We continue with the news from GTC 2016 event, and this is none other than the announcement of the first GPU based on the new generation architecture of the company ‘Pascal’. There is no doubt in it that the newly announced Pascal based Tesla P100 GPU is the fastest GPU to date. Tesla P100 built on a 16nm FinFET process node with die of 600m2 that hides 150.000 million transistors accompanied by the memory of new generation HBM2. Using second generation of High-bandwidth memory means that it will not be arriving in the OEM market until the first quarter of 2017. The new FinFET process allows NVIDIA to gain up to 2 times the performance per watt improvement on Pascal compared to the Maxwell GPUs.
NVIDIA announces 16nm Pascal Based Tesla P100 Featuring HBM2 Memory
Below you can see some of the core features of the Tesla P100:
- Extreme performance—powering HPC, deep learning, and many more GPU Computing areas;
- NVLink—NVIDIA’s new high speed, high bandwidth interconnect for maximum application scalability;
- HBM2—Fastest, high capacity, extremely efficient stacked GPU memory architecture;
- Unified Memory and Compute Preemption—significantly improved programming model;
- 16nm FinFET—enables more features, higher performance, and improved power efficiency.
Unfortunately, Jen-Hsun Huang did not reveal the number of CUDA cores, frequencies of silicon, or any other detail except the use of 16GB of HBM2 memory and floating point performance is 10.6 TFLOPs (FP32) and 5.3 TFLOPs (FP64). These numbers aren’t outperforming the AMD Radeon Pro Duo with 16 TFLOPs, as Pro Duo is based on the two dual-Fiji core.
[Update]: Tesla P100 is cut down GP100 with 10.6 TFLOPs. Full GP100 is 12 TFLOPs.