White Papers

Extreme GPU Computing
Page 7
3. The Tesla K80 GPU accelerator board
The latest HPC-focused Tesla series General Purpose Graphic Units (GPU) released from NVIDIA is the
Tesla K80. From an HPC perspective, the most important improvement is the 1.87 TFLOPs (double
precision) compute capacity, which is about 30% more than K40, the previous Tesla card. The K80 auto-
boost feature automatically provides additional performance if additional power head room is
available. The internal GPUs are based on the GK210 architecture and have 4,992 cores, which
represent a 73% improvement over K40. The K80 accelerator board has a total memory of 24GBs, which
is equally divided between the two internal GPUs; this is a 100% more memory capacity compared to
the K40. The memory bandwidth in K80 has improved to 480 GB/s. The rated power consumption of a
single K80 is a maximum of 300 watts.
Figure 5: The K80 GPU board with two internal GPUs connected via a PCIe switch
Combining K80s with the latest high GPU density design from Dell, the PowerEdge C4130 provides an
extra-ordinarily powerful compute node. The C4130 can have up to four K40 or K80 GPU boards in a 1U
form factor. In addition, it is available in several workload-specific configurations, potentially making
it a better fit for specific HPC codes.