The Open-Source / Linux Letdowns Of 2016

[phoronix]

Phoronix: The Open-Source / Linux Letdowns Of 2016

Last year I had written about the The Open-Source Linux Letdowns of 2015 and then Other Letdowns For Linux / Open-Source Users From 2015, which ended up being among the most viewed articles of 2016. So I figured I'd once again share a list of what personally was disappointing not to see happen in 2016 within the Linux/open-source space...

http://www.phoronix.com/scan.php?page=news_item&px=2016-Letdowns-For-Linux

[Krejzi]

xf86-video-intel-3.0 ... Man, I would've been happy with any release and not having to rely on git snapshots. Last (development) release was in Dec. 2014. Disappointing.

[Beherit]

An updated Linux-emulation layer for FreeBSD would be nice, to be able to run Linux Steam games.

[doublez13]

No Andromeda

[ldo17]

As long as there is no serious competition to Nvidia in the GPGPU space, OpenCL will continue to languish.

Why is there no serious competition? Could it be that the GPGPU market simply isn’t big enough?

[DMJC]

Biggest disappointments for 2016: XHCI Driver instability when running Linux on a Macbook Pro 13" 2015, Thunderbolt/Displayport autodetection failures. Improper wake from sleep. It's 2017 guys, please get this stuff working. Lots of Linux game engine projects still shipping without decent editing tools. (I take back everything I've said about Warhammer Dark Omen. I love the WINE project and you guys have finally fixed it!!!!)

[defaultUser]

As long as there is no serious competition to Nvidia in the GPGPU space, OpenCL will continue to languish.

Why is there no serious competition? Could it be that the GPGPU market simply isn’t big enough?

The market is big and most important is one of higher margins. Way better than consumer oriented GPU's. However AMD apparently was never committed enough to challenge the Nvidia in this segment. Actually even Intel is having problems (for multiple reasons) to make a dent on this segment. Nvidia created a infrastructure around CUDA not only in terms of continually improving the CUDA API, but also creating technical/research centers on universities and funding researchers to port codes to GPU or develop algorithms more suitable to this architecture . Most importantly they managed to offer real increase of performance for each of their main architecture revisions. Even if this meant to reduce performance in one segment (for instance nvidia offers gpu with improved single precision performance for machine learning or tuned to double precision for HPC). Said that for desktop/workstations applications people still trying to use opencl as ways to avoid consumer lock in. For instance if Adobe used cuda. Their applications would not run on Mac

[ldo17]

The market is big and most important is one of higher margins

It is axiomatic that, if there was lots of money to be made, then competitors would enter the market.

So what is keeping them out?

[DMJC]

Cost of entry, GPUs are something which only 3-4 companies in the world have gotten right, and making them into GPGPUs is something which only one company has really done well. There's probably a bunch of patents in that mix along with just good old secret sauce.

[mitcoes]

What is worse, in my opinion, is the lack of AAA Vulkan games, plus the lack of a good "around 300 USD "cheap console" Steam Machine at big stores" (perhaps a 720p by default one). May be this new AMD ZENs, if they make SoCs as they did with the A10, but more powerful, or some intel radeon affordable combo and good enough, at least for playing AAA titles at 720p will make it happen in 2018. And please 64bit games, how is that there are no (or not a lot) 64bit games in 2017?