AA Performance Issue?
An issue that we weren’t expecting, is traditional Multi-Sample or Super Sample Anti-Aliasing performance.
Based on our testing there is indication that MSAA is detrimental to AMD Radeon RX Vega 64 performance in a big way. In three separate games, enabling MSAA drastically reduced performance on AMD Radeon RX Vega 64 and the GTX 1080 was faster with MSAA enabled. In Deus EX: Mankind Divided we enabled 2X MSAA at 1440p with the highest in-game settings. The GeForce GTX 1080 was faster with 2X MSAA enabled. However, without MSAA, the AMD Radeon RX Vega 64 was faster. It seems MSAA drastically affected performance on AMD Radeon RX Vega 64.
In Rise of the Tomb Raider we enabled 2X SSAA at 1440p. Once again, we see AMD Radeon RX Vega 64 drop in performance. GeForce GTX 1080 was faster with 2X SSAA compared to Radeon RX Vega 64 with SSAA. Finally, in Dirt 4, which is playable at 8X MSAA, was faster on GTX 1080.
This is combined evidence enough that enabling forms of anti-aliasing like MSAA or SSAA are for some reason performance-impacting on AMD Radeon RX Vega 64. We need to do more testing on this for sure.
The conclusion so far is thus, when using shader based AA methods like SMAA, or FXAA or Temporal AA, or CMAA AMD Radeon RX Vega 64 performs much better and can compete with GTX 1080. However, if enabling any level of MSAA or SSAA then performance will decrease more on AMD Radeon RX Vega 64 and GTX 1080 will give more performance in that scenario. Therefore currently, AMD Radeon RX Vega 64 is best played with shader based AA methods versus traditional MSAA or SSAA in games for now. It will be interested to see if this can get addressed in a driver update.