Slashdot Mirror


AMD Catalyst Is the Broken Wheel For Linux Gaming

An anonymous reader writes: Tests of the AMD Catalyst driver with the latest AAA Linux games/engines have shown what poor shape the proprietary Radeon driver currently is in for Linux gamers. Phoronix, which traditionally benchmarks with open-source OpenGL games and other long-standing tests, recently has taken specially interest in adapting some newer Steam-based titles for automated benchmarking. With last month's Linux release of Metro Last Light Redux and Metro 2033 Redux, NVIDIA's driver did great while AMD Catalyst was miserable. Catalyst 14.12 delivered extremely low performance and some major bottleneck with the Radeon R9 290 and other GPUs running slower than NVIDIA's midrange hardware. In Unreal Engine 4 Linux tests, the NVIDIA driver again was flawless but the same couldn't be said for AMD. Catalyst 14.12 wouldn't even run the Unreal Engine 4 demos on Linux with their latest generation hardware but only with the HD 6000 series. Tests last month also showed AMD's performance to be crippling for NVIDIA vs. AMD Civilization: Beyond Earth Linux benchmarks with the newest drivers.

28 of 160 comments (clear)

  1. ATI/AMD has had shitty drivers for 20 years by msobkow · · Score: 4, Informative

    ATI's drivers sucked in the '90s. They sucked in the '00s.

    Why, praytell, would we expect them not to suck in the '10s?

    --
    I do not fail; I succeed at finding out what does not work.
    1. Re:ATI/AMD has had shitty drivers for 20 years by red_dragon · · Score: 3, Interesting

      Because they do have a tendency to improve. Jerry Pournelle used to write regularly about his problems with ATI cards in his column on BYTE. They typically followed the same pattern: install new card; install drivers; see computer crash regularly; upgrade drivers; see computer crash less often; upgrade drivers again; see computer run more or less stably.

      Then he'd upgrade to the next shiny ATI card and do it all over again, since the new drivers bore little resemblance to the old ones.

      --
      In Soviet Russia, Jesus asks: "What Would You Do?"
    2. Re:ATI/AMD has had shitty drivers for 20 years by HBI · · Score: 2

      ATI's hardware was fine hardware back in the day, but their software was regrettable. My first card was an ATI EGA (!) Wonder 800 back just before third-party VGA hit the streets. Then I had a VGA Wonder. Both were fine cards. The 8514/A compatible ATI card was also delicious. Then things headed downhill.

      I see no reason to buy their equipment anymore.

      --
      HBI's Law: Frequency of calling others Nazis is directly correlated with the likelihood of the accuser being Communist.
    3. Re:ATI/AMD has had shitty drivers for 20 years by brxndxn · · Score: 5, Informative

      AMD is most certainly not a shit company ~ you have a shit opinion. AMD was pounded into the ground financially by Intel competing unfairly when AMD had the clear performance advantage. Intel made their agreements so companies like HP would actually save money if they went 100% Intel even though the market was clearly calling for AMD processors. This was made obvious when AMD offered to give free processors to HP and HP still refused. Since Intel is basically a monopoly and our regulatory agencies are run by ball-less cowards, AMD has a tiny research and development budget compared to both Intel and Nvidia. AMD continues to exist with research money being their only real limitation. History shows that AMD can create processors that outperform Intel with less than a quarter of Intel's research budget. AMD has nowhere near a quarter of Intel's research budget at the moment. As far as ATI/Nvidia competition, Nvidia tends to make things as proprietary as possible while AMD makes them more open. AMD's graphics hardware tends to be more advanced while having a simpler design than Nvidia. However, AMD's very limited software/driver development budget keeps AMD graphics cards performing less than optimally. Further, my experience of gaming on Linux a few months ago gave me no issues with the Metro 2033 Redux and my Radeon 7970.. so I am going to take this 'benchmark' with a grain of salt. Perhaps the latest driver is slow or has a bug - but they do tend to get fixed from my experience. Even if you hate AMD as a company, you can thank them for the reasonable prices of CPUs and graphics cards. Without AMD, both become extremely expensive.

      --
      --- We need more Ron Paul!
    4. Re:ATI/AMD has had shitty drivers for 20 years by ihtoit · · Score: 3, Interesting

      my first offboard GPU was a 4MB Rage Pro - which I've still got, have since upgraded it to 8MB with the simple addition of a SODIMM. It read 16MB as 8MB and ran OK if slow, it read 8MB and ran as fast as with the base 4MB but was unstable as hell, so I did some hunting and found a 4 and used that.

