Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
AMD Businesses Open Source Linux

AMD Closes OSRC, Lays Off Several Linux Kernel Developers 94

Posted by Unknown Lamer from the reversal-of-fortunes dept.
From the H reporting on LinuxCon Europe comes news that several Linux kernel developers have been laid off by AMD as part of its workforce reduction. From the article: "OSRC staff primarily worked to develop the Linux support for AMD's server processors, but they also wrote code and extensions for related desktop and notebook CPUs – for example, they looked after the code to support CPU frequency scaling for the PowerNow and Turbo Core technologies. While working on the kernel's IOMMU and KVM support, one of AMD's former employees contributed to the development of the "IOMMU groups" feature that was integrated into Linux 3.6; this feature provides the basis for a new Linux 3.6 technology that allows a host's PCIe devices to be passed through to virtual machines and can also be used with Intel CPUs." Looks like the group was doing interesting research on hypervisors, lockless data structures, and multi-core synchronization primitives among other things. The Open Source Radeon driver developers are not affected by this at least.
This discussion has been archived. No new comments can be posted.

AMD Closes OSRC, Lays Off Several Linux Kernel Developers More

AMD Closes OSRC, Lays Off Several Linux Kernel Developers

Comments Filter:

  • I guess they don't want me to buy their products (Score:5, Insightful)

    by h4rr4r ( 612664 ) on Wednesday November 07, 2012 @12:08PM (#41909133)

    The AMD/ATI linux drivers suck, they are laying off their kernel folks, and no indication they have any plans to change. I hope they survive, but convincing me not to buy your products is not going to help.

    • AMD is betting the farm on ARM-64. If it fails to take off in the server world, there won't be anything left of the company. Too many cuts and too deep. The worst part of that is that not only would we lose competition in the x86 space, but graphics competition at the high end would also be gone (unless Intel starts working miracles).

      • Re: (Score:1)

        by Anonymous Coward

        AMD is betting the farm on ARM-64. If it fails to take off in the server world, there won't be anything left of the company. Too many cuts and too deep. The worst part of that is that not only would we lose competition in the x86 space, but graphics competition at the high end would also be gone (unless Intel starts working miracles).

        Is there a good reason Intel doesn't get more serious about graphics hardware? With the fabs and expertise and funding they've got I am convinced they could do it if they wanted to. Why wouldn't they as a company want to expand into this market? Wouldn't it be a way to diversify?

        And I mean, the drivers on Linux for intel graphics Just Plain Work. Even more hassle-free than nVidia's linux drivers which are quite good, just of course not bundled with the kernel like almost every other driver.

        And no

        • Re: (Score:3)

          by snadrus ( 930168 )
          I suspect Intel doesn't want to give any more help to GPU processing efforts which are making inroads at obsoleting their main CPU line for large workloads.

          • Re: (Score:1)

            by Anonymous Coward

            More to the point, they have no need to produce better graphics. People who play more than casual games buy a discrete GPU, so they only need to support desktop displays and basic games; about a third of the die in my i7 system is wasted on an IGP I don't use, and I really don't want that situation to become worse in future.

            • I agree that IGP can be a waste, but why should Intel limit themselves to integrated graphics? AMD seems to have done fine with both discrete and integrated GPUs.

        • Re: (Score:1, Offtopic)

          by lister king of smeg ( 2481612 )

          i have been having a pulse audio problem on my linux box any suggestions for a replacement just use alsa or anything else?

          • i have been having a pulse audio problem on my linux box any suggestions for a replacement just use alsa or anything else?

            I personally just use ALSA for everything.

            I use Gentoo so this system has never had PulseAudio installed (the way Gentoo works, I would only get Pulse by putting it there myself, which I won't).

            Ubuntu and most major distributions have wiki pages concerning PulseAudio and how to remove it. Most of the time it's as simple as running a command or two involving your package manager. Binary distros tend to build programs with all features enabled and they simply won't use functionality you don't actually

            • Re: (Score:2)

              by garaged ( 579941 )

              Honest question, how do you get multiple apps doing sound without pulseaudio?

              Im pretty hardcore linux user, and I dont know a better option then pulse audio, and it does not crashes on me, it is a little PITA on first install, but after that it works great for me, maybe I am lucky? ( currently using it on a dell laptop, asus lapto, and amd desktop)

              • You have always been able to get multiple apps doing sound. You've just not been able to individually control the sound levels before. OSSv4 allows you to do this too.

                • Re: (Score:2)

                  by garaged ( 579941 )

                  For real, that is totally not my experience, just one app will work, any other simultaneous app trying to send sound would lose the ability to until the working one is killed and sometimes a restart of both apps is required

                  This must be related to sound card model I guess, but I have never been able to do multiple apps with plain OSS or ALSA

        • Re: (Score:3)

          by Kjella ( 173770 )

          Is there a good reason Intel doesn't get more serious about graphics hardware?

