Announcement

Collapse
No announcement yet.

Intel Announces 13th Gen "Raptor Lake" - Linux Benchmarks To Come

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

  • #91
    Originally posted by AdrianBc View Post
    Moreover, the lower SMT gain for floating-point applications is also well known and expected, because the execution time of such programs is dominated by loops with perfect branch prediction and they include a large percentage of computational instructions that can be overlapped over loads, and most data is reused several times, so it is loaded from various cache memory levels, not from the main memory.

    Because of this, most floating-point applications can achieve a very high percentage of use of the execution units, so there are few opportunities for executing the second thread of a core. For floating-point applications, it is not uncommon to achieve better performance by disabling SMT.​
    Thanks for acknowledging that point. I hope you'll further acknowledge that it blows a hole in your theory that 1E = 1P2T / 2, or that such a thing was even a design requirement of Intel's. This is too simplistic.

    They created the Thread Director specifically to aid the OS in more effective thread scheduling, in an acknowledgement of the challenges it poses.

    Comment


    • #92
      Originally posted by coder View Post
      I'll grant you this one point: that TDP is misleading when the actual PPT is 1.35x that much. Based on my simplistic understanding, I don't know why they're not equal. If someone can point me at a compelling rationale, I'd appreciate it.
      TDP is a made-up marketing term that is meaningless, beyond a basic correlation with power use. Only correlation within the same CPU lineup, not across generations, as AMD and Intel fully reserve the right to change variables in their made-up calculations at any point. An example factor that goes into it is "room temperature during testing". Which room temp did they use to get the numbers they are advertising? No idea, they won't tell you that.

      They advertise a 125W TDP part because marketing thinks it sounds better than saying it's a 142W part. Simple as that.

      Comment


      • #93
        I'd say TDP is a rough estimate for thermal solution, which guarantees that CPU is going to work at least at base frequency. This is mostly for shit tier PC builders, which can put a minimal and cheap cooler and more or less be sure CPU won't throttle. Actual CPU power consumption is higher, so if you want to sustain high boosts you need a better thermal solution. This is my interpretation why TDP parameter exists and why it's lower then actual consumption.

        As for ZEN4, it's obvious that AMD made stock power parameters stupid to compete with Intel's stupid PL2. So in reality we should evaluate CPU performance in some sane power range, say 65-150W range. I don't see the point in performance graphs where CPUs draw 240+ W on mainstream desktop. It's insane. AFAIK performance gain in running ZEN4 beyond 150W are basically negligible, so extra 100W for <~15% is just irrational. It should not be a default.
        Last edited by drakonas777; 02 October 2022, 09:07 AM.

        Comment


        • #94
          Originally posted by coder View Post
          Try 13%.



          I don't really get what you're complaining about. Isn't it the dream of overclockers to have a CPU that's only thermally-limited? If you want to impose lower power limits, you can do it in BIOS.


          That's not at all atypical, when you do extreme overclocking, which is essentially what he did.


          I'll grant you this one point: that TDP is misleading when the actual PPT is 1.35x that much. Based on my simplistic understanding, I don't know why they're not equal. If someone can point me at a compelling rationale, I'd appreciate it.


          power consumption != power efficiency. Also, power efficiency changes, depending on the SKU and TDP configuration. It's not a single number that characterizes all models in all configurations.

          Because of that, it really matters why you're looking at it. If you just want to compare the microarchitecture and manufacturing process, then you will want to compare comparable models running in a similar power envelope (and not a similarly-named power envelope, but as close as you can get to one that's actually equivalent).

          If you want to compare the typical end user power efficiency, then compare comparable models at stock settings, with a normal case & cooler, running on a defined workload.

          People tend to take the the highest number from the most extreme part, in the highest-power configuration and use that to characterize the entire product line. However, that's only applicable to those intending to run that part in that configuration.
          GN got 251W on stock not by extreme overclocking. Only thing you need to get 250W+ is very good cooler

          Comment


          • #95
            Originally posted by drakonas777 View Post
            I'd say TDP is a rough estimate for thermal solution, which guarantees that CPU is going to work at least at base frequency.
            That used to be the case, but if PPT is 1.35x of TDP and unlimited in time, then it renders TDP a complete fiction for any purpose.

            This situation has gotten so ridiculous that we really need governments to start dictating how these metrics should be quantified, like they do the fuel-efficiency metrics for automobiles. That's the only way to stop this madness.

