XServe cluster achieves 217 GigaFlops

[medea]

http://daugerresearch.com/fractaldemos/JPLXServes/JPLXServeClusterBenchmark.html
Dauger Research today announced that its software achieved over 217 billion floating-point operations per second on a cluster of 33 XServes from the Applied Cluster Computing Group at NASA's Jet Propulsion Laboratory (JPL). Using software provided by Dauger Research, the JPL group has begun "running parallel computing code on what is presently the largest working XServe cluster known to exist." Among the codes executed on the cluster, the "AltiVec Fractal Carbon demo was one that achieved over 1/5 of a TeraFlop. This result is the largest calculation yet performed on an XServe cluster."

 

[medea]

chart

 

[MacBandit]

Was this all in one rack? How many XServes can be put in one rack?

 

[Chaszmyr]

Originally posted by MacBandit
Was this all in one rack? How many XServes can be put in one rack?

All one rack indeed, 42 Xserves can go on one rack.

 

[Mr. Anderson]

I wonder if they were dual XServes or single? I'm guessing they went all out and got as many as they could. Quite impressive results, I wonder what they're going to use it for?

D

 

[unreg]

dual 1gig

 

[Ifeelbloated]

Hopefully pull us out of this economic slump?

 

[shadowfax0]

this is fantastic! I think the big news is when they compare it to the 76 node cluster of dual-533's, the powermacs outperform this cluster only by having TWICE AS MANY nodes! Amazing!!

 

[al256]

Well, Apple could make the worlds fastest super computer by clustering 6,105 XServes together. It would be around 37 terraflops of raw computing power. Just a thought.