Announcement

Collapse
No announcement yet.

AMD Announces Ryzen 7 5800X3D Shipping On 4/20, New Mainstream CPUs

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

  • #31
    Originally posted by cb88 View Post

    This is exactly what power gating already does... turns off sections of cores not in use and it works extemely well on AMD CPUs and is why its pointless.

    The only thing big.Little buys you is more cores in less silicon... granted these aren't "full" cores.

    There quite frankly is little to no reason to have little cores on AMD... its a well known design misstep, and Intel suffered from it with Atom where they were actually better off to implement regular cores and sprint to idle.
    Well it could be an idea if the E-cores where really really efficient, like say some really low powered 500Mhz ARM cores or something that could handle routine OS background tasks.

    Comment


    • #32
      I'm seriously considering getting a 5800X with stepping B2. I need to contact suppliers to ensure the serial codes match...

      Off topic:
      This is an interesting take on getting Nvidia to go open source



      Edit: I just heard that you can't overclock the 5800X3D... wondering what that's about
      Last edited by Jabberwocky; 16 March 2022, 08:31 PM. Reason: Can't OC 5800X3D

      Comment


      • #33
        Originally posted by F.Ultra View Post

        Well it could be an idea if the E-cores where really really efficient, like say some really low powered 500Mhz ARM cores or something that could handle routine OS background tasks.
        That's kind of the problem... for those you are better off having a fast core periodically run, if you are running such tasks periodically that also means you can put the entire system in low power state and not just the CPU and power them back up to run the task in question.

        You'll find that per work done a 500Mhz arm core running constantly is worse than a fast x86 core sprinting to idle... due to that factor.

        The main reason companies end up doing 2 core sizes is to boost specs and save silicon and because thier CPU doesn't have power gating like AMDs CPUS do.

        Note I am not contesting the point that you can build very low power CPUs... just pointing out that their perf/W is often worse than you'd expect... and when fully powered up you kind of have to count that part of the die as dead weight...
        Last edited by cb88; 16 March 2022, 01:21 PM.

        Comment


        • #34
          Originally posted by cb88 View Post

          That's kind of the problem... for those you are better off having a fast core periodically run, if you are running such tasks periodically that also means you can put the entire system in low power state and not just the CPU and power them back up to run the task in question.

          You'll find that per work done a 500Mhz arm core running constantly is worse than a fast x86 core sprinting to idle... due to that factor.

          The main reason companies end up doing 2 core sizes is to boost specs and save silicon and because thier CPU doesn't have power gating like AMDs CPUS do.

          Note I am not contesting the point that you can build very low power CPUs... just pointing out that their perf/W is often worse than you'd expect... and when fully powered up you kind of have to count that part of the die as dead weight...
          Perhaps Intel can add some xtreme high E core that does not draw wallpower but you wind a spring instead :-)

          Comment


          • #35
            From what I heard heard elsewhere, there may be a small error on the processor table. The R5 5500 should be a 6nm Cezanne die, and listed as Zen3+. That also explains why it has less cache and only supports PCIe 3.0

            Originally posted by Venemo View Post
            Why isn't there a 5900X3D or a 5950X3D?
            I don't know their reason, but I am not surprised about the 5900X not being included, since even though it is popular they would be applying their expensive new process to what is in the end a flawed die. I have two guesses about the lack of 5950X3D - thermal issues lowering clocks and not wanting to encroach on Zen 3 Threadripper or Epyc.

            I guess the 5800X3D is released for two reasons. One is that they said they would, and they value people believing their statements. Another is they want to take the 'gaming crown' for a bit before Zen 4 is available.

            Edit: The R7 5700X should be a nice part. Gamers Nexus benchmarked an R7 5800, an OEM only part which may be the same thing, and found it very close to 5800X performance while honouring the 65W TDP.
            Last edited by Teggs; 16 March 2022, 06:34 PM.

            Comment


            • #36
              Originally posted by rclark View Post
              I was hoping for low end 4 core zen 3 processors with the graphics built in for file server builds where a stand-a-lone graphics card isn't necessary but still needed for bios updating and such. Like the 2400G .... at a $100 max price point.
              What's missing from their consumer-level APUs is ECC support. They reserve that for the OEM-exclusive Ryzen Pro APUs.

              You could instead buy a board that has a BMC chip, but those are very expensive and typical BMC performance is horrendously bad, if you actually need to do a significant amount of GUI stuff, locally. I have an ASRock Rack board with an ASpeed 2500. Let me tell you: "speed" doesn't belong anywhere in the name.

              Comment


              • #37
                Originally posted by Venemo View Post
                Why isn't there a 5900X3D or a 5950X3D?
                Doesn't make as much sense.

                Probably, the 3D stacking hurts peak clock speeds. Also, we know that having more cache is mostly a benefit if it's local to the die. If you have to go through the I/O die to reach L3 cache on another chiplet, then the advantage isn't as big.

                So, a single-chiplet CPU is where they get the greatest benefit from it.

                Comment


                • #38
                  Originally posted by schmidtbag View Post
                  Except as you already said, more cores in less silicon. That means more cores at a lower cost.
                  Yup. The data fully supports this. Not only did it increase perf/area and perf/$, but also perf/W.

                  It's not easy to see the latter, because the damn P-cores boost so high. However, if they'd replaced those two quad-Gracemont clusters with two more Golden Cove P-cores, I'm sure the net result would be not only worse perf, but also worse perf/W.

                  Comment


                  • #39
                    Originally posted by Jabberwocky View Post
                    I'm seriously considering getting a 5800X with stepping B2.
                    Would you tell us why?

                    Originally posted by Jabberwocky View Post
                    This is an interesting take on getting Nvidia to go open source
                    I don't like it and I don't expect it'll work.

                    Originally posted by Jabberwocky View Post
                    Edit: I just heard that you can't overclock the 5800X3D... wondering what that's about
                    Stacking cache dies on top of your cores isn't exactly great for keeping them cool.
                    Last edited by coder; 17 March 2022, 05:50 AM.

                    Comment


                    • #40
                      Originally posted by cb88 View Post
                      for those you are better off having a fast core periodically run, if you are running such tasks periodically that also means you can put the entire system in low power state and not just the CPU and power them back up to run the task in question.
                      That's not true. Mobile SoCs can typically scale back the amount of L3 cache they use. They also adjust DRAM clocks based on demand, similar to GPUs.

                      Originally posted by cb88 View Post
                      You'll find that per work done a 500Mhz arm core running constantly is worse than a fast x86 core sprinting to idle... due to that factor.
                      You clearly have no idea just how much more complicated & power-hungry modern performance cores are.


                      A single E-core is using only 15 W total CPU package power vs. 78 W for the P core. And yet, the E-core is still about half as fast as the P-core.

                      Source: https://www.anandtech.com/show/17047...d-complexity/4

                      Originally posted by cb88 View Post
                      when fully powered up you kind of have to count that part of the die as dead weight...
                      E-cores can add meaningful amounts of performance. ARM's Cortex cores aren't the best example, though. However, speaking of die area, the other nice thing about E-cores is they're small.

                      Now, let's see how mobile cores compare:


                      Source: https://www.anandtech.com/show/16983...re-efficient/2

                      So, the Apple's A15 P-cores take 9.3x the power to deliver 3.0x the performance. In terms of energy-consumed, the P-core needed 3.1x to do the same work as the E-core. That's shows "sprint to idle" is definitely not a winning strategy.

                      Comment

                      Working...
                      X