpose processing on GPU, Nvidia will offer a breakthrough in performance and capabilities.

However, at present the majority of mobile applications do not take advantage of Nvidia Kepler graphics and compute technologies. With Kayla, software vendors will be able to develop new programs that utilize extensive graphics features of Kepler as well as can use its compute capabilities, or just port existing GPU-accelerated x86 apps to ARM-based platforms. According to Nvidia, porting GPGPU-accelerated programs to a new microprocessor architecture is relatively easy. NAMD, a widely-used molecular dynamics simulation package, took two days to move to Kayla, whereas Nvidia’s Optix GPU ray-tracer took one day.

Nvidia Kayla platform is based on Tegra 3 application processor (with four high-performance ARM Cortex-A9 cores as well as one low-power ARM Cortex-A9 core) and features a GeForce graphics processor powered by Kepler architecture.

Kayla development platform can run real-time ray tracing as well as graphics demos Nvidia used to do on massive GPUs, including those that take advantage of PhysX technology. The second-generation CUDA-on-ARM can run Linux operating system.

The Tegra 5 “Logan” will bring extensive GPU compute capabilities to mobile platforms, which will allow developers to port even demanding applications to smartphones and tablets. But a true revolution will occur along with Tegra 6 ", which will feature Nvidia’s own custom general-purpose ARMv8-compatible core code-named Denver as well as code-named Maxwell graphics processing core with GPGPU support. The chips will be made using FinFET process technology, which likely means that it will be made using GlobalFoundries 14nm XM [extreme mobility] manufacturing process.