422
edits
Changes
Talk:GPU
,→CPW GPU article: new section
== Nvidia architectures == Afaik Nvidia did never official mention SIMD in their papers as hardware architecture, with Tesla they only referred to as SIMT . Nevertheless, my own conclusions are: * Tesla has 8 wide SIMD, executing a Warp of 32 threads over 4 cycles. * Fermi has 16 wide SIMD, executing a Warp of 32 threads over 2 cycles. * Kepler is somehow odd, not only about running multiple threads in sure how the compute units are partitioned. * Maxwell and Pascal have 32 wide SIMD, executing a Warp respof 32 threads over 1 cycle. Wavefront * Volta and Turing seem to have 16 wide FPU SIMDs, it is more about running multiple waves of Warps respbut my own experiments show 32 wide VALU. Wavefronts [[User:Smatovic|Smatovic]] ([[User talk:Smatovic|talk]]) 10:17, 22 April 2021 (CEST) == SIMD + Scalar Unit == It seems every SIMD unit has one scalar unit on GPU architectures, executing things like branch-conditions or special functions the same SIMD unit ALUs are not capable of. [[User:Smatovic|Smatovic]] ([[User talk:Smatovic|talk]]) 20:21, 22 April 2021 (CEST) == embedded CPU controller == It is not documented in the whitepapers, but it seems that every discrete GPU has an embedded CPU controller (e.g. Nvidia Falcon) who (speculation) launches the kernels. [[User:Smatovic|Smatovic]] ([[User talk:Smatovic|talk]]) 10:36, 22 April 2021 (CEST) == CPW GPU article == A suggestion of mine, keep this GPU article as an generalized overview of GPUs, with incremental updates for different frameworks and architectures. GPUs and GPGPU is a moving target with different platforms offering new feature sets, better open own articles for things like GPGPU, SIMT, CUDA, ROCm, oneAPI, Metal or simply link to hide memory latenciesWikipedia containing the newest specs and infos. [[User:Smatovic|Smatovic]] ([[User talk:Smatovic|talk]]) 21:29, 27 April 2021 (CEST)