          Well, I'd say that the biggest reason they wouldn't is because then they'd have a problem "explaining" why they use up your CPU die for an IGP. Intel wants your GPU to be something that comes with your CPU, because that's obviously a huge advantage for Intel. That's why they've made real effort to improve Sandy Bridge, Ivy Bridge and Haswell promises to take this even further. Reportedly their fastest IGP configuration GT3 is supposed to have 40 EUs compared to 16 in IVB, obviously it's unreleased yet but I

      • Re: (Score:2)

        by Sir_Sri ( 199544 )

        , there won't be anything left of the company.

        There's nothing left of the company now, and what is left has the vultures, notably qualcomm, feasting on its remains. The main potentially profitable enterprise from the old AMD was globalfoundries, and that's a separate company now.

        AMD is basically just a chip design firm now, they sold their ARM business to qualcomm a couple of years ago (the snapdragon and adreno products), and they are going to struggle to catch back up to the big ARM guys at this point, and the GPU products business will probably be

      • Competition in terms of price? Possibly. But competition to push the envelope in terms of performance? I thought game makers pushed hardware makers to innovate?

    • Re:I guess they don't want me to buy their product (Score:5, Informative)

      by Baloroth ( 2370816 ) on Wednesday November 07, 2012 @12:17PM (#41909203)

      This has nothing to do with graphics drivers at all, those were completely unaffected. It might impact implementation of some new server features on Linux, but it is strictly about CPU and related features, not APU or GPU stuff.

      • Re: (Score:2)

        by h4rr4r ( 612664 )

        I realize that. I only meant they were already crap, so this was basically their only way of making those products less attractive to me.

    • Re: (Score:2)

      by zixxt ( 1547061 )

      The AMD/ATI linux drivers suck, they are laying off their kernel folks, and no indication they have any plans to change. I hope they survive, but convincing me not to buy your products is not going to help.

      AMD is the only choice in my mind. Intel is big and evil monopolist who hates consumers , and Nvidia hates Linux and OpenSource.

      AMD is the only alternative to these goons in the PC market.

  • Not sure I understand... (Score:5, Insightful)

    by Nemyst ( 1383049 ) on Wednesday November 07, 2012 @12:10PM (#41909157) Homepage

    The server market, usually Linux-based, appears to be AMD's most stable market. Opterons are very often preferred over Xeons for a variety of reasons. So why exactly would AMD start axing developers in areas related to that? If anything, it'd make more sense to throttle down consumer processors and focus on graphics and server processors, no?

    • Re: (Score:2, Insightful)

      by Anonymous Coward

      I guess the problem is that the work of these guys affected real server performance in contrast to generic benchmarks. What they achieve is real life performance improvements in real applications running on Linux. But generic benchmarks rarely capture these advances, so their work has limited marketing value.

      • Re: (Score:2, Insightful)

        by Anonymous Coward

        I'll agree with this. Under Linux, much of the performance edge for Intel chips over AMD chips disappear. But what websites have these benchmarks? Phoronix has some simple ones but nothing that relevant. So we end up with review after review of Windows + Games even though AMD server optimizations in Linux would make it a compelling purchase option IF PEOPLE FREAKING KNEW ABOUT IT!

        • Re: (Score:1)

          by Anonymous Coward

          Do wonder how much this disappearance has to do with the compiler used.

          That is allegations floating around that the Intel compiler puts checks in the binaries that will only turn on MMX and other advanced bits when running on a Genuine Intel. On any other hardware it drops back to 386 generation code.

          On Linux the most common compiler is GCC however, something that is more likely to be CPU agnostic.

          • Re: (Score:1)

            by Desler ( 1608317 )

            Sure, if all that Windows software was compiled with Intel's compoler. That is highly unlikely.

            • Re: (Score:2)

              by 0123456 ( 636235 )

              Sure, if all that Windows software was compiled with Intel's compoler. That is highly unlikely.

              When I was writing high-performance Windows software it was compiled with the Intel compiler. Since Intel CPUs were 99% of our market, not optimising for them would have been silly.

              Your word processor probably doesn't use the Intel compiler, but when was the last time you complained that your word processor's code wasn't optimised enough?

              • Re: (Score:1)

                by Desler ( 1608317 )

                When I was writing high-performance Windows software it was compiled with the Intel compiler

                Good for you? Plenty of other people use MSVC, GCC, and various other compilers.

                Since Intel CPUs were 99% of our market, not optimising for them would have been silly.

                Sure, if no other compiler optimized their code generation for Intel
                CPUs. This isn't the case, though. Also ICL's benefit on the average case in code generation over MSVC or GCC is marginal.

                Your word processor probably doesn't use the Intel compiler, but when was the last time you complained that your word processor's code wasn't optimised enough?

                Yes, and neither are most applications. MSVC is quite dominant for softwate on Windows.

          • Re:Not sure I understand... (Score:5, Informative)

            by Rockoon ( 1252108 ) on Wednesday November 07, 2012 @01:12PM (#41909849)

            That is allegations floating around that the Intel compiler puts checks in the binaries that will only turn on MMX and other advanced bits when running on a Genuine Intel

            Its not just allegation...

