Announcement

Collapse
No announcement yet.

One-Line Patch For Intel Meteor Lake Yields Up To 72% Better Performance, +7% Geo Mean

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

  • #11
    I hate to say it, but I think that Im getting to be grumpy old man that no longer keeps ups with all the codenames... especially wrt processors. I'd rather see them referencenced by generation and/or part name...

    I mean between all of the lakes, canyons and whatnots I've lost track since they were largely meaningless differentiators... Im not sure what idiot manager at Intel decided that naming processors after geological features was bright idea(TM) was a good thing, but this is utterly ridiculous... it's worse when other Intel parts have cutesy names but are regularly referred to by part names, and then to think of it, whats even worse is that same design is referenced by cutesy geological features for consumer parts, YET going to commercial products, nah, we're good with numeric/part numbers...

    Amd is almost as bad, but they do not seem to push their codenames as much as Intel, or seemingly, Michael is doing for Intel, which makes it an incomprehensible muddled mess... but.... CPUs are a muddled mess ATM but I for one would still prefer part numbers of the myriad of codenames which mean nothing to me... especially since even the same generation CPU have have MANY codenames for 'features'...

    Has Intel done the dumb thing and abandoned the i7 etc. designations? They need to rethink that for one thing IMNHO. I mean <generation> i9 i7 etc give me a baseline of performance, and maybe features depending upon how closely that I track it, meanwhile canon lake , meteor lake, useless... I do not work for Intel so those internal codenames are USELESS... and lets be honest they're not marketable either... which is why I guess AMD does not use their internal codenames much other than what is referenced by outside articles...

    Sorry, Ill go back to yelling at the clouds now...

    Comment


    • #12
      Nobody in the world really understands how Intel names their processors, it is a full mess, they did a big mistake with that.
      I guess Intel MARKETING PEOPLE should be FIRED now , they are the worst in the whole universe ... Shame on them !

      Comment


      • #13
        It seems a lot of people here are very easily impressed. 6.7% more performance for 4.9% more power is indeed a net win, in terms of efficiency.

        Let's not forget these are primarily laptop CPUs, though. So, burning more power isn't always a winning proposition. Yes, if you're using the balance_performance power profile, you probably want peak perf/W, which this does improve.

        Comment


        • #14
          Originally posted by Anux View Post
          No regressions in sight, makes me wonder on which basis they chose the initial values.
          When is the last time you ever saw a CPU get slower by increasing its power budget?? That's essentially all their change does!

          Comment


          • #15
            Originally posted by edxposed View Post
            Does this mean that the actual performance gains for lunarlake/arrowlake are almost nil?
            No, Lunar Lake delivers real IPC benefits on a better process node (TSMC N3B). There's no reason to expect it won't perform markedly better, with similar CPU performance governor parameters.

            Comment


            • #16
              Originally posted by sophisticles View Post
              I think it's obvious that AMD has bribed Linux and friends to sabotage Intel products so that AMD CPUs look better in comparison.
              This is too boneheaded even to work as a joke. First, the code they modified is the CPU frequency scaling driver written by Intel, specifically for Intel CPUs. If there's anything not optimal about it, they have no one but themselves to blame.

              Secondly, on your conjecture that someone "sabotaged" it, the git history will prove no such thing happened. andyprough said as much, but I suppose that probably went over your head.

              Lastly, they should've seen such a merge request go by, if they were paying attention as they should be doing.

              Originally posted by sophisticles View Post
              It's the only possibility that makes sense.
              Well, it's a good bet you're not shilling for Intel, since comments like this only make them look worse.

              Comment


              • #17
                Originally posted by panikal View Post
                Having done this actual job at Intel once upon a long time ago (specifically, recommend kernel/bios defaults for new product initiatives in the server division) I can say that these defaults (at least when I was doing it) were the result of setting 'sane' defaults and no one on either the kernel, bios, or product teams realizing that changing it and running a benchmark = big change.
                Maybe that's why companies like Dell have long used their own.

                Originally posted by panikal View Post
                ​Stuff like this is often coded by a dev in a vacuum (red cover guide and maybe a prototype) , with minimal benchmarking on the initial roll-out scenarios because in order for the more main stream Engineering Product teams to test it has to be in a kernel, so tuning parameters like this....fall through the cracks. Someone chose a conservative default that sounded good at the time and no one revisited that decision until now..but someone at least DID revisit it!
                Intel has an entire team devoted just to overclocking!
                Sure, maybe these folks are focused on Windows gaming, but don't tell me there's been nobody doing performance tuning on Linux! Perhaps the reason this slipped through the cracks is specifically because it's a laptop CPU? I could understand if tuning their frequency scaling driver for Linux laptops falls a bit further down the priority stack, at Intel.

                Comment


                • #18
                  Originally posted by cutterjohn View Post
                  I hate to say it, but I think that Im getting to be grumpy old man
                  After writing such a huge rant in a random thread like this, I'd say you're definitely acting like a grumpy old man!

