[nhkcode]

> It's not like AMD is above artificially limiting the performance of certain workloads when it suits them, they do it with CAD applications unless you pay the workstation tax.

I'm curious about that, do you have an example? I only know of nvidia consumer drivers giving you abysmal performance if you try to use fixed function pipeline line or wireframe rendering with opengl. Or at least they did a couple of years ago. As soon as you used shaders to render the same lines everything got super fast. I always suspected it's their way to force users of CAD systems to buy their pricey professional cards. No game would use opengl line rendering without shaders or if any game did in the long gone past it's probably so old and the line count so low that it would be still fine with the artificial slowdown.

[jsheard]

You can see it in action here: https://techgage.com/article/specviewperf-13-viewport-perfor...

The consumer RX580 and workstation WX7100 are based on the same Polaris silicon, and the consumer card is configured with a higher power limit of 185W vs 130W, but the workstation card usually performs similarly or much better anyway. It's most egregious in the Siemens NX test where the workstation card is about 8 times faster.

I think AMD strategically throttles legacy OpenGL paths similar to Nvidia, Autodesk Max/Maya use modern shader-based viewports and in those the consumer card pulls ahead as you'd expect given its higher power limit.

[nhkcode]

That is in interesting comparison. I really wonder if the differences between consumer and pro cards are all just in the driver or if there is a hardware reason for it too. The difference between the P2000 and the gtx 1080 ti in the Siemens NX benchmark looks insane.