The eight GPU (eight Tflop), 4U, rack-mounted supercomputer was rendered even more impressive by its specially engineered motherboard – sporting no less than eight PCIe Gen 2 x16 slots, ensuring that each C1060 Tesla GPU gets blasted with data over full bandwidth
Backing up the eight GPUs with their whopping 1,920 processor cores (8 x 240), the system also sports two Intel Xeon (Nehalem) DP quad-core W5590 processors, up to 144GB of DDR3 system memory, two internal 2.5in SATA drives, two 1,200W (2,400W) non-redundant power supplies, four Intel 82574l Gigabit Ethernet controllers, IPMI 2.0W/IKVM support, integrated ASPEED AST 2050 VGA controller and a Linux OS.
The basic configuration costs around $21,140
PREDICTION: Thousand+ CPU workstations- mainstream chip vendors 2009-2012
By counting CPU cores, this server has 1920 CPUs from mainstream chip vendor Nvidia.
In other supercomputing news, The University of Tennessee’s supercomputer, Kraken, has broken a major barrier to become the world’s first academic supercomputer to enter the petascale, performing more than 1 thousand trillion operations per second, a landmark achievement.
Kraken, a Cray XT5 computer, also has a massive amount of memory to store the information used in scientists’ large-scale projects. With 129 terabytes of memory, Kraken can store the equivalent of more than 10 million phonebooks.
As the first computer managed by a university to pass this milestone, Kraken puts UT in front of other major computing centers across the country, while enhancing the national research effort through Kraken’s role in NSF’s nationwide network of computers called TeraGrid, the largest computational platform for open scientific research. Kraken is housed in the computing facilities at Oak Ridge National Laboratory, which are also home to another petascale computer, called Jaguar.