The new product, launched during SC14, is pitched as double-your-everything: twice the performance and twice the memory bandwidth.
The K80 has 2x Kepler GK210 GPUS that claim 8.74 teraflops single-precision and 2.91 teraflops double-precision. Each board has 24 GB of GDDR5 memory (12 GB for each GPU) and 480 gigabytes/second memory bandwidth. There are 4,992 CUDA cores, and NVIDIA’s GPU booster (automatic overclocking) and dynamic parallelism support.
The product is targeted at HPC market, doubling the number of GPUs that can fit in the same server or workstation form factor as the K40.