AMD Ryzen 5 9600X & Ryzen 7 9700X Linux Performance With 105 Watt cTDP

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • qarium
    Senior Member
    • Nov 2008
    • 3438

    #21
    Originally posted by skeevy420 View Post
    I run my 7800X3D in 65W mode. At the end of the day it's a frequency capped CPU so anything that helps keep the temps down to prevent any potential thermal throttling can have a potential net gain in performance. Heck, running Ryzen Master to help find the optimal UV is one of the very few reasons I keep a Windows install. If any of y'all use Ryzen Master, the Ryzen Master stability test is insufficient. Prime95 fails where Ryzen Master reports successful and stable. That's really useful for finding optimal per-core values. Also, I have a 6700 XT running at 1440p. I'm GPU-bound in nearly every single game. Giving the CPU more power won't fix that.
    I really wish that AMD would release a firmware update that would increase the 7800X3D and 7900X3D to the frequency of the 7950X3D. To me, there's no reason that more cores should be faster than less cores when the issue is thermal limitations.
    sadly AMD do limit the clock speed on the smaller cpus to sell more of the expensive cpus...
    it has marketing reasons means PR... and the money nobility part they claim they make more profit this way.

    but i fully agree with you that this is nonsense. a gamer who buy a 7800X3D or future 9800X3D will only pay money for a higher clocked 7800x3D or higher clocked 9800X3D but they will not buy a 7900X3D/7950X3D or 9900X3D or 9950X3D

    AMD has really a wrong understanding here. i know no one as gamer who buy 7950X3D instead of 7800€ only because of 100-200mhz more clock speed.

    so this is true AMD just sapotage itself.
    Phantom circuit Sequence Reducer Dyslexia

    Comment

    • coder
      Senior Member
      • Nov 2014
      • 8922

      #22
      Originally posted by loganj View Post
      only 6% more performance for so much more power consumption.
      Yeah, you can easily see why AMD went with a lower TDP. Not only that, but single-threaded and gaming performance got almost no benefit from going to 105 W, however giving it a higher TDP means a bigger & more expensive cooler is required.

      I think it was okay for AMD to ship with 65W and let anyone with MT-heavy workloads just enable PBO and use a bigger CPU cooler.

      Originally posted by raystriker View Post
      6% geomean,,, not all workloads scale linearly with power consumption.
      Okay, the best got 23% more performance for 61.5% more power. That's not good.

      Below that, four more clustered around 21%. The first big cluster is around 17%.

      IMO, 105W is too damn high, for the 9700X. They should've picked a slightly lower value, like 90W. I'll bet that would've unleashed nearly the same benefits.

      Originally posted by raystriker View Post
      Also undervolting is a VERY big thing with modern Zen.
      Undervolting is something you do at your own peril. I'd rather not risk my system stability or data integrity.
      Last edited by coder; 12 September 2024, 03:15 AM.

      Comment

      • coder
        Senior Member
        • Nov 2014
        • 8922

        #23
        Originally posted by sophisticles View Post
        Basically some motherboards now allow overclocking, that's really what the story is.
        Nope. Changing the power limits doesn't affect the frequency limits. Overclocking is exactly what it sounds like: exceeding the manufacturer-specificed clock frequencies.

        Originally posted by sophisticles View Post
        I have never been a big fan of overclocking, and i certainty would not run a processor that originally was sold as a 65w part at a 105w setting, you're just begging to damage something.
        The CPU will throttle, if temperature limits are exceeded. The motherboard won't let you draw more power than it can provide.

        Increasing power limits is actually one of the safer ways you can increase CPU performance. So much so, that Intel has long allowed you to do it, even on the non-K models and non-Z motherboards (neither of which support overclocking). It won't void your warranty, either (which is not true of overclocking).