            Due the the FTC ruling against Intel, its still not too late to get reimbursed by Intel if you purchased Intels compiler. [compilerre...rogram.com] Only a few more months left on that.

            • I guess getting reimbursed by Intel if you purchased AMD's processor is out of the question... :p

    • Re: (Score:1)

      by Anonymous Coward

      The company is being run by people who don't understand the technology. I expect a lot of question marks in the coming months.

    • Re:Not sure I understand... (Score:4, Funny)

      by dmbasso ( 1052166 ) on Wednesday November 07, 2012 @12:36PM (#41909441)

      No, they see the obvious: Windows 8 will dominate the supercomputer OS market [Unicode: U+2E2E]

    • Re: (Score:1)

      by Anonymous Coward

      You're exactly right, so the only available conclusion is that they're getting out of the x86 market. See also their recent ARM deal.

      It's too early and too risky, IMO. The might know something I don't know, but I suspect they're just being foolish.

      • Getting out of the x86 market would be the wrong move they are after all responsible for the AMD64 architecture. If I where them I would look into a multi-architecture computing. Have a board support running on ARM for low power and fire up the x86_64 core when you need more power. Possibly put both cores in the same possessor similar to the APU design where the CPU and GPU are built one on top of the other. Or maybe I'm just crazy. I admittedly know little about the processor design if someone would be so

        • Re: (Score:2)

          by Jeng ( 926980 )

          If I where them I would look into a multi-architecture computing. Have a board support running on ARM for low power and fire up the x86_64 core when you need more power.

          They are doing multi-architecture computing.

          http://en.wikipedia.org/wiki/AMD_Fusion [wikipedia.org]

          How much of a benefit do you think they would receive from adding an ARM core to the mix?

    • The current trend is: focus on tablets. Discard anything else as unimportant.

    • Re: (Score:1)

      by Desler ( 1608317 )

      AMD's marketshare in server is less than 6% and that was after the 1.5% estimated drop back in March. Opterons "very often" preferred over Xeons? Sorry, but the actual data doesn't bear out this claim.

      • You can't pack 64 cores in a 1U server for ~$5000 (with moderate RAM and HDDs) with Intel.

        If you need massively parallel, high density, x86_64 compatible servers then AMD is king.

    • Re:Not sure I understand... (Score:4, Interesting)

      by Kjella ( 173770 ) on Wednesday November 07, 2012 @01:01PM (#41909731) Homepage

      The server market, usually Linux-based, appears to be AMD's most stable market. Opterons are very often preferred over Xeons for a variety of reasons.

      If all you read is slashdot, yes. Actually AMD's server market share has been in big decline from 15% in 2007 to below 5% today. They're quite well represented in the TOP500 as apparently they give good quantity discounts for the PR, but mainstream companies are >95% Intel now. And with AMD now betting on ARM servers that's certainly not going to improve in the short term.

    • So why exactly would AMD start axing developers in areas related to that?

      Because maybe it's one step away from closing the front doors? AMD has their financial ass in a financial sling. I hope they pull through it but a cut like this is a bad sign.

    • No, because since the economy failed companies are getting by on 2-3+ year old servers. Consumer computer purchases are still going strong, though.

      When you have a datacenter full of mostly idle servers, it makes no sense to upgrade them all. Especially because virtualization now allows one to consolidate 20+ mostly idle machines into a single physical computer.

    • The server market, usually Linux-based, appears to be AMD's most stable market.

      Servers are a market that AMD has been losing for years now.

      AMD stole over 20% market share from Intel while they were fucking around with the Pentium 4 disaster, and they've been losing it ever since the release of Conroe. Nehalem with high-scaling QPI interconnect was the last nail in the coffin - AMD's server share has been in free-fall since then.

      2010: 7% server market share! [seekingalpha.com]

      This year: 5.5% server market share [fudzilla.com]

      Like the first

  • Too bad for AMD and all of us (Score:3, Insightful)

    by Anonymous Coward on Wednesday November 07, 2012 @12:12PM (#41909167)

    Wish they'd start at the top... just look for the fuckers preparing their golden parachutes and sack them before they can deploy.

    • Like a smoker shooting themselves in the head for getting lung cancer? After all, it was the brain that made the stupid decisions... :p

  • Time to look at alternatives...

    I wish I knew this before buying a quad core A6 laptop, their graphics card driver is abysmal compared to nVidia, which is increasing the performance of their drivers even more.

    AMD is doing a very risky bet, the ARM64 has not been well tested in real server workloads, it may as well flop and take the company with it, their opteron line was excellent for virtualization thanks to the many cores, hardware virtualization support and a good performance/price point this new ARM64 ha

  • How does one go about laying off a hobbyist?

Slashdot Top Deals

If all the world's economists were laid end to end, we wouldn't reach a conclusion. -- William Baumol

Close