Announcement

Collapse
No announcement yet.

Arm Cortex-X3 + Cortex-A715 Announced As Second-Gen Armv9 CPUs

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

  • Arm Cortex-X3 + Cortex-A715 Announced As Second-Gen Armv9 CPUs

    Phoronix: Arm Cortex-X3 + Cortex-A715 Announced As Second-Gen Armv9 CPUs

    Arm today announced their second-generation Armv9 CPU designs with the Cortex-X3 and Cortex-A715. Arm also refreshed the Cortex-A510 to allow for more cores and a power reduction...

    https://www.phoronix.com/scan.php?pa...X3-Cortex-A715

  • #2
    When will there be an ARM SoC that can compete with Apple?

    Comment


    • #3
      That might be interesting.

      Current Cortex-X2 powered laptop designs are maxing at around 1100 on single-core Geekbench scores
      (compared to 1500-1800 on contemporary Intel/Amd/Apple laptops).

      If Cortex-X3 really adds another 25% speed-up and gets to about 1400, then that's in the ballpark of a usable laptop provided that the price is right.


      Edit:
      Apparently not.
      Wikichip is saying the performance (IPC) improvement is approximately 11% on real-world applications.
      Their article is well worth reading: https://fuse.wikichip.org/news/6855/...ore-cortex-x3/
      Last edited by pkese; 28 June 2022, 04:48 PM.

      Comment


      • #4
        Originally posted by pkese View Post
        That might be interesting.

        Current Cortex-X2 powered laptop designs are maxing at around 1100 on single-core Geekbench scores
        (compared to 1500-1800 on contemporary Intel/Amd/Apple laptops).

        If Cortex-X3 really adds another 25% speed-up and gets to about 1400, then that's in the ballpark of a usable laptop provided that the price is right.


        Edit:
        Apparently not.
        Wikichip is saying the performance (IPC) improvement is approximately 11% on real-world applications.
        Their article is well worth reading: https://fuse.wikichip.org/news/6855/...ore-cortex-x3/
        Thank you for the very informative link about cache, pipelines and branch predictors. Will definitely bookmark!

        With the Cortex-X2, the underlying circuitry used for handling 32-bit architectural-related elements was removed, saving on transistors and simplifying some structures.
        Finally!!!

        On a side note, I was kinda hoping armv9 would become a target at which Linux distros would offer packages (kinda like x86-64v3 that Arch was looking at) for those juicy extensions. But oh well, I guess you can't beat gentoo!

        Comment


        • #5
          Originally posted by MastaG View Post
          When will there be an ARM SoC that can compete with Apple?
          Sorry this is off topic since this has nothing to do with Cortex.

          A startup company named Nuvia founded by three former Apple engineers that worked on Apple's A-series chips was acquired by Qualcomm in 2021 and Qualcomm announced that Nuvia's chips will be a PC computing first product released next year.

          Comment


          • #6
            What's up with these number bumps? First we were in Ax, then A15, then A5x and now A7xx?

            What's next... Cortex-A81235?

            Comment


            • #7
              Originally posted by kvuj View Post
              On a side note, I was kinda hoping armv9 would become a target at which Linux distros would offer packages (kinda like x86-64v3 that Arch was looking at) for those juicy extensions. But oh well, I guess you can't beat gentoo!
              There's not much gain for regular software to target armv9 specifically.

              Most of changes are related to either:

              1) Security, which consists of a few instructions that kernel or hypervisors may use (and no compiler will emit anyway - system software calls those instructions through custom hand-written assembly)

              2) Some additional vector instructions ... Vectorizable code is usually built and distributed through multiple dynamic libraries, similarly as how video encoders provide several libraries targeting e.g. various levels of SSE and AVX on x86 CPUs ... and then get dynamically loaded depending on how CPU identifies its capabilities.

              Comment


              • #8
                Originally posted by tildearrow View Post
                What's up with these number bumps? First we were in Ax, then A15, then A5x and now A7xx?

                What's next... Cortex-A81235?
                Their previous generations were a bit confusing, no clue about why they went up from A5X to A5XX, A7X to A7XX though.

                ARM probably has one of the easiest naming conventions since they changed their naming up. A5X - "LITTLE" cores, A7X "big" cores, X - high-performance.

                Comment


                • #9
                  Originally posted by pkese View Post
                  That might be interesting.

                  Current Cortex-X2 powered laptop designs are maxing at around 1100 on single-core Geekbench scores
                  (compared to 1500-1800 on contemporary Intel/Amd/Apple laptops).

                  If Cortex-X3 really adds another 25% speed-up and gets to about 1400, then that's in the ballpark of a usable laptop provided that the price is right.


                  Edit:
                  Apparently not.
                  Wikichip is saying the performance (IPC) improvement is approximately 11% on real-world applications.
                  Their article is well worth reading: https://fuse.wikichip.org/news/6855/...ore-cortex-x3/
                  Its 25% total speed up (IPC+Clocks), and still a valid improvement.

                  Comment


                  • #10
                    Originally posted by pkese View Post

                    2) Some additional vector instructions ... Vectorizable code is usually built and distributed through multiple dynamic libraries, similarly as how video encoders provide several libraries targeting e.g. various levels of SSE and AVX on x86 CPUs ... and then get dynamically loaded depending on how CPU identifies its capabilities.
                    Sure, specialized software like kernels and video decoders already have handwritten assembly, but the Arch guys saw some impressive improvements for a wide utilization of x86-64v3 (10% in Firefox without even rebuilding all of the dependencies). On the ARM side, I would be curious how SVE2 instructions compare to vanilla armv8.

                    Comment

                    Working...
                    X