[kllrnohj]

Whenever AMD's CPUs have been price/performance competitive they've always been a fantastic choice. Especially now, when with Ryzen/Threadripper/Epyc you not only got more cores for the less money, you also got better security.

AMD's GPU drivers have been perfectly fine since the 9700 Pro. That's a 16 year old myth at this point, let it go.

[TomVDB]

https://twitter.com/_humus_/status/1018846492273119233?lang=...

ATI started rewriting their OpenGL driver in 2004-2007. Long after 9700 Pro’s prime.

As told by the author, it was a long lasting disaster.

[kllrnohj]

Do you have an actual argument here? The story, while interesting, doesn't really contribute either way. Per the author's own story the switch to the new driver broadly didn't happen until after it was finally stable, and the legacy driver continued to receive performance optimizations in the meantime.

It's an interesting story of project management nightmares, but it doesn't provide any argument to the state of AMD's drivers as experienced by end users either way.

In terms of stability we do have some large-scale metrics on that front, such as Vista's crash blaming. Those metrics don't support claims that AMD's drivers are less stable than Nvidia's, as the Nvidia was responsible for 28.8% of Vista crashes while ATI was 9.3%. Given more Nvidia than ATI users that's not necessarily damning, but it also clearly disagrees with the notion that ATI is deeply unstable while Nvidia is rock solid.

And keep in mind during the 9700 Pro's prime all the way up to today OpenGL is used by nearly nothing on Windows. We're already talking about the niche use case.

[TomVDB]

If the driver became highly unstable after 2004, you can’t claim that the driver has been stable since 2002. That is all.

[kllrnohj]

> driver became highly unstable after 2004

The ATI dev on twitter made no such claim. In fact he never made any claim about stability at all. Just that the new driver was missing functionality, so only specific things (like Doom3) got it and they were shipping 2 OpenGL drivers as a result during 2004-2007. End-users weren't broken during that timeframe. The old driver didn't suddenly break and get super unstable. If anything the complaint is that the old driver was too stable, it wasn't getting new features & changes fast enough.

[throwaway2048]

Its really not a myth, AMD drivers on windows have completely terrible openGL performance for instance.

[kllrnohj]

Do you have any benchmark comparisons to support this claim? All I can find in searching are random threads of people saying this, but nobody providing any actual evidence or comparisons.

The few games using OpenGL on Windows I can find show AMD's performance being perfectly competent:

https://www.pcgamer.com/doom-benchmarks-return-vulkan-vs-ope...

https://www.hardocp.com/article/2014/05/21/wolfenstein_new_o...

The nvidia cards were on average a bit faster, but that was true in DX as well. Meaning it wasn't just "lol amd opengl drivers"

And of course Linux testing, where OpenGL is far more common, shows no major disparity, either, which you already know hence the qualifier of "on windows" but that qualifier makes no sense. It's going to be the bulk of the same code between windows & linux for the driver, as has also been rather well tested & verified.

Which gets back to "it's a goddamn 16 year old myth" that literally spawned out of Nvidia's hyper aggressive opengl optimization of Quake 3.

[TomVDB]

You’re confusing performance with stability. See my other reply about OpenGL drivers.

I’ll take the word of a well respected former ATI employee.

[kllrnohj]

You're clearly confused as to the topic you're replying to. I responded with benchmarks to someone who said performance was bad. Stability was not the topic of discussion in this subthread.