        Originally posted by sophisticles View Post
        It's also ironic that AMD, which in addition to more cores, has been marketing its processors as being lower power consumption than Intel's should suddenly allow their processors to be taken to setting that consumes an extra 40w.
        Yeah, they're flailing to damage-control a botched launch of the Ryzen 9000 series. One thing a lot of reviewers and commentators focused in on (I think wrongly, for the most part) was the 9700X 65W TDP, yet it's meant to be a successor to the 7700X that has a 105W TDP. AMD is trying to address the various criticisms they got, including this (mostly misguided) point.

        Lucky for them, the launch of Arrow Lake has now slipped to the latter part of October. Already, gaming performance on Ryzen 9000 is looking better than it did in the initial launch reviews.
        Last edited by coder; 11 September 2024, 11:12 PM.

        Comment

        • blackshard
          Senior Member
          • Oct 2009
          • 602

          #24
          Originally posted by qarium View Post

          sadly AMD do limit the clock speed on the smaller cpus to sell more of the expensive cpus...
          The reason is very simple: this way the single thread benchmarks (and gaming, often) nicely stack up from the higher range products to the lowest.

          That thing does not affect the real performance all around at all, because really intensive workloads are usually spread over multiple cores, or at least developers struggle to do so.

          Ps: intel lineup has the very same 100/200mhz small decrement ranging from higher to lower range for the same reason

          Comment

          • haplo602
            Senior Member
            • Aug 2009
            • 259

            #25
            Originally posted by skeevy420 View Post

            You sound like someone who bought a 7950X3D and wants to be able to lord their precious performance gain over everyone else. Whoop-de-doo.

            With every other Zen CPU with an X in the name when you want your system to run faster than stock you enable PBO, the system will OC itself, and then you can fine tune it even further from there if you're willing, except with X3D cores for supposed stability concerns. I say supposed due to the product line limitations and BCLK OCs. I could accept the product line limitations if they'd give us the option and opportunity to fine tune our performance and stability on our own like how we can with every other consumer Zen CPU. It sucks that we're not given the option to find out outside of BCLK OCs which are only so stable due to their effect on everything, which is a bit of a tease because it does show that the X3Ds can be ran a hair faster and still be stable enough to pass tests.

            I knew what I was getting into when I bought my 7800X3D for less than $350 after tax. Talk about an offer you can't refuse. I just think I'd be nice to have the option to try to succeed or fail like with all the other X models.
            The problem here is the 3D V-cache on the X3D CPUs as it is much more thermally limited. So you cannot apply the same rules as to the other CPUs. That's the main reason why the 7800X is faster clocking than the 7800X3D. The dual CCD X3D parts have an uncapped CCD without V-cache and a capped CCD with V-cache.

            Comment

            • raystriker
              Phoronix Member
              • May 2023
              • 53

              #26
              Originally posted by sophisticles View Post

              Are saying you don't bother to proofread your posts?

              I don't care what year it is, overclocking a computer component is asking for trouble.

              Overclocking is great of you want to run a few benchmarks but of you want to get a long project finished you need stability and you will not get that from an overclocked CPU.

              That's goes for any CPU from any manufacturer.
              Maybe you're a visual learner so this will help.

              When AMD is designing a chip, they leave a lot of things undecided (like the final frequency, or the tdp of the chip etc) till it is productized.

              Do you seriously think the 7700 and 7700X are different chips? No they're not. They're just binned differently. For the Zen 4 launch, due to competition from Alder Lake, AMD decided on a tdp of 105W for the 6 and 8 core X skus.

              But with Zen 5, AMD thought efficiency might attract people so they decided on 65W (to look much better when compared to Intel). Gamers did not like that, so now mobo manufacturers are giving a VERY easy way to just up the TDP as if the 9700X was a 105W part to begin with. This is no different from AMD deciding before launch whether to launch the 9700X as a 105W or a 65W part.

              Also another "logic" check for you...the 9700X is a 65W part but the 9950X is 170W? So from 8 cores at 65W to 16 at 170W HOW DID THEY DO THAT WOWOW MUST BE A DIFFERENT Zen 5 CCD...right? /s
              image.png
              Last edited by raystriker; 13 September 2024, 08:26 PM.

