Talk:How to enable your AppleTV as an Xgrid node

From AwkwardTV
Revision as of 23:56, 10 April 2008 by Bonessk (talk | contribs) (Notes)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search


  • I've found it to work fastest when the ATV is plugged into ethernet, instead of using AirPort.
  • This could (but probably won't) be huge in large cluster/datacenter installations, seeing how the ATV draws 15-20 watts. At a full load, that's between 50 and 66MHz per watt. For example, my (dualcore) 1.83GHz MacBook came with a 60 watt power adapter, which works out to be about 61MHz per watt. Almost half the performance per watt as the AppleTV, but compared to an 20" iMac Core Duo (dualcore as well) does about 42MHz/W.

Let's compare this to a distributed computing environment. Something I've seen before at a friend's workplace (an atmospheric research lab where they run weather models on big off-the-shelf clusters) with about 16 off-the-shelf 350W (5600W total) power supplies and 16 2.4GHz single core procs (38.4GHz total). By my calculations, that chugs through to be 6MHz per watt.

This is of course not considering how much space they take up...

ProfessaFresh 21:45, 18 May 2007 (CEST)

Let's also consider the heat generated by an Apple TV and it's general lack of air-flow. I imagine this would potentially be kinda bad for your ATV's processor :(

Dobedobedoh 12:12, 19 May 2007 (GMT)

Yes, but one of the ATV advantages is that no fan exists, doesn't pull in dirt. I have set up a few clusters and various forms of distributed rendering over the years. These little boxes are rather impressive. Currently I am distributing several thousand that all chat happily to each other and handle the specific tasks very well. The heat stays constant, not as hot as my first generation Intel based MacBooPro 17" which one cannot touch at certain points on it's case. Current tests suggest ATVs could last longer than systems costing more while performing the same tasks. In fact they run cooler than many of the technologies I've worked with previously and seem to perform quite well under severe load 24/7. --JasonBelec 20:43, 15 September 2007 (CEST)

Has anyone tried this in Take 2 yet?