Finally, after hunting for a decent power supply and attempting to navigate through Adapteva’s somewhat, detail lacking, documentation, my Parallella’s are finally churning away.
In picture, Rho, the first of two nodes, running the Blobubska real-time ray-tracing demo.
First thoughts, its a little sluggish, but realize, any similar OpenGL demo would be operating on a graphics card like a nVidia Tesla with 512 Cores while this little puppy is doing a respectable job on only 16 Epiphany cores. I am eager to see how the 64-Core Parallella handles the same demo when Adapteva finally releases it.
My goal of this test, was mostly to measure the operating temperature performance with a medium load. Adapteva suggests, anything below 70C is good.
Since this is essentially a Rev A board, its lacking in some features that the Rev P boards have, namely a heat sink slab for both the Zynq FPGA and the Epiphany-16, I have modified the setup slightly, I added my own heat sink to the Epiphany chip and I placed a 5V fan to blow across both heat sinks.
The good news, the improvements worked better then expected. Its been running all day on the Blobubska demo and hasn’t gotten above 59.1C (sitting in an already hot office). As well, system load has maxed at 0.31 (as compared to running the demo in HostOnly mode [i.e. runs without the Epiphany chip] at a load of 1.5 and terribly sluggish video by comparison). Heat and load monitor below…
What was funny, since I am the paranoid security type, I always keep a close eye on /var/log/auth.log to see if anyone is banging on the box. Between the Parallella and my Raspberry Pi Model B+, I’ve had to block 7 networks from China, one from Denmark and one from Spain.
Damn crackers picking on my poor little computers.