Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Sunday March 07 2021, @04:43AM   Printer-friendly
from the notsogood-things-come-to-those-who-wait? dept.

Intel's next-generation "Rocket Lake" CPUs will be some of Intel's last desktop models on a "14nm" node, and include "backported" Willow Cove cores (referred to as "Cypress Cove") from "10nm" Tiger Lake mobile CPUs, with improved instructions per clock. Notably, the lineup only goes up to 8 cores, instead of 10 cores for the previous Core i9. The review embargo ends on the launch date, March 30th, but some retailers have been selling the CPUs early. AnandTech obtained an 8-core i7-11700K and wrote a review of it. The results were not great.

Power consumption of the 125 W TDP chip peaked at 224.56 W when running an AVX2 workload, compared to 204.79 W for its i7-10700K "Comet Lake" predecessor and 141.45 W for AMD's Ryzen 7 5800X. The i7-11700K reached 291.68 W with an AVX-512 workload.

The i7-11700K not only failed to beat the 5800X in many benchmarks, but trailed the previous-gen i7-10700K in some cases. The major exception is performance in AVX-512 workloads. Gaming performance of the i7-11700K was particularly bad, in part due to an increase in L3 cache and core-to-core latency.

It's possible that there will be some improvements from a final microcode update before launch. There's also models like the Core i9-11900K, which have the same 8 cores but can clock up to 300 MHz higher.

See also: Intel Core i7-11700K 8 Core Rocket Lake CPU Review Published By Anandtech – Very Hot, Consumes More Power Than Core i9-10900K & Slower Than AMD In Core-To-Core Tests

Related: Linus Torvalds: Don't Hide Rust in Linux Kernel; Death to AVX-512
Former Intel Principal Engineer Blasts the Company
Gigabyte Confirms Intel Rocket Lake Desktop CPUs Will Launch in March


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 4, Interesting) by fakefuck39 on Sunday March 07 2021, @06:09AM (7 children)

    by fakefuck39 (6620) on Sunday March 07 2021, @06:09AM (#1120969)

    the tests they ran seem to completely disagree with every other person who ran these tests. in fact, not only does their overall conclusion disagree, the individual tests are not matching either. this makes me thing they screwed up their tests. but let's not stop the anger based on likely misinformation - this is soylent in 2021 after all, not soylent in 2019.

    https://cpu.userbenchmark.com/Compare/Intel-Core-i7-11700K-vs-Intel-Core-i7-10700K/4107vs4070 [userbenchmark.com]

    here we see the new one beating the old one in every category, despite being clocked lower. why did they clock it lower? one can theorize. AVX512 normally clocks the core at half speed. This is terrible when you're mixing in other instructions - so as you normally would in a mixed workload with multiple apps or VMs running. So they figured out a better way, had to clock a little lower, but no longer take the huge hit.

    so, much better CPU. Oh look - if we look at other sources it beat the crap out of the Ryzen they tested too.

    So, let me correct their conclusion to the real world:
    "The i7-11700K beat the 5800X in all benchmarks, as well as the previous-gen i7-10700K in all cases. We are a bunch of lying faggots who are not competent to conduct testing."

    Starting Score:    1  point
    Moderation   +2  
       Interesting=2, Total=2
    Extra 'Interesting' Modifier   0  
    Karma-Bonus Modifier   +1  

    Total Score:   4  
  • (Score: 3, Insightful) by Anonymous Coward on Sunday March 07 2021, @06:33AM (6 children)

    by Anonymous Coward on Sunday March 07 2021, @06:33AM (#1120975)

    It's cool that you based your opinion on the most unreliable and biased site - userbenchmark - against one of the oldest and most consistent sites - anandtech.
    Also AVX-512 very rarely, if ever, goes as low as you claim, which you'd know if you'd actually used a CPU with those instructions yourself.

    • (Score: 2) by fakefuck39 on Sunday March 07 2021, @06:53AM (5 children)

      by fakefuck39 (6620) on Sunday March 07 2021, @06:53AM (#1120980)

      AVX512 drops to about half the normal max (turbo) frequency in every test I've done. Yes, it's on a windows laptop, so the extra heat effect may be affecting it more. The other tests I did was on a 2000+ core esx farm. same thing. Yes, on paper the hit is 25% in performance. I'm talking about clock speed. when you go half the clock speed, you do not go half the performance.

      so I think anandtech may have mixed in some light AVX512 instructions into their tests, and this caused the chip to clock down more.

      and FYI, something that's common knowledge for people: anandtech went to shit after Anand left. Which was years ago. So now it's not a trustworthy source. It's a source that was trustworthy, the ownership and leadership changed, and now it's a site with a narrative, catchy headlines, bad testing, and results that aren't trusted. Welcome to 2016.

