AMD Ryzen AI 9 HX 370: 100+ Benchmarks Validate Zen 5's Captivating Power Efficiency & Performance

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • sophisticles
    Senior Member
    • Dec 2015
    • 2591

    #31
    Yawn!

    As already pointed out, the performance is not that great when you consider 12C/24T for the new chips vs 8C/16T for the older chips, gee, I wonder which one is going to be faster?

    The power consumption is a nice improvement, very impressive, though I expect Lunar Lake to take these chips to school in that department.

    Overall, it looks like a hard pass is in order for these chips from Linux users perspective, at least until Linux support is up to snuff.



    Comment

    • ddriver
      Senior Member
      • Jan 2014
      • 722

      #32
      Originally posted by avis View Post
      Core to core latency is alarming:
      Typical penalty for traversing between hybrid core clusters.

      In fact, AMD"s numbers are QUITE GOOD, and the penalty is only when traversing clusters. The in-cluster latency is very good.

      Intel's big core best case is around 35 vs 25 for amd, and "economy" core is 50-75 for intel vs 35-40 for amd. That's also better than crapple and crapcomm. Should the entire industry be alarmed?

      I don't expect the cluster traversal to matter that much in scenarios that can actually saturate all the cores.

      And besides, amd will likely retain non-hybrid skus for the time being, since they already offer ample performance and features, and their manufacturing cost is lower. So there will be a solution for latency sensitive workload laptop niche.
      Last edited by ddriver; 28 July 2024, 12:45 PM.

      Comment

      • scottishduck
        Senior Member
        • Jun 2011
        • 498

        #33
        Originally posted by Chugworth View Post
        What sort of apps could take advantage of Ryzen AI? Could it potentially replace Frigate's need of a Google Coral chip?
        It’s there to appease shareholders. AI hardware still has very few or zero use cases on a laptop.

        Comment

        • coder
          Senior Member
          • Nov 2014
          • 8930

          #34
          Originally posted by drakonas777 View Post
          Both HX 370 and 7840HS are using the same lithography.
          From what I heard, Zen 5 is using TSMC N4P, which is a different process than the N4 node used by the 7840HS.

          The difference won't be massive, but they're not the same node.

          Comment

          • coder
            Senior Member
            • Nov 2014
            • 8930

            #35
            Originally posted by sophisticles View Post
            I expect Lunar Lake to take these chips to school in that department.
            I think Lunar Lake is set to launch on TSMC N3B (shown above as just N3), giving it a somewhat natural node advantage over these Ryzen models. Like Zen 5, Intel seems to have focused primarily on IPC improvements, so I expect its efficiency to be good.

            Its biggest disadvantages will be the loss of Hyperthreading + competing with a maximal configuration of just 4P + 4E.​ On heavily-multithreaded workloads, Intel will have no answer until Arrow Lake launches for laptops, probably at the end of this year or early next.

            Comment

            • coder
              Senior Member
              • Nov 2014
              • 8930

              #36
              Originally posted by ddriver View Post
              Typical penalty for traversing between hybrid core clusters.
              Uh, not for Raptor Lake, at least.

              The 174 - 186 ns numbers in avis' post are more than 3x as bad as Raptor lake and like almost as bad as going to DRAM!

              Originally posted by ddriver View Post
              ​I don't expect the cluster traversal to matter that much in scenarios that can actually saturate all the cores.
              Yeah, I really can't say how widespread the cases are where it would have an impact. It would mainly be an issue for multithreaded apps with lots of data-sharing between threads.

              Comment

              • sdack
                Senior Member
                • Mar 2011
                • 1725

                #37
                Originally posted by Michael View Post
                Efficiency doesn't matter on the desktop side? It very much does...
                Maybe also explain why and not just be the contrarian. It helps people to better understand the reasons and arguments.

                Regarding power efficiency, does it help with keeping a core's temperature low, which in return allows for higher clock speeds. It also allows for more cores to be packed onto a die and to run more parallel threads and processes. Last but not least does it help to save costs and reduces the CO2 footprint of computers.

                Comment

                • NekoTrix
                  Junior Member
                  • Jun 2023
                  • 22

                  #38
                  Originally posted by avis View Post

                  Yeah, I was about to say the same, the new part has 50% more cores yet it doesn't excel over 7840HS/U that much.

                  It's extremely power efficient though, I'll give it that. Zen 4 owners may not need to apply, previous generations laptop owners will see massive improvements.

                  Overall a decent job but I expected a tad more. It's kinda jarring to see 7840HS beating this new CPU in multiple benchmarks. I don't remember that being the case previously.

                  As a 7840HS owner, I'm looking forward to Zen 6/RDNA 4.0 or if Intel finally digs itself out of the grave, whatever Intel comes up with next (and I'm not talking about Lunar Lake).
                  We are looking at a 45% average power reduction from the 7840HS to the HX 370 with 10% more performance on top, what are you expecting more?... Evidently, this chip is not the one to get if you want to upgrade from a Zen 4 laptop, but a future SKU with a higher TDP.

                  Comment

                  • coder
                    Senior Member
                    • Nov 2014
                    • 8930

                    #39
                    Originally posted by sdack View Post
                    Maybe also explain why and not just be the contrarian. It helps people to better understand the reasons and arguments.
                    Michael is a busy guy, and he has lots of smart folks, like yourself, to help pick up the slack on detailed explanation of things. I'd rather see him focus on the high-value contributions, like benchmarking, and crowdsource most of the commentary.

                    As far as commentary goes, I agree with your explanation.

                    Comment

                    • coder
                      Senior Member
                      • Nov 2014
                      • 8930

                      #40
                      Originally posted by NekoTrix View Post
                      We are looking at a 45% average power reduction from the 7840HS to the HX 370 with 10% more performance on top, what are you expecting more?... Evidently, this chip is not the one to get if you want to upgrade from a Zen 4 laptop, but a future SKU with a higher TDP.
                      I've read this SoC has a configurable TDP of 15 to 54 W. Anandtech tested the same laptop model and found evidence of significant thermal throttling.

                      Don't you wonder what its performance would be like, in a laptop with a more substantial thermal solution and equivalent TDP to that 7840HS?
                      Last edited by coder; 28 July 2024, 03:42 PM.

                      Comment

                      Working...
                      X