346 Designing Scientific Applications on GPUs
running at 2.4GHz. It provides a RAM memory of 12GB with a memory band-
width of 25.6GB/s and it is equipped with two NVIDIA Tesla C1060 GPUs.
A Tesla GPU contains in total 240 cores running at 1.3GHz. It provides 4GB
of global memory with a memory bandwidth of 102GB/s, accessible by all its
cores and also by the CPU through the PCI-Express 16x Gen 2.0 interface
with a throughput of 8GB/s. Hence, the memory copy operations between the
GPU and the CPU are about 12 times slower than those of the Tesla GPU
memory. We have performed our simulations on a cluster of 24 CPU cores and
on a cluster of 12 GPUs. Figure 13.4 describes the components of the GPU
cluster of tests.
Linux cluster version 2.6.39 OS is installed ...