The Interesting Tale Of AMD's FirePro Drivers

The Maya test with SPECViewPerf 9 is just like the EnSight results. The Catalyst 9.2 driver did well but then regressed in Catalyst 9.3 and its performance was living sub-standard for a year until the spike in Catalyst 10.2 where it was a huge spike compared to Catalyst 10.1, but only up marginally compared to Catalyst 9.2. The driver then regressed and this summer on AMD's second try the performance went back up and it's been living that way since, but it's not any faster compared to Catalyst 9.2: 218 then versus 214 now. The peak was at 256 in Catalyst 10.2. Worth noting though is that between Catalyst 9.2 and 9.3 is that the OpenGL image quality was the same and there weren't any visible differences like in some past driver tests where some areas of the screen were black or miss-rendered that would result in an inflated result.

The Pro/Engineer results are like the other tests where there were there are the two spikes in 2010, but after the second jump the performance has since regressed.

With the SolidWorks test, the performance began to mature in late 2009 after dropping between Catalyst 9.2 and 9.3 and it too shared the Catalyst 10.7 performance boost, but since then has fallen. The Catalyst 10.12 performance has actually fallen compared to where it was at with Catalyst 9.2 (52 vs. 50 now for the weighted geometric mean).

With the TCVIS test, like SW, there were some improvements in late 2009 and then again this summer, but they were not sustainable. The Catalyst 10.12 driver is not any faster than it was with Catalyst 9.2.