      --
      Political debates have me rolling my eyes so much I think I got optical whiplash. I should sue. - Foamy The Squirrel
    5. Re:ATI/AMD has had shitty drivers for 20 years by TFlan91 · · Score: 2

      byte.com -> http://www.informationweek.com... -> 404...

      Such a shame

    6. Re:ATI/AMD has had shitty drivers for 20 years by Wootery · · Score: 2

      As far as ATI/Nvidia competition, Nvidia tends to make things as proprietary as possible while AMD makes them more open.

      Correct: CUDA, SLI, Physx, G-Sync, Shield Portable/GRID. AMD tend to either hop on board the open follow-up technologies (OpenCL, OpenGL compute shaders, DirectCompute) or create their own technologies which tend to be more open (FreeSync, Crossfire).

      Mantle is an interesting one: it's open, but Nvidia aren't interested. This seems kinda reasonable: Mantle seems to have served well as a wake-up call that it's possible to create more efficient graphics APIs, so we can expect the next-gen OpenGL and Direct3D APIs to catch up to Mantle and far reduce its relative merits.

      Whether this trend of openness with AMD is because they're just nice like that, or because they're generally playing catch-up... I guess Mantle would imply they do it 'voluntarily', but again, it was never going to have the sort of lock-in effect that CUDA does.

    7. Re:ATI/AMD has had shitty drivers for 20 years by drinkypoo · · Score: 2

      Then he'd upgrade to the next shiny ATI card and do it all over again, since the new drivers bore little resemblance to the old ones.

      Two points. 1, they now bear striking resemblance to the old ones, they now are the old ones with support for new cards. That's how modern video drivers work. 2, they bore striking resemblance to the old drivers then, too, as they were shit release after release.

      I've been watching ATI drivers crash Windows since Windows 3.1 and the Mach32. Others have seen it even longer. Why anyone ever gives them money is beyond me. The last time I did it, I took a chance on some integrated graphics based on an old core and they never did work right and caused me endless headaches. Now I just don't run anything that uses 3D graphics on that machine, and I use it for OBD-II diagnosis. Since it's running windows, I had to upgrade it to a dual-core processor just to have it do that decently, but that's a different rant.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    8. Re:ATI/AMD has had shitty drivers for 20 years by Tough+Love · · Score: 2

      Sounds like your granddaddy's ATI hardware...

      $ uptime
        15:55:56 up 171 days, 2:22, 25 users, load average: 0.76, 0.98, 1.26

      $ lspci | grep AMD
      01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Park [Mobility Radeon HD 5430]
      01:00.1 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] Cedar HDMI Audio [Radeon HD 5400/6300 Series]

      Zero glitches, hangs or weirdness in the last several years with this or my other Radeon cards. Using the Xorg drivers. Includes heavy OpenGL hacking and way too much Civ V.

      --
      When all you have is a hammer, every problem starts to look like a thumb.
    9. Re:ATI/AMD has had shitty drivers for 20 years by Tough+Love · · Score: 2

      The last bout I had with NVidia was really annoying. Switch to text console and, whee, nice black screen. Yum. Got to love that Quadro experience. That was just one of many severe glitches. Haven't run that binary NVidia crap for many years now, very happy with the open source Radeon drivers and nice hardware. I really need to wonder if/why you're trolling. Got anything to add based on your last ten years experience? Didn't think so.

      Maybe one day Nouveau will catch up to where the Xorg Radeon drivers got to years ago. Wake me up when that happens, maybe then I will take another flyer on NVidia.

      Just say no to crap binary video driver blobs, I don't care what hardware or bespoke OS you are talking about, proprietary crap is proprietary crap.

      --
      When all you have is a hammer, every problem starts to look like a thumb.
    10. Re:ATI/AMD has had shitty drivers for 20 years by Lonewolf666 · · Score: 2

      Well, AMD certainly has their own less than stellar moves too. Ever since AMD bought ATI in 2006 they've been talking about synergies but to be honest, I'm not seeing it. An "APU" performs very, very similar to the same CPU+GPU if you compare cores on the CPU side and shaders on the GPU side.

      Depends on which kind of system we're talking about.

      On low-end APUs, the concept works fine and not needing a discrete GPU is a nice cost advantage. But Intel's HD graphics is already becoming a serious competitor in that product range.

      At the top end of the (desktop) APU spectrum, the APUs tend to become bottlenecked by memory and a similar combination of cores on the CPU side and shaders on the GPU side tends to win the benchmarks. The cost advantage of the APUs still makes them interesting, but check out offers with discrete GPUs too and read some reviews.