              Comment

              • skeevy420
                Senior Member
                • May 2017
                • 8633

                #27
                Originally posted by haplo602 View Post

                The problem here is the 3D V-cache on the X3D CPUs as it is much more thermally limited. So you cannot apply the same rules as to the other CPUs. That's the main reason why the 7800X is faster clocking than the 7800X3D. The dual CCD X3D parts have an uncapped CCD without V-cache and a capped CCD with V-cache.
                Regular cores aren't the entire issue here, but they do apply since they do bring an increase in heat generation.

                The problem is that 3D cores of the 7950X3D are clocked higher than the 7900X3D which are clocked higher than the 7800X3D. As you go up the product line each model gains more cores clocked higher which means more that has to be cooled. If thermal limitations were 100% the issue then 7900X3D or 7600X3D would be the highest clocked or boosting capable 3D models since they have the least amount of 3D cores to reduce heat from. If 3D cores just get screwy above a certain frequency, then they should all just boost to 5.25Ghz like the 7950X3D and allow the regular cores to make up the difference in performance. That would make the thermal limitations claim seem a lot more credible and believable.

                Historically, the way practically every CPU product line worked until very recently, the more cores it had then the slower it was clocked due to thermal concerns. Now we're being told that thermal limitations are the problem on an 8c 3D CPU while an 8c + 8c 3D CPU runs faster and hotter from the factory. That smells like some bullshit and not something that's caused by bad binning creating multiple product lines like the 7600X3D or 5700X3D coming out so much later. As a consumer, this issue is making me not trust AMD because it makes no sense scientifically. It only makes sense capitalistically. It makes AMD executives look like a bunch of liars because they're saying things that go against the laws of thermal dynamics. If thermals are really the issue then the things with more thermal headroom should be able to make use of it, but everything works backwards in X3Dland.

                The lower core 3D models should be like the old Intel Westmere X5698 which had reduced cores and a nearly 1ghz higher clock speed than the X5690 and other high end Westmere CPUs. I used to have a dual [email protected] setup which is why I remember that X5698 and how unique it was. That's how the 7600X3D should perform -- the fastest 3D cores possible due to having less to keep cool. Instead, the 3D CPU with the most thermal headroom is the slowest from the factory. Binning, Bullshit, or a Bit o' Both?

                Comment

                • sophisticles
                  Senior Member
                  • Dec 2015
                  • 2591

                  #28
                  Originally posted by coder View Post
                  Nope. Changing the power limits doesn't affect the frequency limits.
                  Of course it does.

                  Where do you think the extra performance comes from?

                  Higher frequencies obviously.

                  Why are you being obtuse?

                  Is it deliberate?

                  Comment

                  • user556
                    Phoronix Member
                    • Jul 2019
                    • 114

                    #29
                    Yeah, the cooling needs beefed up to match. As others have observed, the average running temperature affects the peak boost frequency in bursty type tests like compiling and web page rendering.

                    Comment

                    • coder
                      Senior Member
                      • Nov 2014
                      • 8922

                      #30
                      Originally posted by sophisticles View Post
                      Of course it does.

                      Where do you think the extra performance comes from?

                      Higher frequencies obviously.​
                      No, the effect of raising power limits is to enable the CPU to increase multi-core speeds closer to the frequency limits. No matter how much power you let it use, it still won't exceed the hard frequency limits. Enabling PBO lets you bypass those frequency limits, but at the expense of voiding the warranty.

                      The same is true for Intel CPUs. You can let them use more power, to the point that they're either frequency-limited or thermally-limited. What power limits you use is separate from what frequency limits are in effect.

                      Originally posted by sophisticles View Post
                      Why are you being obtuse?

                      Is it deliberate?
                      No, I'm trying to emphasize nuances that you were steamrolling over. Power and frequency are related, but they each have distinct limits that exist for different reasons.

                      The main thing that kills CPUs is voltage. Voltage increases with frequency. That + potential for instability are the reasons frequency is capped.

                      Comment

                      Working...
                      X