Tensor Core equivalent Likely to Get Embedded in AMD rDNA3
AMD software programmers have begun to distribute new fixes for the forthcoming GFX11 architecture, also known as RDNA3. According to a recent patch, AMD is working on their own instructions that can operate on matrices.
AMD's RDNA3 graphics IP is on its way, and we're learning more about the new architecture. AMD developers added a new instruction to the LLVM compiler's back end called Wave Matrix Multiply-Accumulate (WMMA). This instruction will be available on the GFX11 RDNA3 GPU architecture. Tensor Core has been present on NVIDIA GPUs since the Volta design, and AMD WMMA might be viewed as a response to that. NVIDIA has released a sophisticated instruction to boost NVIDIA's AI-based super-resolution technology known as DLSS.
Intel also has its own XMX/DPAS instructions that work on matrixes and can be utilized to accelerate upcoming XeSS technologies. AMD will be able to handle 16x16x16-dimensional tensors in FP16 and BF16 forms with WMMA. AMD is implementing new methods for handling matrix multiply-accumulate operations with these instructions. This is similar to what NVIDIA does with Tensor Cores.
“rocWMMA provides a C++ API to facilitate breaking down matrix multiply accumulate problems into fragments and using them in block-wise operations that are distributed in parallel across GPU wavefronts. The API is a header library of GPU device code, meaning matrix core acceleration may be compiled directly into your kernel device code. This can benefit from compiler optimization in the generation of kernel assembly and does not incur additional overhead costs of linking to external runtime libraries or having to launch separate kernels.
rocWMMA is released as a header library and includes test and sample projects to validate and illustrate example usages of the C++ API. GEMM matrix multiplication is used as primary validation given the heavy precedent for the library. However, the usage portfolio is growing significantly and demonstrates different ways rocWMMA may be consumed.”
Gigabyte Announces HPC Systems Powered by NVIDIA A100 Tensor Core GPUs - 05/15/2020 08:00 AM
Gigabyte today disclosed four NVIDIA HGX A100 platforms under development. These platforms will be available with NVIDIA A100 Tensor Core GPUs. NVIDIA A100 is the first elastic, multi-instance GPU tha...
TYAN Launches AI-Optimized Server Platforms Powered by NVIDIA V100S Tensor Core GPUs - 05/08/2020 08:23 AM
TYAN launched the latest GPU server platforms that support the NVIDIA V100S Tensor Core and NVIDIA T4 GPUs for a wide variety of compute-intensive workloads including AI training, inference, and super...
NVIDIA Announces Tesla T4 Tensor Core GPU - 09/13/2018 09:00 AM
Fueling the growth of AI services worldwide, NVIDIA today launched an AI data center platform that delivers the industry’s most advanced inference acceleration for voice, video, image and re...
Senior Member
Posts: 1821
Joined: 2013-06-04
Games will look the same (or slightly worse) but with more frame rates. At least that's what I understood from what Nvidia is doing.
Senior Member
Posts: 2327
Joined: 2015-06-11
ROCm isn't used in games at all. It is used manly with Instinct cards on linux.
Senior Member
Posts: 929
Joined: 2007-09-03
I guess offloading operations like FSR calculations to there, or they could be doing ray tracing with them.
Senior Member
Posts: 12593
Joined: 2013-01-17
Glitch in brain made me read the title "Tenor Core equivalent ..."
Senior Member
Posts: 1639
Joined: 2012-11-06
Right, ive read through that and its all Greek to me, in a nutshell will games look better but with the same/improved framerates?