You are currently browsing the daily archive for December 15, 2009.

Want to know what it’s like living in Pacific Grove California? Well, it’s a wonderful small town with very little crime and a cost of living that is off this planet.

In Everett Washington where we used to own a house average home cost is 6.2 times the cost of the average family income. Expensive. Here in Pacific Grove? The average home is 18.3 times the average family income. And the houses are smaller, older, and cost way more to heat than is reasonable. My last months power bill was $429 for one month for a small 3 bedroom house. Water was $210 last month.

So, for perspective reasons think of paying for 3 homes in Everett for every 1 home in Pacific Grove. This is why we rent. I’d love to own a home here, but it isn’t going to happen.

A few other places we have lived:

  • Bothell Washington – 5.3 times
  • Everett Washington – 6.2 times
  • Fairbanks Alaska- 4.3 times
  • Biloxi Mississippi- 3.8 times
  • Butte Montana- 3.2 times
  • Pacific Grove California – 18.3 times

Statistics are from 2007 and found at

http://www.city-data.com/city/Pacific-Grove-California.html

and

http://www.city-data.com/

Most people just don’t understand the scale of building a supercomputer. Supercomputer? That’s just a big computer right?

In an article over on hpcwire Dave Turek (IBM Deep Computing) made a comment that sticks in my head. When building an exaflop system, 1 quintillion or 1 with 18 zeros, flops, or FLoating point Operations Per Second, think 1+1=2 as a flop. (none built yet, but planning in the near future). Using current designs, the memory system alone, think the RAM in your home computer, would require roughly 80 Megawatts of power. The equivalant of  1.5 Million light bulbs burning at the same time. I guess that would be enough lighting for more than 100,000 homes or so. Just for the memory.

An IBM Power7 MCM (MultiChipModule), think 4 CPU’s, pulls about 800 watts of power per Teraflop of performance, and yes, that’s good, really good compared to say your PC at home. So, 800 watts per Teraflop. 1000 Teraflops per Petaflop, 1000 Petaflops per Exaflop, or roughly 800 Megawatts for the CPU’s. We’re up to 880 Megawatts so far for memory and cpu’s and haven’t even turned on any disk space, cooling, facilities, networking, lights, infrastructure, etc, etc, etc,

So, frankly if your datacenter doesn’t have at least a Gigawatt of power free and clear, well, don’t bother trying for the next generation of supercomputers. Anyone got a spare Nuclear plant laying around?

And let’s not even get into what I’ll be charging to get the thing working. (-;

Blog Stats

  • 60,766 hits
December 2009
S M T W T F S
« Nov   Jan »
 12345
6789101112
13141516171819
20212223242526
2728293031