The NVIDIA Tesla™ K40 represents the latest and most advanced Kepler GPU architecture from NVIDIA. Based on the new GK110B GPU, new levels of supercomputing processing capability is now available in a wide range of HP Z-workstations. Based on the massively parallel re-worked CUDA architecture, it features many “must have” HPC capabilities including 12GB ECC memory and an L1/L2 cache design for improved bandwidth and data sharing. The Tesla K40 features innovative technologies like SMX, Hyper-Q, and Dynamic Parallelism. - Delivers up to 1.5 Tflops of double-precision peak performance, enabling a single workstation to deliver ~3 Tflops with two K40's. The K40 single precision peak performance is 3.6 Tflops. - New CUDA core architecture delivers up to 3x more performance per watt compared to the previous generation Fermi GPU based Tesla cards. - When enabled, converts power headroom to higher clocks and achieve even greater acceleration for various HPC workloads on Tesla K40. - Enables GPU threads to automatically spawn new threads. This allows elimination of unnecessary program control transfers between CPU and GPU and enables GPU acceleration of a broader set of algorithms. - This feature enables multiple CPU cores to simultaneously utilize the CUDA cores on a single Kepler GPU. The result is a dramatically increased level of average GPU utilization. - Meets a critical requirement for computing accuracy and reliability for workstations. Offers protection of data in memory to enhance data integrity and reliability for applications. Register files, L1/L2 caches, shared memory, and DRAM all are ECC protected. - Doubles the size of datasets/models that can be loaded into the Tesla K40 onboard memory. Maximizes performance and reduces data transfers by keeping larger data sets in local memory that is attached directly to the GPU.