            Comment


            • #96
              Originally posted by piotrj3 View Post
              Anyway my issue is that AMD changed definition of their TDP (what video exactly mentions). Before TDP was actually the power your CPU drew from EPS rail.
              ... What?! No it wasn't, and it was never that even back in the days when Intel - which is, incidentally, not the same company as AMD - still even pretended it was.

              > Intel meanwhile implies 2 things one is base power draw and boost power draw and with exception of some AVX512 workloads you will not break that boost power draw.

              Again: ... What?! Intel has been lying about TDP since before you were born.

              coder

              re TDP in general: it's a fiction. It's always been a fiction, but now it's just absurdly so (in much the same way that lies in, say, politics grow over time, like boiling a frog).

              That's it. That's literally all there is to it.

              The "rough" TDP is an imaginary thermal output under an imaginary load with an imaginary cooler in an imaginary environment, which once upon a time was based on vaguely-realistic numbers for at most one of those terms.
              The "official" TDP is whatever nice round number is vaguely within cannon range of the original fictitious number that Marketing likes, and is slightly lower per perf unit than whatever the competition's is.

              That's *before* the creation of boost clocks, let alone the infinite boost clocks that have been around since ?Sandy? ?Ivy?.

              AMD's 90W sustained in 65W mode is, I suspect, a bug rather than willfully deceptive **, but Intel has been "off" by staggering amounts at times for years now on multiple chips, so it's possible AMD is just following suit because it has to. Obviously, if one company is misrepresenting its CPUs by well over 50W, and one is "only" doing so by 10W, the latter is going to get creamed on perf/W judgements unless they have a design that's 20%+ more performant at a given power draw. We've seen one of those in the past decade, thanks to 14++++ vs TSMC7 Zen, so it can happen, but it's pretty rare and not something you can bet the company on year after year for a decade or more.

              ** Simply because at the other power draw targets it's pretty close, but the 65W behavior is an outlier so I'm giving them the benefit of the doubt. I won't be shocked if no AGESA/microcode update fixes it for several months though, or indeed, ever.

              As far as "then it renders TDP a complete fiction for any purpose" goes? Yeah, pretty much. I'm not sure exactly when it went from "kinda sorta at least *vaguely* representative of reality" to "not even remotely so" - like I say, it was boiling a frog - but it's been there for a very long time now, and I don't see that ever improving. (Unless forced to by some EU regulation or something, but they've got bigger fish to fry right now, and Intel has deep pockets).

              edit> I'm buried in post-vacation backlog right now, so I'm trying to minimize posting, but if you need a couple of examples to understand how the process is perverted let me know.
              Last edited by arQon; 03 October 2022, 02:46 AM.

              Comment


              • #97
                Originally posted by coder View Post
                That used to be the case, but if PPT is 1.35x of TDP and unlimited in time, then it renders TDP a complete fiction for any purpose.

                This situation has gotten so ridiculous that we really need governments to start dictating how these metrics should be quantified, like they do the fuel-efficiency metrics for automobiles. That's the only way to stop this madness.
                I'm not sure I understand your point. Well, at least in theory it's possible for TDP be lower than PPT given only a fraction of PPT is required to sustain base frequency. For example, TDP is 125W, PPT is 170, you put on 125W solution, your CPU starts to draw more than 125W until it reaches a thermal limit during which algorithm will drop frequency and voltage until power draw is reduced at say ~130W which would be sufficient for base frequency. I'm not saying this is the case for ZEN4, i'm saying this is possible in theory. Though it feels to me that 105W for 7700X and 170W for 7900X/7950X should be really enough to sustain base, power scaling shows that 170W gives around 80-85% (if i recall correctly nuilzoid video) performance for 7950X so it seems you can actually even get some boost there. The price is of course CPU is going to be at very high temp.
                Last edited by drakonas777; 03 October 2022, 08:12 AM.

                Comment


                • #98
                  Originally posted by drakonas777 View Post

                  I'm not sure I understand your point. Well, at least in theory it's possible for TDP be lower than PPT given only a fraction of PPT is required to sustain base frequency. For example, TDP is 125W, PPT is 170, you put on 125W solution, your CPU starts to draw more than 125W until it reaches a thermal limit during which algorithm will drop frequency and voltage until power draw is reduced at say ~130W which would be sufficient for base frequency. I'm not saying this is the case for ZEN4, i'm saying this is possible in theory. Though it feels to me that 105W for 7700X and 170W for 7900X/7950X should be really enough to sustain base, power scaling shows that 170W gives around 80-85% (if i recall correctly nuilzoid video) performance for 7950X so it seems you can actually even get some boost there. The price is of course CPU is going to be at very high temp.
                  TDP is supposed to be thermal design for extended workload, not transient. Of course i don't mind if CPU has temponary higher transient load because thermal mass of radiatiors/coolers will be enough to handle short power spike.

