🎮
Topic: How the
PlayStation 5 Pro implements machine-learning-based upscaling
without traditional tensor cores (specialized AI hardware found in many PC GPUs). (
YouTube)
Main points likely covered in the video:
- PS5 Pro uses machine learning upscaling technology similar in purpose to DLSS (NVIDIA) or XeSS (Intel) — but the console doesn’t have dedicated tensor cores like those found on many PC GPUs. (YouTube)
- Instead of tensor cores, Sony and AMD have optimized the existing GPU hardware to perform the matrix and tensor-like calculations required for AI upscaling. This involves efficient use of shader hardware and lower-precision math that’s suitable for neural network inference. (Reddit)
- The upscaling implemented in the PS5 Pro aims to boost image resolution and visual fidelity (e.g., upscaling to 4K) while keeping performance high. This likely refers to Sony’s PlayStation Spectral Super Resolution (PSSR) upscaling technology, which is an AI-style neural upscaler built into the console’s GPU pipeline. (TweakTown)
- The video probably explains the architectural differences between having dedicated tensor cores versus doing ML work on general GPU compute units — showing how Sony gets around the lack of dedicated hardware by optimizing standard GPU execution paths. (Reddit)
🔧
What this means in practice:
Instead of relying on dedicated neural-network hardware, the PS5 Pro performs machine learning tasks using its existing graphics processing units in a highly optimized way — letting developers include better upscaling in games without special tensor accelerators. (
YouTube)
es waren 60-70 fps. Wenn man nicht mal stabile 60 fps hat, soll man FG doch gar nicht erst einschalten. 
