Turing Tensor Cores: Leveraging Deep Learning Inference for Gaming

Though RT Cores are Turing’s poster child feature, the tensor cores were very much Volta’s. In Turing, they’ve been updated, reflecting its positioning as a gaming/consumer feature via inferencing. The main changes for the 2nd generation tensor cores are INT8 and INT4 precision modes for inferencing, enabled by new hardware data paths, and perform dot products to accumulate into an INT32 product. INT8 mode operates at double the FP16 rate, or 2048 integer operations per clock. INT4 mode operates at quadruple the FP16 rate, or 4096 integer ops per clock.

Naturally, only some networks tolerate these lower precisions and any necessary quantization, meaning the storage and calculation of compacted format data. INT4 is firmly in the research area, whereas INT8’s practical applicability is much more developed. Regardless, the 2nd generation tensor cores still have FP16 mode, which they now support in a pure FP16 mode without FP32 accumulator. While CUDA 10 is not yet out, the enhanced WMMA operations should shed light on any other differences, such as additional accepted matrix sizes for operands.

Inasmuch as deep learning is involved, NVIDIA is pushing what was a purely compute/professional feature into consumer territory, and we will go over the full picture in a later section. For Turing, the tensor cores can accelerate the features under the NGX umbrella, which includes DLSS. They can also accelerate certain AI-based denoisers that cleanup and correct real time raytraced rendering, though most developers seem to be opting for non-tensor core accelerated denoisers at the moment.

Turing RT Cores: Hybrid Rendering and Real Time Raytracing The Turing Trio: TU102, TU104, & TU106
POST A COMMENT

112 Comments

View All Comments

  • Wwhat - Wednesday, October 17, 2018 - link

    What the article ignores is that ray tracing went through a long revolution, and one of the findings at one point for example was that triangles weren't covering the need for advanced ray-tracing, after which things like NURBS were thrown into the mix.
    My point being that you would think the RT for gaming development would not start from point A and slowly meander to all the evolutionary steps with constant hardware updates. But it's not clear at the moment if the Nvidia team is 'keeping it simple' for now in their package, or if that's just how it is presented for easy presentation to the crowd.
    Reply
  • falak - Wednesday, September 9, 2020 - link

    https://www.thenora.co.uk/hacking-apps-games/ Reply

Log in

Don't have an account? Sign up now