AMD at SIGGRAPH 2017 uncovered their new effective element called the Rapid Packed Math, which essentially parts the drifting point estimations in handling an information ask for, bringing about a faster appropriation of data.
AMD and Ubisoft uncovered that Wolfenstein 2: The New Colossus and Far Cry 5 will bolster Rapid Packed Math, this component starting at now is restrictive to the RX Vega arrangement of GPUs. As per the engineers at AMD, RPM will process two math directions for the time devoured for one.
This implies both the diversions will perform considerably speedier gliding point 16 FP16 processing than with the standard Floating point 32bit figuring. Fast Packed Math is as of now a selective AMD highlight and NVIDIA has made no declarations what so ever to help it. Since it's an equipment highlight, proprietors of current NVIDIA GPUs won't profit by it.
Certain aficionados additionally censured the thought depicted by AMD for utilizing 16 bit FP register against the FP32. In view of the talks from Neogaf discussion, FP16 is a lower quality estimation being utilized by the GPU, specifically influencing the picture quality being created, aside from when it's utilized as a part of those ranges of the diversion where protest detail would not make any difference and will be less detectable to the eyes, may bring about certain level of execution pick up without a visual quality misfortune. The degree of which is yet to be seen.
Aside from the way this is an equipment include and without the greater part of the diversion designers supporting this element, it will wind up plainly futile. You may Google for 16 bit exactness VS 32 bit accuracy to check picture quality contrasts.
AMD likewise guarantees that it's the total help for MS DX12 API. The reality still remains that if the engineers will exploit all the DX12 highlights, and will they demonstrate sufficient to finish with NVIDIA GPUs. Considering the way that DX12 and VULCAN are as yet attempting to get standard consideration from diversion engineers.