      What could help AMD here is HBM as VRAM in future APUs, that would remove the memory bottleneck...

      --
      C - the footgun of programming languages
  2. Who want to play games on Linux? by Anonymous Coward · · Score: 4, Interesting

    Uhh, the average age of gamers is above 30 http://www.statista.com/statistics/189582/age-of-us-video-game-players-since-2010/

  3. Re:Who want to play games on Linux? by jedidiah · · Score: 4, Insightful

    A number of my colleagues would dump Windows in a heartbeat if they could run their PC games on some other OS.

    --
    A Pirate and a Puritan look the same on a balance sheet.
  4. Breaking old cards by phorm · · Score: 4, Insightful

    One of the reasons why I will probably not buy an ATI/AMD (for graphics), is that support for older hardware is pretty terrible. I have an Asus laptop which worked *beautifully* in both Windows/Linux.
    Apparently, some people (not me) had issues with brightness control not working on the fglrx driver. AMD fixes that, and on my laptop (and others, according to Google) the backlight breaks. As soon as X initializes my backlight goes dark. In a bright room I can barely see that X otherwise started successfully and is displaying a login window.. It's been over a year. I've seen lots of chatter on fixes for the brightness-control button, but pretty much zip about the broken backlight.

    I can use the Radeon driver so that X will work, but video is choppy and since I'm working on actually developing GL code, it's pretty much useless for that. So... core i7 processor, lots of RAM, decently powerful GPU, and a farked video driver that renders the whole thing useless.

    I had actually been migrating more towards AMD from nVidia since their graphics drivers had shown promise since ATI was acquired, but frankly the nightmare of bug-support is pushing me back towards nVidia. It especially sucks for a laptop since I can't exactly replace the GPU on what it otherwise fully functional hardware.

    Currently I'm picking at firegl_public.c and related modules attempting to merge the 13.25 driver with the 8.960 driver (I've been told that reverting to the older driver will allow the backlight to work, but in my case it won't compile under DKMS).

    To any AMD Linux driver devs listening: I would be happy to work with you on this. Hell, I can ship you the damn laptop for a few months if you believe that would help develop a driver that works again.

    1. Re:Breaking old cards by dkman · · Score: 3, Interesting

      My 4 year old Sager laptop has a GPU module (or slot, so it's essentially like RAM and can be replaced). When the graphic card decided to flip out a few weeks ago I searched around and they were around $230 shipping from China. Even though everything said to me that the problem was just the video card I decided to spend $1600 on a new Sager laptop. Since the old one was now disposable I decided to do the "bake the video card" trick (375 degrees for 10 minutes, in case you're interested - just remove all screws, heat sinks, and thermal paste). I let it cool, applied thermal paste, and gave it a shot - bam, worked like new. Since the new laptop was already in "processing" I decided to let it come anyhow.

      The old one is an ATI (HD 6990M). It handles linux gaming alright, it really depends on the game. Windows gaming it's great at - I just don't boot window often. The new laptop has an nvidia because I do feel that the nvidia drivers will be better in linux. Over the past 20 years I've given both companies some love.

      --
      I refuse to sign
    2. Re:Breaking old cards by spire3661 · · Score: 2

      Windows 8 can run well on a 16 GB partition and 1 GB of ram with a 1.3 GHZ Bay Trail Atom CPU.....

      --
      Good-bye
  5. Waffle much? by mandark1967 · · Score: 4, Interesting

    From 2 weeks ago:

    "...the latest Phoronix end-of-year tests show the AMD Catalyst Linux driver is beating Catalyst on Windows for some OpenGL benchmarks. The proprietary driver tests were done with the new Catalyst "OMEGA" driver. Is AMD beginning to lead real Linux driver innovations or is OpenGL on Windows just struggling?"

    (http://linux.slashdot.org/story/15/01/03/1426208/amd-catalyst-linux-driver-catching-up-to-and-beating-windows?sdsrc=rel)

    --
    Sig Follows: "Suppose you were an idiot. And suppose you were a member of Congress. But I repeat myself." -- Mark Twain
    1. Re:Waffle much? by Creepy · · Score: 2

      It shows the transition - OpenGL has been a second class citizen for ATI/AMD for years and now is first class. With the PS4 using their chip and OpenGL they finally have a good reason to have extremely good OpenGL performance and dedicate development time to it, but I still remember when they wouldn't even support extensions.

      nVidia has always had OpenGL as a first class citizen (for different reasons - CAD, then PS3, now mobile and Linux GRID arrays such as the one I use with VMWare VMs for GPU support).

