Parallella....ella....ella....ella - 72 cores and not a lot of watts

Way back in 2012 I backed the Adapteva Parallella Kickstarter. I was going through a phase of gadgets, and was experimenting will all sorts of electronic boards.

Unfortunately, as work and family evolved, I sort of ran out of time and the kickstarter acquired cluster never made it out of its box other than to look at it.

Roll on to 2017, and despite having made it from Aberdeen to Qatar, where it stayed boxed for over 2 years, then it was time to move on again. This time it made it all the way from Qatar to Cyprus.

One day I was sat in the villa, and for whatever reason I had the urge to at least open the box and power it up.

The backing level I supported was the mini-cluster kit. The kit consisted of;
4 x Parallella 16 cards
4 x heat sink
4 x 16GB micro-SD cards
4 x Power supply
4 x small Ethernet Patch leads
1 x 5-port GB Network Switch (D-Link DGS-1005G)
bag of standoffs for stacking the cards.

That equates to 16 Epiphany cores and 2 ARM cores per card or 72 cores in total.

I downloaded the latest version of both the headless and HDMI images, and wrote the headless one to the supplied SD cards. I didn't have a spare keyboard and mouse to attach to the micro-usb port on each card, so went with the headless image to start with.

Assembling the kit, sticking on the heatsinks and hooking the network switch to my main wifi router, inserted the SD cards and powered up the Parallella cards one by one. Looking up the router connected devices to find the assigned address, then creating a static address allocation so it doesn't change in the future.

One thing I need to get in the future is a small fan as this is recommended by Adapteva.

So, here it is stacked and powered up:

I poked around a couple of the demos, but until I get a fan, don't want to risk overheating. Even with minimal workload the CPU reports around 60'C using the provided

So, at least it is working, now just need to find something to do with it!

More info on both the company and the product can be found at the links below.
Adapteva website and Parallella website

There are also a heap of resource to be found on the Parallella Github at the link below;

Have you got one? What have you done with it? Any ideas, or general comments, leave them below.



Popular Posts