General Purpose Desktop GPU Information Page

This page contains one table listing key attributes of general purpose desktop GPU of Nvidia and AMD (ATI). It includes representative GPU products for each architecture but not mean to be comprehensive.

Notes:

     *:The evolution of GPGPU architecture of Nvidia is: Tesla, Fermi, Kepler, Maxwell, Pascal.

     †:The evolution of GPGPU architecture for AMD (ATI) is: TeraScale, Graphics Core Next (GCN). TeraScale has three iterations: TeraScale 1, TeraScale 2 (VLIW5) and TeraScale 3 (VLIW4). TeraScale2 is used in "Evergreen" series GPUs. GCN also has three iterations: GCN 1.0 (Southern Islands), GCN 1.1 (Sea Islands), GCN 1.2 (Volcanic Islands).

     ‡:The number of streaming processors is the number of unified shader units.

     #:In TeraScale architecture, the SIMD number is equal to the number of compute unit. However, each compute unit in GCN architecture contains 4 SIMDs.

     **:The data of AMD (ATI) GPU is from [1] and [2], while the data of Nvidia GPU mainly from [3] and [4] and whitepapers of the different architectures.

This page is maintained by members of the VLSI Computation Laboratory at UC Davis.

Last update: June 18, 2016