Nvidia’s New Pascal GPU Architecture With Tesla P100
Today at the 2016 GPU Technology Conference in San Jose, Nvidia has announced their new Tesla P100 GPU for computing – the first one based on the next generation GPU architecture of the company that is coming to succeed the current Maxwell architecture. The new Pascal architecture is based on the new Pascal GP100 GPU and specifications wise is supposed to deliver very high performance for HPC, technical computing, deep learning, and many computationally intensive datacenter workloads. We are expecting to hopefully soon also see Pascal-based releases for the consumer GeForce series of GPUs and the details that were revealed for the Tesla P100 are making us even more excited to see the consumer GPUs based on Pascal.
The key new features for the Nvidia Pascal architecture are:
– 16nm FinFET process that enables more features, higher performance, and improved power efficiency.
– NVLink, a new high speed, high bandwidth interconnect for maximum application scalability.
– HBM2 video memory, a faster and higher capacity, extremely efficient stacked GPU memory architecture.
– Unified Memory and Compute Preemption — significantly improved programming model.
Among the interesting things to note are the higher operating frequencies for the GPU, especially for Tesla-class GPU, the improved bandwidth and performance with not much of a power usage increase. The new GP100 Pascal GPU also comes with support for the new Compute Capability 6.0 and CUDA 8.0. Of course the consumer GeForce GPUs will be slightly different, but it will still be interesting how the new features and capabilities will do for crypto currency mining. It seems that it will not be that long before we are able to check for ourselves. After the release however there still might be need for some additional optimizations in the available GPU mining software for it to be able to take advantage of the new Pascal-based GPUs. Do not forget that AMD is also working on their next generation of GPUs that will be made using 16nm process and will feature HBM2 memory offering better performance and power efficiency similar to that of Nvidia.
– For more details about the new CUDA 8.0 introduced with the Pascal GPU architecture…
– For more information about the new Pascal GPU architecture and the Tesla P100 GPU…
Other Similar Publications:
- In: General Info
- Tags: 16nm FinFET, Compute 6.0, CUDA 8, GP100, HBM2 memory, Nvidia GP100, Nvidia Pascal, Nvidia Tesla P100, NVLink, P100 GPU, Tesla, Tesla P100