AMD Radeon Software Crimson Edition Is A Letdown On Linux

[Passso]

I think Michael himself can do a favor and maybe stop talking about anything AMD for couple days, this sarcasm is so much 5 in a row . He simply likes to talk about nVidia more Even when hardware is expensive and slow for the money like Jetson X1

I agree 5 articles is too much, they could be joined in 2-3.

But do not deviate to high level corruption theory : Michael simply put facts. Whatever you think he only write that NVidia delivers performance for money and watt and provide quality drivers, and AMD does not. This you cannot deny.

[spiff]

Actually, AMD's future never looked any brighter: ...

While I agree that AMD has a lot of interesting efforts and in general has a much better approach from a free software / open standards point of view, they have been promising great HSA results in the "next generation of hardware" every single time since 2010, and have never fully delivered on this. Zen has seen a brain drain in the past half year, and while AMD has been saying "the work is mostly done, we don't need them", I don't see this as a promising development. Why run out on an ambitious process near completion and lose the acclaim that comes with such high profile success?

I really hope they can turn things around, with an open source Linux driver, top of the line Mantle performance (it looks like they have an edge on Nvidia with regards to close-to-the-metal operation, based on the DirectX 12 results), and a competitive Zen CPU line. It would be great to have an alternative to Intel with good Linux support.

But I'm not holding my breath.

[F1esDgSdUTYpm0iy]

Does that mean you can reproduce it with 15.11? Someone complained for year+ on amd forums and now commented how it is fixed for him

https://community.amd.com/thread/169542

Part of the mouse cursor corruption is that it is very hard to reproduce reliably; one example of an application that does it for me is Heroes of the Storm (while run using Wine). However, I can play HOTS fine for an entire day without it happening... or, I can have it happening multiple times in a row upon loading a map. So, I'm going to be somewhat skeptical about that user for whom it is fixed. It might simply not have shown itself yet.

Additionally, I can't even install 15.11 at this time since support for non-stock and/or very recent kernels is sub-par. I'm on Liquorix 4.3-2, not sure 15.11 would allow itself to be installed.

[artivision]

That i don't understand is why vendors don't use only free code for OpenGL, Vulcan, OpenCL. So if you want OpenGL or Vulcan or OpenCL then install MESA. The same on Windows, install Catalyst for D3D and if you want the rest co-install MESA. Is that simple.

[MoonMoon]

Um, so why can I play OpenGL4 games like Civ Beyond Earth on my Radeon 5570 using Mesa 11 and a recent kernel? Right! Because all the extensions needed are supported. Just give it a try, it might already work today

Civilization: Beyond Earth is listed on Steam to run with a GTX260, which is OpenGL 3.3, not 4.

[tomtomme]

I use this driver, i am interested to read about it

If you don't why you read it, i dunno that Articles should be written for people to read, especially for those who use that driver like me.

And i don't agree at all with how all those are written, because it is not true

Let alone all the marketing and compare drivers, nothing much happened here to deserve that... it is normal 15.11 driver release, update like any other, with some stability and perfromance improvments that is.... only what make sense if Michael want to do contra marketing to AMD because they don't send him cards anymore or something like that Otherwise it does not really align to reality

so the metro regression and the flat bioshock bench are no letdown for you?
and the wrong / useless changelog is no letdown for you?
and that they did not give us the qt gui?

if nothing would have been known about crimson I would be ok with this release - but you must agree, that AMDs linux marketing failed with the leaked slides, the glx-gears-joke and a new win-only-gui. And Michael - he went for it, he was excited to see the perf improvements the new ui etc. and nothing happens but regressions. He does not know about the xonotic-sdl problem, does he? Or do you expect him to read every comment?
I really believe him, that he wanted to see an improved driver, but he didn?t. We didn?t. What improvements are there except xonotic? can you tell us? would be good to hear that it is more than thos 2 bug fixes and kernel 4x.

[tomtomme]

Eeeh, what about testing with a more recent kernel, and seeing if that makes a difference?

why should the kernel impact fglrx? you seem to confuse this with the opensource driver where kernel is relevant.

[tomtomme]

I know, but something might use let say one CPU core on that machine in all 3 runs.

One never knows, as this happened on Hawaii only i would first double check what that machine is doing.

and why didn?t this happen with 15.9 fglrx? those are very wild speculations dungeon. I really respected you for most of your contributions in this forum but this time you seem to cling to a straw.

[arunbupathy]

As a researcher I use linux quite a lot for my research and I play a bit of games too. From what I read, fglrx seems to be a huge code base, and I didn't expect it to improve any time soon. If there is a shimmer of hope for AMD, it is in the open source drivers. I have PCs with a HD 5770 and a R9 M265x and the open source drivers have always worked the best for me. Gaming in linux? Well, I'm fine with Windows for now and I'll happily wait for a year or so to see if AMD delivers. But AMD+linux for research? Not so much because OpenCL is too complicated for me. nVidia with their CUDA is a no-go for me simply because I would never use any proprietary software to do research.

Correct me if I am wrong, but the way I see it, IPC gains are all about parallism in many of the common algorithms that we use and the proper way to exploit it would be to use massively parallel cores or GPUs. If you have parallelized all the parts of your code that could be parallelized, then IPC would mean nothing. So Bulldozer having weak IPC is quite understandable by me (Bulldozer is inefficient as hell, but that is not going to melt all of Antartica's ice overnight, whereas, you could nearly double your power bill by leaving your lights on when you don't need them!). Sadly, even after 5 years, I'm still waiting for the programming languages + drivers + compilers required to tap the power of GPUs! I am hoping that C++17 or whatever it is, combined with the ground work AMD has laid for HSA will make things easier for us researchers and also that AMD will deliver soon enough. Or else I will have no other option but to go Intel/Nvidia.

[gamerk2]

will lead to AMD winning over NVIDIA, thanks to async compute support & better double precision performance (at least when comparing GCN vs. MAXWELL) [plus with the consoles using GCN, optimizations in engines for it]

Except GCN has it's own performance issues, and frankly, the arch is getting a bit long in the tooth at this point. AMD is probably hanging on to it too long at this point. Meanwhile, NVIDIA is moving on from Maxwell, and guess what? It's async compute support is getting fixed.

ZEN:
AMD becomes competitive against Intel again, especially in the high-margin server field!

Unlikely. We have a fairly good idea where Zen performance is going to be. AMD claims a 40% IPC improvement, which sounds need until you remember clocks are going to be lower due to the process node and shorter pipeline. So per-core performance is going to be about 20-30% faster, which puts it around Ivy Bridge/Haswell level performance. AMD is still going to be behind Intel in performance, even after Zen.

HSA:
Once APUs get HBM, things will get really interesting...

Not happening, simply because on-die HBM will make APUs too expensive for their price point, and much for the same reason, HBM won't replace DDR as system level memory. APUs with HBM is simply a pipe-dream.

AMD is up 5.41% at end of regular trading and up another 1.28% in extended hours according to Google Finance at time of this reply. Wall Street hungry for good drivers? Makes sense to me.

Beware short terms trends.

AMD is actually up over 30% since I got in at $1.6, simply because the stock will jump as we near the Zen release. Wouldn't be surprised to see them hit a ceiling north of $3, honestly. But at the end of the day, they have over a Billion in dept repayments due in 2018. and at some point, AMD needs to start turning a profit to pay off those debts. $1 Billion in new debt is not a good thing when your companies valuation is hanging around $800 Million. I'd be shocking if AMD, as currently constructed, exists in 2020.