      • (Score: 2, Interesting) by Anonymous Coward on Sunday March 07 2021, @07:52AM (4 children)

        by Anonymous Coward on Sunday March 07 2021, @07:52AM (#1121000)

        AVX512 drops to about half the normal max (turbo) frequency in every test I've done.

        That's not consistent with what's supposed to happen unless your tests only use the heaviest instructions (https://en.wikichip.org/wiki/intel/frequency_behavior). It's also not consistent with my own testing.

        so I think anandtech may have mixed in some light AVX512 instructions into their tests, and this caused the chip to clock down more.

        Really? They are running standard benchmarks. Have you even read their conclusions and cache latency analysis? I'll tl;dr: the IPC increase is there, but it gets negated by increased cache latencies, which is the same problem earlier Ryzens had.

        One theory is that Rocket Lake will scale better with faster memory (again, just like Ryzen), but running above specified maximum 3200MHz for Rocket Lake 1) requires a Z-series motherboard, which is expensive 2) invalidates the warranty (at least outside of the EU) since Intel considers this overclocking.

        I found no fault in Anandtech's testing methodology apart from using a ridiculous 2kg copper cooler on Intel platforms, but that still makes the comparison between 10700K and 11700K valid since they used the same hardware for both.

        We'll have to see what other reviews say. One interesting dynamic is that Anand's review used a retail sample, and not a press sample which possibly might've been binned better...

        • (Score: 2) by fakefuck39 on Sunday March 07 2021, @08:36AM (3 children)

          by fakefuck39 (6620) on Sunday March 07 2021, @08:36AM (#1121009)

          the site i linked ran standard benchmarks and shows the opposite result. more importantly, does it make sense to you that the new and improved generation of a cpu is slower than the previous? does it make sense to you that a cpu with a smaller cache performs worse? yes, we know a larger cache increases latency slightly. and as is always the case, the speedup from the larger cache is expected to increase the overall performance.

          anandtech results don't match all the others out there - google it. they also don't make any common sense. what's interesting is people look at this one result, an uncommon and illogical result, and fight tooth and nail that it's correct, instead of the obvious -they're either lying for some reason, which they've been known to do under the new management, or they ran a bad test - also something they've been known to do under the new management.

          so take one of the rest of the tests - just from google - that shows the new cpu is faster. and examine That for the test methodology. find anything wrong? no? oh, so nothing wrong by your opinion in tests showing opposite results? then something's wrong with how you determine the validity of the test.

          • (Score: 2) by Reziac on Monday March 08 2021, @02:30AM (1 child)

            by Reziac (2489) on Monday March 08 2021, @02:30AM (#1121272) Homepage

            My cynical little voice wants to know if anandtech has an advertising contract with AMD.

            • (Score: 2, Informative) by Anonymous Coward on Monday March 08 2021, @03:15AM

              by Anonymous Coward on Monday March 08 2021, @03:15AM (#1121279)

              That's from the Intel playbook.

          • (Score: 0) by Anonymous Coward on Monday March 08 2021, @05:17PM

            by Anonymous Coward on Monday March 08 2021, @05:17PM (#1121426)

            more importantly, does it make sense to you that the new and improved generation of a cpu is slower than the previous?

            No... but can you tell we with a straight face that Pentium 4 was better than Pentium 3? And do you remember the scandal when some some Centrino CPUs were better than Pentium ones (and Intel subsequently gimped the Centrino ones to maintain market segmentation). Stuff happens.

            Frankly, given how much Intel games benchmarks, plus all the other scandals (e.g. The security flaw... rowhammer? Whatever, the one they tried to snowball the world and Linus into saying "not a bug, and even if it is, AMD has it too.), I'd trust Anandtech over them. They also have more vested interest in bias news than Anandtech does. Plus, as mentioned by others, Anandtech is a old and vaunted name in tech journalism and testing.