After successfully launching a Pascal-based TITAN X that promised 11 teraflops (TFLOP) in July 2016 and also improved the performance of a tad with a fully unlocked Pascal GP102 core in the form of TITAN Xp (12 TFLOP), recently NVIDIA is reportedly preparing to roll back an entirely new GPU aimed at researchers and scientists, TITAN V.
The graphics card has been announced to coincide at the Conference and Workshop on Neural Information Processing Systems (NIPS) in Long Beach, California. The new TITAN V is based on the NVIDIA Volta architecture, and with that shift comes a big jump in calculating performance with a certain workload. According to NVIDIA, the redesigned Tensor Core TITAN V has a parallel integer and independent floating-point data line, helping to trigger a 9x whopping increase in Tensor Performance compared to TITAN Xp that delivers 110 confusing TFLOPs for Deep Learning workloads. Single-Precision performance is likely to be present in 14 better TFLOPs, and can process depending on the frequency of the existing GPU.
All that power is incorporated into the 21.1 billion GV100 GPU transistors that are manufactured using the high-performance TSMC 12nm FFN process, which has been customized for NVIDIA. TITAN V itself reportedly will be filled with 5120 CUDA core, 640 Tensor Core, and has a base / boost clock of 1200MHz and 1455MHz respectively. HBM2 memory runs at 1.7Gbps with 3072 bit interface (for an effective bandwidth of 653GB / sec). While the GV100 GPU that drives the NVIDIA Tesla V100 accelerator is paired with 16GB HBM2, the TITAN V comes with 12GB of HBM2. While other details that NVIDIA is willing to share at this time include mentioning that the upcoming TITAN V will be supplied with new combined L1 data cache and shared memory units to improve overall performance. The power efficiency itself is reported to have increased dramatically with Volta,
As stated by NVIDIA CEO Jensen Huang, NVIDIA’s vision for Volta is to push beyond the limits of high-performance computing and AI by solving new problems with processor architectures, instructions, numerical formats, memory architectures and new processors. With TITAN V, it will put Volta into the hands of researchers and scientists around the world so as to help bring about a new breakthrough.
TITAN V will be added to the NVIDIA GPU Cloud, which provides quick access to in-depth learning software optimized for artificial intelligence (AI) development. This is certainly very welcomed by one of the NVIDIA Inception partners namely Deep Cognition through Chief Operating Officer (COO) Jeremy Antoniuk.
Given TITAN V is designed for use in a research environment, then do not expect the price of an enthusiastic GPU class. Subject to its own existence, as can be seen in the NVIDIA online store , TITAN V reportedly is now marketed with the official price of $ 2,999.