|Machine type||RISC-based shared-memory multi-processor.|
|Models||PRIMEPOWER 900, 1500, 2500.|
|Operating system||Solaris (Sun's Unix variant).|
|Compilers||Parallel Fortran 90, OpenMP, C, C++.|
|Vendors information web page:||http://primepower.fujitsu.com/en/index.html|
|Year of introduction||2002.|
|Model||PRIMEPOWER 1500||PRIMEPOWER 2500|
|Clock cycle||1.89 GHz||1.82 GHz|
|Theor. peak performance|
|Per proc. (64-bits)||3.78 Gflop/s||3.64 Gflop/s|
|Maximal||121 Gflop/s||466 Gflop/s|
|Memory/node||≤ 4 GB||≤ 4 GB|
|Memory/maximal||≤ 128 GB||≤ 512 GB|
|No. of processors||4—32||8—128|
|Aggregate||17.2 GB/s||133 GB/s|
We only discuss here the PRIMEPOWER 1500 and 2500 as the smaller models have the same structure but less processors (maximally 16 in the 900 model). In many respects this machine is akin to the SUN Fire E25K. The processors are 64-bit Fujitsu implementations of SUN's SPARC processors, called SPARC 64 V processors and they are completely compatible with the SUN products. Fujitsu's SPARC64 processors have extensive RAS features for all CPU components, caches, registers, execution units, etc., that make them very reliable. The processors are available in a 1.89 GHz and 1.82 GHz for the 1500 and 2500 models respectively. Also the interconnection of the processors in the PRIMEPOWER systems is like the one in the Fire E25K: a crossbar that connects all processors at the same footing, i.e., it is not a NUMA machine.
Unfortunately, there is no sound technical information available beyond the data sheets that are provided via Fujitsu's web site. These data sheets omit almost all information about the bandwidth of the interconnect be it point-to-point, bi-sectional, or aggregate. Only for the PRIMEPOWER 2500 an aggregate bandwidth is quoted: a very respectable 133 GB/s stemming from the two 66.5 GB/s crossbars. Judging from the available information the 1500 system is more positioned as a commercial server than as a high-performance computer while the 2500 model is also targeted at the HPC market. Fujitsu offers the possibility to link many PRIMEPOWER 2500s together by a very fast (4×4 GB/s) optical link to build very large configurations.
In  a cluster of 18 fully configured PRIMEPOWER 2500s with the older 1.3 GHz processors is used to solve a linear system of order N= 658,800 This yields a performance of 5.4 Tflop/s with an efficiency of 45% on 2,304 processors.