    2. Re:Waffle much? by emblemparade · · Score: 2

      Right: OpenGL on AMD/Linux is catching up with OpenGL on AMD/Windows.

      But both suck relative to NVIDIA's implementation of OpenGL. AMD and NVIDIA are competitive on D3D, but that's no comfort to Linux users, where they simply cannot use their GPUs to their fullest potential.

  6. Re:Who want to play games on Linux? by CaptainOfSpray · · Score: 3, Informative

    700+ games ifor Linux in the Steam store. Clearly lots of people want to play games on Linux.

    --
    "Cock Up Your Beaver" does not mean what you think. This sig is intended to clog filters and annoy do-gooders
  7. But the Open Source drivers are good by Atmchicago · · Score: 4, Informative

    My Radeon 6850 runs TF2 great on the OSS drivers. This is where things are headed, and if AMD keeps it up then Nvidia will have catching up to do. We're nearing the point where you can buy a graphics card, plug it in, and it "just works." The main issue is that it could take months for the bleeding edge to make it into the latest kernel, so brand new GPUs could be problematic. More to the point, in a few years an AMD APU might be both "good enough" for gaming, and also "just work." On Linux. That's saying something.

    --

    You can lead a horse to water, but you can't make it dissolve.

    1. Re:But the Open Source drivers are good by AqD · · Score: 2

      Being able to utilize only 10% of your hardware is NOT good - there is a zero missing!

      It sucks and you know it and that's why you're testing TF2 rather than real games on it. By real games I mean it should at least heat your GFX card to 80C and fans running in maximum speed.

    2. Re:But the Open Source drivers are good by Tough+Love · · Score: 3, Informative

      We're nearing the point where you can buy a graphics card, plug it in, and it "just works." The main issue is that it could take months for the bleeding edge to make it into the latest kernel, so brand new GPUs could be problematic.

      +1. All my latest Radeon installs have been: buy it, plug it in, it works. That is because I always check this matrix before deciding which card to buy. Note that it is now green all the way to the right on nearly every hardware feature you care about. A notable exception is OpenCL, which is WIP. If I wanted to develop with it right now, that would move me back to fglrx, and only that. Notice how the latest chipsets are the ones with OpenCL closest to prime time.

      --
      When all you have is a hammer, every problem starts to look like a thumb.
  8. Re:Who want to play games on Linux? by Immerman · · Score: 2

    Right - all those gamers should be either working themselves to death trying to get ahead in a game that's rigged against them, or out getting drunk and shagging strangers. You know, socially acceptable ways to pass the time.

    --
    --- Most topics have many sides worth arguing, allow me to take one opposite you.
  9. Jaguar: Do the math by tepples · · Score: 2

    PlayStation 4 and Xbox One are essentially PCs built around AMD's Jaguar laptop APU, except with a locked-down BIOS instead of standard UEFI. Do the math.

  10. Their drivers might be garbage, the silicon's OK by jhantin · · Score: 4, Interesting

    AMD got the $6 billion to buy ATI by spending the cash reserves they had to build their next generation fab. The result is that after they bought ATI they had to sell their manufacturing operations sliding even further into irrelevance as their costs are much higher than Intel.

    It's not like they don't actually have a sensible plan, though. While they might not be able to catch Intel in the short run on high-end CPUs, some of their newer APUs (some of them outright SoCs) are surprisingly efficient little beasts built for the low-power market segment: silent or fanless mini PCs, tablets, ultraportables, and an assortment of bespoke embedded gadgets. While the CPU side trails Intel's, on-die GCN soundly demolishes any integrated graphics Intel puts out there.

    --
    ...when you're writing a game...tweak the difficulty of "Easy" to something [your mother] can cope with. -- onion2k
  11. Re:Why do people still go AMD? by moogaloonie · · Score: 3, Interesting

    Brand loyalty I suppose... I chose AMD over Intel for CPU when putting my first PC together because I didn't want to support certain business practices of Intel. When AMD bought ATI, I started using AMD GPUs instead of nVidia (thinking they would work together more smoothly being from the same company). My previous system was a 4-core with Radeon HD, my current system is 8-core with R285. I have been quite happy with all of the AMD components I have used, and see no compelling reason to switch. I'm not a hardcore gamer, I mainly play Just Cause 2, Spin Tires and a few others... But 3D performance is still quite important to me for content creation.

  12. Re:Who want to play games on Linux? by Anonymous Coward · · Score: 2, Informative

    The Steam Hardware Survey shows Linux usage at 1.16%. Clearly, few people actually play games on Linux. Windows 8 on the other hand, the OS that trolls claim no one wants to use, is at 31.29% and climbing.