Jetson/Graphics Performance

From eLinux.org
< Jetson
Revision as of 13:49, 21 June 2014 by Deppman (talk | contribs)
Jump to: navigation, search

The Jetson TK1 SOC

The Tegra K1 SOC in the Jetson TK1 is targeted for embedded GPGPU applications as well as general purpose use in power-constrained devices such as super-phones, tablets, laptops, set-top boxes, and low-power desktop computers.

GPGPU Capabilities

The Tegra K1 SOC GPU provides excellent GPGPU performance. Nvidia claims Tegra TK1 can attain 326 GFLOPS, whereas its closest contemporary competitor, the SnapDragon 805, may achieve estimated 200 GFLOPS. Imagination Technologies announced a PowerVR GX6650 design which they claim can challenge the Tegra K1 performance. However, the design favors FP16 operations which may limit its usefulness for GPGPU tasks. As of June 20 2014, public bench marks are not available, and the GX6650 may not ship to consumer until 2015. By that time, the NVidia Erista (the Maxwell successor to the Tegra K1) should be available.

NVidia leapt over the competition by using the same Kepler GPU architecture that it has used for years to power the worlds fastest desktop GPUs and super computers. This allows them to leverage existing, high-quality code to provide superior breadth of capabilities. APIs include OpenGL ES 3.0, CUDA 6, OpenCL 1.2, OpenGL 4.4, and a bevy of others.

In my testing, GPGPU tasks have required the highest power draw by a wide margin. Typical non-GPGPU applications, including demanding OpenGL games, rarely required more than 5W for the SOC + RAM. GPGPU applications that harnessed the power of all CUDA cores, however, could push the SOC + RAM numbers to 9W.

Frames Per Seconds (FPS) Comparisons

More details soon.

Power Consumption

According to my testing, graphics performance on the Jeston TK1 is about on par with Intel HD 4600 graphics, but with superior OpenGL and GPGPU capabilities. Power draw for graphic-intensive tasks has been surprisingly low. It appears that typical graphical applications probably default to lower-power FP16 operations. GPGPU tasks, by comparison, do increase power draw significantly.

More details soon.