Benchmarks Of The NVIDIA Tegra K1 & Its Hotness
Here's our first public benchmarks of the NVIDIA Jetson TK1 ARM development board powered by the Tegra K1 SoC with quad-core+1 Cortex-A15 and NVIDIA Kepler GPU. There's also some thermal metrics for those concerned about the active-cooling on this development board.
Coming out later today or tomorrow will be my first "official" Jetson TK1 benchmarks comparing the performance of this new sub-$200 ARM development board against some other x86 and ARM configurations that have been benchmarked in the past on Phoronix... Intel Atoms, AMD AM1 APUs, Cortex A15 Chromebook, PandaBoards, etc. Anyhow, for those anxious to see some numbers or to run some cross-comparisons yourself, I have uploaded some standalone benchmarks to OpenBenchmarking.org.
Via 1405010-KH-NVIDIATEG94 you can see the first of the Jetson TK1 benchmarks running Ubuntu 14.04 LTS with the Tegra K1's specialized Linux 3.10 kernel. If you want to see how any of your own Linux hardware compares to the performance in these benchmarks, simply install the Phoronix Test Suite and then run phoronix-test-suite benchmark 1405010-KH-NVIDIATEG94. It should be that easy with our fully-automated, open-source, turn-key benchmarking software!
With all of these computational-focused benchmarks I also monitored the Jetson TK1's temperatures for each of the tests... That's done by simply setting the MONITOR=cpu.temp environment variable prior to running the Phoronix Test Suite benchmarks. Anyhow, since yesterday's article there were a number of concerns and complaints by Phoronix readers that the Jetson TK1 board is equipped with a heatsink fan to cool the Tegra K1. NVIDIA likely did this to minimize costs and lower the foot-print of the $192 ARM board. The heatsink fan is attached to the PCB via dual push-pins in opposing connectors and if finding a passive heatsink with the same dimensions, it should work. The fan on this board is audible but not obnoxiously noisy like some cheap, small fans.
With the active cooling, the Tegra K1 peaked at 41 Celsius, bottomed out at 30 Celsius when idling, and averaged to 36 Celsisus during the many benchmarks run within the aforelinked result file. That's not too hot and hopefully for interested parties will be able to find an appropriate passive heatsink that can fit, get the job done, and not raise the TK1's temperatures too much higher. My personal take on the matter is I don't mind too much... It's a development board, not a end-user product. The little bit of noise is worth it for greater cooling capacity so as some assurance I can push the board a bit hard and hopefully not have to worry about any issues, in the past I've overheated and wrecked NAS devices and other embedded ARM systems (not meant for benchmarking) with passive-cooling when applying to much stress during testing.
P.S. For anyone interested in the cpuinfo or other system logs for the Jetson TK1 they are within the aforementioned OpenBenchmarking.org result file for reference.
Latest Articles & Reviews
Latest Linux News
Most Viewed News This Week