Announcement

Collapse
No announcement yet.

AMD Announces Ryzen Embedded R2000 Series With Zen+ Cores, Radeon Graphics

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

  • AMD Announces Ryzen Embedded R2000 Series With Zen+ Cores, Radeon Graphics

    Phoronix: AMD Announces Ryzen Embedded R2000 Series With Zen+ Cores, Radeon Graphics

    AMD is using the Embedded World conference in Nürnberg to launch the Ryzen Embedded R2000 series for industrial use-cases along with IoT, thin clients, and edge computing...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    They switched from GF 14 nm to 12 nm but no real improvement on efficiency. The 2C4T variant looks to be exactly the same CPU and GPU as R1606G, just with 100 MHz more. Same TDP, this seems strange.

    Probably only cheaper production vor AMD but who wants > 15W embedded CPUs?
    Last edited by Anux; 21 June 2022, 09:34 AM.

    Comment


    • #3
      Originally posted by Anux View Post
      They switched from GF 14 nm to 12 nm but no real improvement on efficiency. The 2C4T variant looks to be the exactly the same CPU and GPU as R1606G, just with 100 MHz more. Same TDP, this seems strange.

      Probably only cheaper production vor AMD but who wants > 15W embedded CPUs?
      There's no reason for them to change the TDPs they were already targeting. There is probably a different embedded lineup somewhere with ~6W parts.

      Not that it's difficult to passively cool a 15W chip TDP-downed to 12W, in a bulky slot machine or whatever they are going into.

      From CNX:

      IBASE plans to launch SI-324-N2314 fanless digital signage player and INA1600 desktop uCPE/SD-WAN appliance
      Last edited by jaxa; 21 June 2022, 09:22 AM.

      Comment


      • #4
        Originally posted by jaxa View Post
        There's no reason for them to change the TDPs they were already targeting. There is probably a different embedded lineup somewhere with ~6W parts.
        They did change the TDP from 6 to 15 W in the 1000 series to 15 - 35 W. So even if you use cTDP at 12 W it still doubled the consumption of the old lineup.

        Comment


        • #5
          I wonder if they can backport zen 3 to a 12 nm quad core embedded solution. It shouldn't be that big or power hungry and the lower clocks won't matter, if it can be made on the cheap. Maybe even backport a more adequate igpu solution while at it. That would be a hit in the developing world - anything from decent computers to industry equipment.
          Last edited by ddriver; 21 June 2022, 11:08 AM.

          Comment


          • #6
            Originally posted by ddriver View Post
            I wonder if they can backport zen 3 to a 12 nm quad core embedded solution. It shouldn't be that big or power hungry and the lower clocks won't matter, if it can be made on the cheap. Maybe even backport a more adequate igpu solution while at it. That would be a hit in the developing world - anything from decent computers to industry equipment.
            This appeared in rumors and hasn't been heard about since.

            Zen 2/3 + RDNA2 on 12LP+ seems like it would be cheaper than Mendocino on 6nm.

            Comment


            • #7
              I thought they were over Vega graphics but apparently no. That is a shame as Vega was the child AMD hated. But there is some good news.

              Brilliant 4K Graphics Resolution
              and Multi-Display Configurability
              A single, small-footprint AMD Ryzen™ Embedded R2000 SoC
              can power up to four independent displays in crisp 4K resolution
              leveraging DisplayPort™ 1.4, HDMI™ 2.0b, or eDP 1.3 interfaces,
              each supporting up to 4K60 resolution. For industrial PC and
              thin client solutions, AMD Ryzen™ Embedded R2000 SoCs help
              enable versatile display and HMI configurations in space and
              thermally constrained environments. Multimedia Hardware
              Accelerator supports decode or encode for select widely used
              video codecs with resolutions up to 4K: HVEC & VP9 10-bit
              decode, H.264 8-bit decode and HVEC & H.264 8-bit encode3.

              Comment


              • #8
                Now if only they would make a ~30W GPU with hardware decoding of everything up to and including AV1. https://en.wikipedia.org/wiki/Unifie...Format_support
                Yeah I know I'm part of an incredibly small niche market that wants to combine a decent CPU with a passively cooled video card. AMD's APUs are almost there: the 5700G is just missing AV1 decode (and half the L3 cache).

                Comment


                • #9
                  Originally posted by Gonk View Post
                  Now if only they would make a ~30W GPU with hardware decoding of everything up to and including AV1. https://en.wikipedia.org/wiki/Unifie...Format_support
                  Yeah I know I'm part of an incredibly small niche market that wants to combine a decent CPU with a passively cooled video card. AMD's APUs are almost there: the 5700G is just missing AV1 decode (and half the L3 cache).
                  The niche could expand if people just give up on chasing higher resolutions and framerates. You can make the argument for 8K resolution (esp. VR) at 1000 Hz, but most people seem to be fine with 1080p60. Rembrandt will do well at that, Phoenix Point should be +50% over Rembrandt, and Strix Point will deliver some improvement. All of those will use DDR5, which has design changes to aid overclocking which can drive up the memory bandwidth that any APU needs. Keep in mind that we don't know if any particular APU family will come to DIY desktops, or if it will be OEM-only (Renoir, until the 4600G recently) or OEM-exclusive for several months (Cezanne, e.g. 5700G).

                  I would like to see AV1 decode everywhere too. AMD was late to the party but it should be ubiquitous on the AM5 socket, since every CPU should have RDNA2 graphics and at least VCN 3.0. So you'd get it from Rembrandt or Raphael. RDNA3 should be paired with AV1 encode, so that could come as soon as Phoenix Point. I would like to see rapid progress on the development and rollout of AV2 to keep H.266 et al. at bay, but that's wishful thinking.

                  On cache, I don't know what to expect. 5700G was already a big improvement, quadrupling the usable L3 per core from the 4700G. L2 cache may be doubled in all Zen 4 products including the APUs. It would be nice to see 3D L3 cache added to some APUs, if only on the top-end parts, like a 7900HX3D or 7700G3D. AMD is probably struggling to get a good supply of 3D cache chiplets from TSMC, and will prioritize Epyc over anything else, so we will see staggered launches like Zen 4 with 3D V-Cache.

                  But you should ask yourself how much it matters. Tripling cache in the 5800X3D only increased average gaming performance by 15% (maybe it would have been 20% without clock speed regressions). Some workloads can get a massive 50-100% boost, others get nothing. I wouldn't necessarily consider the 4800U/4700G/4900H to be bad even though each core can only access 4 MiB of L3 with 8 MiB total. You already got a huge improvement going to the 5700G, with 16 MiB L3 and the unified 8-core CCX.
                  Last edited by jaxa; 21 June 2022, 10:41 PM.

                  Comment


                  • #10
                    Originally posted by jaxa View Post
                    It would be nice to see 3D L3 cache added to some APUs, if only on the top-end parts, like a 7900HX3D or 7700G3D
                    I got the impression that it wasn't possible to manage the heat. Maybe that will change to some extent over time, between more advanced nodes and physical rearrangement, but it seems likely to me that it will simply push the problem around at best for at least a few more years.

                    Comment

                    Working...
                    X