Announcement

Collapse
No announcement yet.

Don't Keep Dreaming For Linux Video Improvements On Intel's Old Platforms

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by e8hffff View Post
    Intel is acting like a two bit company, not a massive player in the silicon market. I don't own that hardware but just fix the damn thing. If you don't have the staff, get more. Ridiculous.
    Do you have any idea how old this hardware is? Believe me, if you have that attitude about Intel then I suggest never buying an AMD product since support will be dropped way way faster.

    Incidentally, this article should point out that you can certainly play back video on the hardware, you just don't get acceleration bells & whistles. I have a nearly 10 year old Core 2 notebook that's even older than these parts and it can playback non-crazy resolution video (including H.264 video) over HDMI just fine. It's not about to do 4K H.265 playback, but expecting that is just stupid.

    Comment


    • #22
      Originally posted by schmidtbag View Post
      Nvidia still maintains drivers for really old hardware. Even their legacy hardware still sometimes gets an update once in a while. AMD is sort of mixed - the mesa drivers still seem to have an interest in supporting R600, but they just don't have the time. However, they're quickly running out of things they need to get done for radeonSI, so perhaps R600 will see some updates again soon.
      My 2010 GTX460(Fermi) TDR-locks* under Windows 10 and doesn't have linux framebuffer support with the property driver (while the Nouveau driver does).

      Originally posted by schmidtbag View Post
      Anyway, pretty stupid Intel has so much funding yet they don't have very long-term driver support. Kind of a Catch-22: by dropping support, it will encourage people to just buy new hardware. But, people aren't going to want to buy new hardware if they know how quickly their support will be dropped.
      Intel is failing on tech, not marketing: A 2010 i7 is clocked roughly 1.5gigahertz below a 2016 i7 and has 10 times less the transistors (600million versus 6billion) but is, at most, 5 times slower compute.
      If they won't drop old hardware support, they'll end up making no sales at all.

      *The driver keeps trying to restart locking the system in an infinite loop.

      Comment


      • #23
        Originally posted by c117152 View Post

        My 2010 GTX460(Fermi) TDR-locks* under Windows 10 and doesn't have linux framebuffer support with the property driver (while the Nouveau driver does).



        Intel is failing on tech, not marketing: A 2010 i7 is clocked roughly 1.5gigahertz below a 2016 i7 and has 10 times less the transistors (600million versus 6billion) but is, at most, 5 times slower compute.


        *The driver keeps trying to restart locking the system in an infinite loop.

        Can you point out a 6 billion transistor Core i7 please? Because they don't exist. A 10 core Broadwell-E has about 3.2 Billion transistors as you can see here: http://hothardware.com/reviews/intel...well-e-arrives

        As for old Core i7's from 2010, the Gulftown parts had over 1.1 billion transistors, so that's only about a factor of 2.9 more transistors in the newest chips. Believe me, if you know what you are doing with using AVX and all the features that Intel uses in its premiere chips, you can do way more than a factor of 2.9 faster than those 2010 parts. And, more importantly, you can do it with massively better performance per watt.

        Funny you should bring up 2010-era i7 parts though since in many ways they are better than AMD's best products in November of 2016.

        Comment


        • #24
          Originally posted by yoshi314 View Post
          i actually have a paperweigh gma500 based tablet at home. i know the support for it won't ever get any better, so it's going to be dismantled one day for spare parts.
          Same here. I was hoping Wayland could help, but turns out even that isn't working well for some reason and nobody can debug it any more.

          Sigh. Too bad there are still little to no good amd64 detachable-keyboard tablets nowadays.

          Comment


          • #25
            Originally posted by chuckula View Post

            Do you have any idea how old this hardware is? Believe me, if you have that attitude about Intel then I suggest never buying an AMD product since support will be dropped way way faster.

            Incidentally, this article should point out that you can certainly play back video on the hardware, you just don't get acceleration bells & whistles. I have a nearly 10 year old Core 2 notebook that's even older than these parts and it can playback non-crazy resolution video (including H.264 video) over HDMI just fine. It's not about to do 4K H.265 playback, but expecting that is just stupid.
            If there's enough Users complaining then Intel has a duty to fix it, or should I say, GET IT RIGHT.

