Last night I watched a major animated feature (the reason many of my coworkers have IMDB entries) that was in cinemas earlier this year and is currently doing well on DVD. Said feature was rendered almost exclusively on the $BIGNUM node cluster, capable of $IMPRESSIVENUM teraflops, that I am currently working on. When does “working on” become “managing”?
So if they release another film, you get a credit?
Yeah, that’s how it works.
Please post cluster porn.
That would probably be against all kinds of stuff in my contract. You’ll have to use your imagination :)
/me imagines 1024 quad dual core AMD 64 nodes on fibre running LinuxBIOS with a liquid cooling network 4x as complex as the data network requiring 2 fulltime plumbers, and frankly I think that’s stupid. What you should be doing is 4096 nodes of those awesome EPIA mini-itx fanless dual cpu boards with a custom linux kernel running off of compact-flash. You can cram 16 of those into 4u. We’re talking roughly 64gHz for every 4u compared to 32gHz per 4u with your REDICULOUS quad dual core cluster. And no plumbers needed. Granted you’re sacraficing throughput with the 100mbps network, but we’re talking about what I assume is a render farm here, not a real-time analytical processing system for Los Alamos.