NVIDIA Volta GV100 GPU Twice As Fast As Pascal P100; Supercomputer Chip To Hit 9.5 TFLOPs FP64 Compute; [Video]
NVIDIA Volta is being prepped for launch in the next generation supercomputers known as Summit and Sierra. There is a little information about Volta GPU specifications but an analysis reveals the details of Summit supercomputer, which says that it can be an insanely fast chip that will be capable of delivering multi-Tflops, compute power in the HPC market.
How insanely fast NVIDIA Volta could be, answer is given back in November, when the Next Platform suggested what could possibly be the speed and power of Volta based on the information on 'Summit' Power Tesla AI Supercomputer.
When NVIDIA had announced their Pascal GP100 GPU at GTC 2016, they have called it the largest chip endeavour in the history of humanity. And the pascal GP100 was indeed the greatest chip of 2016, which aimed to power the HPC and data-centre market with performance that never before seen in the graphics industry.
NVIDIA also had utilized Pascal GP100 GPUs in their own DGX SaturnV, supercomputer that is designed to help them build smarter cards and next generation GPUs. Just a year after Pascal launch in the HPC market, NVIDIA is planning to introduce their next grand chip for the HPC market, codenamed Volta. Details of the chip first emerged back at GTC 2015 when NVIDIA showcased what they predict to be the estimated performance output of their upcoming chips.
According to WccFtech, The only thing that Pascal currently lacks is the 32 GB capacity, but that is mostly an issue due to HBM production which has already ramped up and now we can expect a full GP100 configuration with 32 GB capacity.
The latest details for the Summit Supercomputer have been confirmed and they are incredible if we look from a HPC perspective. The latest Summit Supercomputer has 5-10x improvement in application performance boost the Titan supercomputer that featured the Kepler GK110 GPU architecture. The Titan was comprised of 18,688 nodes rated at 1.4 TF, per node. Where the Summit features around 4,600 nodes with a rated compute output of over 40 TF, per node.
Each node has the 2 IBM Power9 CPUs and 6 NVIDIA Volta V100 GPUs. NVIDIA's NVLINK2 interconnect will be fully integrated between these nodes and the system would consume 13 MW peak power that is just 4 MW more than the Titan supercomputer, which is 9 MW, for over 10x the performance improvement.
NVIDIA previously has said that, NVIDIA Volta GV100 GPUs will deliver Single precision floating General Matrix Multiply of 72 GFLOPS/Watt compared to 42 GFLOPs/Watt on Pascal GP100.
Using the mentioned ration, a Volta GV100 based GPU with a TDP of 300W can theoretically deliver 9.5 TFLOPs of double precision performance. This is almost twice that of the current generation GP100 GPU.
With the final configuration of Volta V100 and IBM Power9 CPUs in place, the Summit Supercomputer will be the top performing machine in the world with performance crossing the 250 Petaflops mark.