                  However if extended workload processor consumes more power by 30-40% then TDP indicates, it means AMD's claim is entirly fictional.

                  Intel actually almost always abides those limits. In Blender/Cinebench/7zip etc. you won't see 12900k consuming for extended periods of time more on average then 241W (+- few watts). Only extreme stress testers like prime95 kind of pushes limit more but i wouldn't call it representative. Meanwhile in Blender AMD made 170W chip push 250W on average in extended period of time. What is more, it is way harder to cool 7950X (comparing to Intel chips) as AMD for some outrageous reason decided to make heat spreader extra tall, and smaller meaning most people can't even make it to 250W without some atrocius setup. der8auer who made direct die cooling (remove heat spreader and put cooler directly on top of chip) decreased tempretures by 22C on 7950X.


                  arQon Power draw from EPS rail (on average) is kind of equalivent to TDP as essentially all electricity you use is thermal in the end. Of course definition varies but TDP was mostly information made for sake of cooler producers in the past. Since then agressive boosting was introduced Intel dropped TDP and introduced PL1 and PL2 and simply set PL2 either unlimited, or limited in time. And since then with small exception of strong AVX loads/stress testers Intel actually abides it. And AMD with Zen 3 and before was also abiding it up to few watts diffrence.
                  Last edited by piotrj3; 03 October 2022, 11:31 AM.

                  Comment


                  • #99
                    Originally posted by piotrj3 View Post
                    Intel actually almost always abides those limits. In Blender/Cinebench/7zip etc. you won't see 12900k consuming for extended periods of time more on average then 241W (+- few watts).
                    But 12900k advertises 125 W TDP.

                    Comment


                    • Originally posted by Anux View Post
                      But 12900k advertises 125 W TDP.
                      https://ark.intel.com/content/www/us...-5-20-ghz.html
                      No, Intel advertises 125W as "Base frequency" power. It also mentions maximum power draw (241W) for boost loads over periods longer then 1 second (quoting)
                      Maximum Turbo Power
                      The maximum sustained (>1s) power dissipation of the processor as limited by current and/or temperature controls. Instantaneous power may exceed Maximum Turbo Power for short durations (<=10ms). Note: Maximum Turbo Power is configurable by system vendor and can be system specific.
                      Also Intel made it public in 12900k annoucment that for K chips PL2 is no longer time limited, and PL1/PL2 time limits are existing in BIOSes as well as official intel extreme tuning utility (and if you visit data sheet link that is more documented above, you can find them how they work too)

                      From volume 1
                      Thermal Considerations
                      The Processor Base Power (a.k.a TDP) is the assured sustained power that should be
                      used for the design of the processor thermal solution, Design to a higher thermal
                      capability will get more Turbo residency. Processor Base Power is the time-averaged
                      power dissipation that the processor is validated to not exceed during manufacturing
                      while executing an Intel-specified high complexity workload at Base Frequency and at
                      the maximum junction temperature as specified in the Datasheet for the SKU segment
                      and configuration.
                      Note: The System on Chip processor integrates multiple compute cores and I/O on a
                      single package. Platform support for specific usage experiences may require additional
                      concurrency power to be considered when designing the power delivery and thermal
                      sustained system capability.
                      So TDP according to Intel is power draw at BASE frequency, without including SoC additional power draw (considering majority of chipset is inside CPU and it is responsible for a lot of IO it makes sense to exclude it). So few watts above power limit (if it happens) comes from IO/SoC related stuff. This happened both for Intel and AMD (5950x by comparison has commonly 105W TDP but likes to load up to 120W, but it is acceptable considering you can't quite count in those watts spent on IO/motherboard etc. related tasks).

                      But because TDP is useless metric in boosting era of processors intel gave you PL1 and PL2 (and PL3/4 but they are disabled by default) to better represent those loads and in chapter just after one i quoted they are very well described how in general they work. Somewhere there is also information about 12th gen K chips being not limited to PL2.
                      Last edited by piotrj3; 03 October 2022, 12:48 PM.

                      Comment

                      Working...
                      X