AMD Radeon Software Crimson Edition Is A Letdown On Linux

[dungeon]

1 -- Comprehensive reading is a valuable skill, adopt it sooner, rather than later.
2 -- I didn't ask for your help, I just shared an experience I had had with AMD's drivers.

Thanks for playing though.

You are welcome On "sooner, rather that later" yes, something they get sooner something else later depends on priority.

Supporting every kernel and x relases on its release dates, obviosly isn't their priority and never has been.

On "AMD's drivers" those are also opensource drivers, those are different story.

[oleid]

Just because Michael called it "leaked" doesn't mean they have to be leaked. Funny, it isn't it? The website Michael linked to in that article doesn't tell us anything about leaks, but that these slides are from an AMD presentation. So yes, you did miss something.

Michael, the guys from reddit and basically the rest of the web. But important information is missing:

1. Where is the talk itself? Slides only are not the full information.
2. What do those footnotes mean? So far, I didn't find an explanation in the slides.
3. What do they take as performance basis? The last beta driver? Or maybe the last stable release? Probably the 2nd slides gives a hint, where "Since AMD Catalyst Omega Driver" is written.

Furthermore, there is no official press release providing these slides. You don't get them from amd.com. How would you call it else, if not leaked information?

I guess users of the proprietary AMD drivers have to wait for the driver providing Vulkan in order to get really new stuff.

[F1esDgSdUTYpm0iy]

You are welcome On "sooner, rather that later" yes, something they get sooner something else later depends on priority.

Supporting every kernel and x relases on its release dates, obviosly isn't their priority and never has been.

On "AMD's drivers" those are also opensource drivers, those are different story.

O.o...

[bridgman]

It also would be great if AMD wouldn't promise performance gains and then fail to deliver, don't you think?

If you mean "it would be great if AMD's Windows SW marketing folks wouldn't promise performance gains (and deliver AFAIK), but fail to make it clear that they were talking about Windows/DX not Linux/OpenGL" then yes I agree.

Usually the presentation materials do spell out the OS & hardware the performance numbers refer to, but it's down in the fine print. As a general rule if an OS is not specified you should assume they are talking about Windows.

[bridgman]

So basically, my A10-6800K is obsolete...

I can't use the older Catalyst that supports the HD 8670D inside it because that doesn't work with 4.x kernels, nor can I use this 'new' Catalyst because it dropped support for pre-GCN cards and the HD 8670D is a pre-GCN card ...

Smooth move, AMD ...

Certainly not obsolete since we're continuing to support it via the open source drivers.

That said, my understanding had been that this Catalyst release still included support for VLIW parts.

[dungeon]

O.o...

Ah sorry, you are about mouse corruption... i tought answering to someone else Cannot remember your forum nick i don't even trying

[dungeon]

[*]What do they take as performance basis? The last beta driver? Or maybe the last stable release? Probably the 2nd slides gives a hint, where "Since AMD Catalyst Omega Driver" is written.

I think most comparison on slides are done on Radeon Software 15.11 beta vs Catalyst 15.7.1. According to this downthere:

https://community.amd.com/community/...rimson-edition

Those are internal results, some are done even month ago, some vs different driver some on different machines of course... most recent is Crimson 2 weeks before release...

And people here want to take Michael results as a fact, that does not fly 100% really... even on different sites there are different results

[oleid]

bridgman : Any idea if it will be theoretically possible to drive a GCN1.0 or 1.1 GPU using amdgpu and the proprietary stack if the amdgpu kernel part was modified accordingly? (i.e. by providing the pciids etc). So does the proprietary blob contain code for GCN1.1 and 1.0, or is it really reduced to 1.2?

[MoonMoon]

If you mean "it would be great if AMD's Windows SW marketing folks wouldn't promise performance gains (and deliver AFAIK), but fail to make it clear that they were talking about Windows/DX not Linux/OpenGL" then yes I agree.

Usually the presentation materials do spell out the OS & hardware the performance numbers refer to, but it's down in the fine print. As a general rule if an OS is not specified you should assume they are talking about Windows.

The slide clearly says: Better Linux performance than previous driver.

[dungeon]

The slide clearly says: Better Linux performance than previous driver.

Where previous driver is again mostly 15.7 actually (same as on Windows)... Slide shows improvment in Bioshock, Portal 2, Dota 2 and Total War.

For first 3 games i am sure there is around that difference as in 15.7 driver profiles were not enabled, but now with Crimson all those are there... for Total War, i guess they think about Attila as in Crimson there i saw profile for it, but that game seems yet is to be released.

In the end driver is better, also AMD seems cut prices for holidays - so happy holidays

AMD Cuts All GPU Prices Including R9 Fury X, Fury And Nano ? Down To As Low As $569, $499 And $549 Respectively