                  Originally posted by cutterjohn View Post
                  that no longer keeps ups with all the codenames... especially wrt processors. I'd rather see them referencenced by generation and/or part name...
                  Oh, but that won't help when they have completely different products for laptop vs. desktop markets, within the same generation. In Gen 14, the desktop market is served by a Raptor Lake "refresh" (or so they call it, but it's literally just a rebadge and better binning of the Gen 13 and some Gen 12 parts). However, their lower-end and mid-range laptop market is served by a very new, tile-based architecture that includes a GPU tile made by TSMC and a CPU tile made on their (then new) Intel 4 node. A similar thing happened in Gen 10 and Gen 11, where laptops got Ice Lake and Tiger Lake, but the desktop market was saddled with 14nm rehashes like Comet Lake and Rocket Lake.

                  Server is likewise very different. Even though Sapphire Rapids nominally used the same Golden Cove P-cores as Alder Lake, they're a bit more like the P-cores in Raptor lake, but with an extra AVX-512 FMA bolted on, AMX, and some other goodies.

                  Originally posted by cutterjohn View Post
                  I mean between all of the lakes, canyons and whatnots I've lost track since they were largely meaningless differentiators...
                  They might sound meaningless, only because codenames are supposed to be devoid of any intrinsic meaning. That's what a codename is!

                  Originally posted by cutterjohn View Post
                  ​Im not sure what idiot manager at Intel decided that naming processors after geological features was bright idea(TM) was a good thing,
                  First, these are internal names. They're not how Intel markets or brands them. Second, you need a naming convention that scales and has enough options that steer clear of controversy and unintended connotations in one language or another, because Intel is a very international company and these codenames obviously leak out.

                  Originally posted by cutterjohn View Post
                  ​​whats even worse is that same design is referenced by cutesy geological features for consumer parts,
                  Wow, that's the first time I ever heard geological features referred to as cutesy! Gosh, I mean, it's not like they're children's cartoon characters, which is what we used at one former job I had.

                  Originally posted by cutterjohn View Post
                  ​​​YET going to commercial products, nah, we're good with numeric/part numbers...
                  I guess you haven't heard they too have names like Cascade Lake, Sapphire Rapids, or Sierra Forest?

                  Originally posted by cutterjohn View Post
                  ​​​​I for one would still prefer part numbers of the myriad of codenames which mean nothing to me...
                  Intel part numbers won't tell you much more. Certainly, not as much as you seem to think they will.

                  Originally posted by cutterjohn View Post
                  ​Has Intel done the dumb thing and abandoned the i7 etc. designations?
                  The iSomethingMeaningless (as SemiAccurate likes to call it) don't tell you very much, because Intel was never consistent about what they meant (i.e. other than that higher i-numbers are more expensive within their product segment).

                  Originally posted by cutterjohn View Post
                  ​​I do not work for Intel
                  🤯

                  Originally posted by cutterjohn View Post
                  ​​​Sorry, Ill go back to yelling at the clouds now...
                  I think we can all understand your frustration. I'm not sure it warranted such a long rant. Next time, perhaps just ask for a cheat sheet to look up Intel code names? Here's the best I could find.

                  Comment


                  • #19
                    Originally posted by Phoronos View Post
                    Nobody in the world really understands how Intel names their processors, it is a full mess, they did a big mistake with that.
                    I won't repeat what I said in my above post about codenames being internal and all of that. One thing I forgot to mention is how you specifically want to avoid names that actually tell competitors anything about the product, in case it leaks out (which they inevitably do).

                    Originally posted by Phoronos View Post
                    I guess Intel MARKETING PEOPLE should be FIRED now
                    If you had a marketing job, you'd be fired a lot faster than any of them would. You clearly don't understand marketing, which aspires to do things like making people want something they don't need and getting them to pay a lot more money than it's really worth to them. You don't accomplish that by being fully transparent about what the product is and what intrinsic value it has.

                    Not that I'm happy about it. I wish as much as anyone that we could have clear, consistent, and straight-forward product names. However, codenames are a different story... those will always be opaque.

                    Comment


                    • #20
                      Originally posted by coder View Post
                      I won't repeat what I said in my above post about codenames being internal and all of that. One thing I forgot to mention is how you specifically want to avoid names that actually tell competitors anything about the product, in case it leaks out (which they inevitably do).

                      If you had a marketing job, you'd be fired a lot faster than any of them would. You clearly don't understand marketing, which aspires to do things like making people want something they don't need and getting them to pay a lot more money than it's really worth to them. You don't accomplish that by being fully transparent about what the product is and what intrinsic value it has.

                      Not that I'm happy about it. I wish as much as anyone that we could have clear, consistent, and straight-forward product names. However, codenames are a different story... those will always be opaque.
                      Intel processors names are impossible to understand and you defend that ? LOL
                      Stop writing please.

                      Comment

                      Working...
                      X