[Swordmaker]

An interesting calculation by user Macamac on Mac News Network shows that the Mac clusters are by far the fastest when considering R-Max per processor basis.

---

Just for fun I calculated the per-processor power of some of these servers by dividing the R-max (Maximal LINPACK performance achieved) by the number of processors.

The Virginia Tech System X took first place with a "score" of

5.56818182

Next highest was the SGI system at NASA-Ames Research center with a score of

5.10531496

The number one system, the IBM at LLNL scored only

2.08740234

the Cray system at Sandia 3.05, and the HP system at Los Alamos

1.69433594

---

If the speeds are scalable, had IBM built BlueGene using Mac xServes, it would have turned in 364,500 gigaflops!

However, since BlueGene processors are running at only 700Mhz compared to VT's xServes at 2.3GHz, IBM might get the same results by using their own PowerPC 2.3GHz processors! I wonder why they didn't.

Apparently, US Army contractor COLSA's 1556 Node (3112 processors) xServe cluster did not participate in this contest. Using the score factor above it would have turned in about 17,300 gigaflops putting it in the number 9 position on the list.

[Swordmaker]

Top 500 Supercomputer list released.

Apple has four xServe Clusters in the Top 500 including Virginia Tech's at number 14.

PING!!!

If you want on or off the Mac Ping List, Freepmail me.

[solitas]

Do any of these installations DO anything actually beneficial?

Or do they just "sit in the parking lot smokin' their tires" at each other?