At the heart of the Tesla V100 is NVIDIA's Volta GV100 GPU, which features a staggering 21.1 billion transistors on a die that measures 815mm2 (this compares to 12 billion transistors and 610mm2 respectively for full-fat Pascal GP100). According to NVIDIA, GV100 is built on a 12nm FinFET manufacturing process by TSMC.
When it comes to brute power, the GV100 has 80 SMs with 40 TPCs bringing a total of 5,120 CUDA cores, compared to 3585 CUDA cores for the GeForce GTX 1080 Ti and previous gen Tesla P100 AI accelerator, for example. The new Volta GPU delivers 15 TFLOPS FP32 compute performance and 7.5 TFLOPS of FP64 compute performance. You’ll also find 16MB of cache and 16GB of second generation High Bandwidth (HBM2) memory with 900GB/sec of bandwidth via a 4096-bit interface. Interestingly enough, NVIDIA has also divulged the Boost clock for the GPU, which comes in at 1455MHz.
But that’s not all — the GV100 has dedicated Tensor Cores (640 in total) that are designed specifically to speed up AI workloads. With these cores enabled, a Tesla V100 is capable of pushing 120 TFLOPS of deep learning performance (which NVIDIA says is equal to 100 traditional CPUs). The dedicated Tensor Cores also allow for a 12x uplift in deep learning performance compared to Pascal, which relies solely on its CUDA cores.
"Artificial intelligence is driving the greatest technology advances in human history," said NVIDIA CEO Jen Hsun Huang. "It will automate intelligence and spur a wave of social progress unmatched since the industrial revolution. "Deep learning, a groundbreaking AI approach that creates computer software that learns, has insatiable demand for processing power. Thousands of NVIDIA engineers spent over three years crafting Volta to help meet this need, enabling the industry to realize AI's life-changing potential.”
NVIDIA is targeting a Q3 2017 release for the Tesla V100 with Volta, but of course we’re unsure of what the timetable is for a GeForce GTX derivative family of consumer graphics cards. However, looking at the architecture surrounding GV100, both big data scientists, machine learning developers and gamers alike are in for a real treat and some serious new firepower.