            Comment


            • #26
              Originally posted by yoshi314 View Post
              it's the nature of the hardware, i guess. it becomes obsolete with time.
              Huh? The issue here is it's perfectly good hardware that's capable of decoding x264 1080p video in Windows. The issue is we can't utilize the video decoder in Linux. Intel announced they were planning on supporting it on Linux and then changed their mind. This has nothing to do with the hardware's capabilities.

              Comment


              • #27
                Originally posted by avph View Post
                On g45 h264 accel works perfectly fine (even 1080p) with the vaapi-g45 branch on the condition the the preallocated vram is high enough (default is 32M, it works fine with 256M or more). This can only be set by the BIOS. I think the Intel vaapi devs oversaw this and therefor never mainlined vaapi-g45. I suspect a similar vram issue is present on Ironlake.
                VRAM is unrelated, unfortunately: https://bugs.freedesktop.org/show_bug.cgi?id=47858#c26

                By the way, I doesn't have this Ironlake laptop anymore, but still have laptop with GM45 that used in school with projector. While CPU in this laptop is powerful enough for HD video, fan make a lot of noise due to higher CPU load. VA-API could be very useful here, however, Intel have different opinion on this matter.
                Originally posted by GreatEmerald View Post
                Sigh. Too bad there are still little to no good amd64 detachable-keyboard tablets nowadays.
                Intel Core-M devices not bad in general (typing this on Dell 7140) and some BayTrail devices could get usable next year (like Dell 5830, if WiFi get fixed) if you need something below 10 inches.

                Comment


                • #28
                  Originally posted by chuckula View Post


                  Can you point out a 6 billion transistor Core i7 please? Because they don't exist. A 10 core Broadwell-E has about 3.2 Billion transistors as you can see here: http://hothardware.com/reviews/intel...well-e-arrives

                  As for old Core i7's from 2010, the Gulftown parts had over 1.1 billion transistors, so that's only about a factor of 2.9 more transistors in the newest chips. Believe me, if you know what you are doing with using AVX and all the features that Intel uses in its premiere chips, you can do way more than a factor of 2.9 faster than those 2010 parts. And, more importantly, you can do it with massively better performance per watt.

                  Funny you should bring up 2010-era i7 parts though since in many ways they are better than AMD's best products in November of 2016.
                  I was (obviously?) adjusting for die size, frequencies, lanes, price and node stepping minus gpu cores. If you're itching for the top Xeons it's around 7.2billion right now versus 2.2billion in 2011.

                  Mind you, I'm not counting core-per-core but mm^2-per-mm^2 which is arguably unfair.

                  A more fair example would be to look on Intel's top transistor density chip which is the 30billion Stratix 10 vs. whatever top density they had back then... But that makes my x10 look conservative even if taking the die size into account.

                  And lets be clear here, your 3 times the performance = 3 times the transistors estimate leave Intel looking like they were sitting 7 years twiddling their thumbs while the material engineers did all the work.

                  Power consumption or AMD comparison wise, seeing how I don't care about x86 comparability while you're perfectly fine with AVX, I'd take your "massive" gains and raise you a 5$ no-name Korean audio DSP's TDP. I mean seriously... I've been unrolling my loop since circa '94 so call me crazy, but the way I see it, what can be parallelized or pipelined in wide issuing (rather then threading) sense is gonna end up in dedicated hardware (GPU core at least) anyhow in the x86 world so adding these extra vector instruction looks like just more wasted silicone from where I'm standing.

                  But hey, your mileage may vary.

                  Comment


                  • #29
                    Originally posted by RussianNeuroMancer View Post
                    Intel Core-M devices not bad in general (typing this on Dell 7140) and some BayTrail devices could get usable next year (like Dell 5830, if WiFi get fixed) if you need something below 10 inches.
                    I know the hardware is generally there, but they're just not assembling it in a way that it would have both a dechable keyboard and also stylus support. Last I checked, only ASUS Transformer Mini fits the bill well enough at the moment.

                    Comment


                    • #30
                      Sorry for your laptops !
                      I think it is a bad news for recycling old laptops (and an i3 M330 is always a good plateform nowadays...)...

                      Comment

                      Working...
                      X