SAN JOSE: CHIP DESIGNER Nvidia said at its GPU Technology Conference (GTC) today that its Tegra 5 chip will be the first to support its CUDA programming language.
Nvidia's Tegra 5 chip codenamed 'Logan' will feature the firm's Kepler graphics core and support OpenGL 4.3. However the most important detail Nvidia CEO Jen Hsun-Huang revealed was that Logan will be the first Tegra chip to run CUDA.
While ARM and Imagination Technologies have been pushing OpenCL support on their respective GPUs, Nvidia has been holding back on bringing its CUDA support on its Tegra GPUs. However the firm will change that with Logan, which will also have a GPU based on the firm's Kepler architecture.
Huang also showed off CUDA demos running on a Kayla test board sporting a previous generation Tegra 3 chip. While Huang's demonstration showing off ray tracing was impressive, Nvidia's Kepler architecture is known to offer significant performance improvements over previous generations.
Not surprisingly Huang was happy to talk up Logan's CUDA support, but he also mentioned that the chip will support OpenGL 4.3 and one of the demos on stage using the Kayla test board was an OpenGL rendering of the ocean. Huang did not mention OpenCL, a word that hasn't been mentioned many times at GTC.
Huang said that Tegra 5 chips will be released at the tail end of 2013, with devices appearing early in 2014. It will also mean that next year all of Nvidia's product line will be based on the same GPU architecture, which should help lower costs. µ
Because that worked out well last time
The Pixel C is going it alone
VirnetX is also calling for another $190m in damages
Acquisition will see chipmaker create deep learning-based apps for autonomous vehicles