Talk:Pascal (microarchitecture)

Rewrite from scratch

This article needs to be rewritten, to be more like the Kepler (microarchitecture) article. The Maxwell (microarchitecture) article is a rather bad example. Possible sources besides Nvidia:

• http://www.anandtech.com/show/10222/nvidia-announces-tesla-p100-accelerator-pascal-power-for-hpc
• http://www.anandtech.com/show/10304/nvidia-announces-the-geforce-gtx-1080-1070
• http://forums.anandtech.com/showthread.php?t=2463869
• etc. User:ScotXW^t@lk 17:57, 14 May 2016 (UTC)

Streaming Multiprocessor "Pascal"

One "Streaming Multiprocessor" corresponds to one Compute Unit. An SMP encompasses 64 CUDA cores (Maxwell = 128, Kepler = 192).

• Each SMP (Streaming Multiprocessor "Pascal") has 64 CUDA cores (Maxwell = 128, Kepler = 192)
• L2 Cache Size of 4096 KiB (Maxwell = 3072 KiB, Kepler = 1536 KiB)
• The Register file size of the entire GP100 GPU is 14336 KiB (Kepler = 3840 KiB, Maxwell = 6144 KiB) User:ScotXW^t@lk 18:05, 14 May 2016 (UTC)