Rocket Lake Intel Core i9-11900K slips past AMD Ryzen 9 5950X in single-thread benchmark by 2.7-5.7% - just don't mention multithread performance, power, or delays

[Eliezer May]

1.  Techradar has a similar article on the same subject by Bill Thomas except that that site does not allow any form of feedback.  I strongly advise my readers not to pay any attention to any sites that do not permit readers to voice disagreement with authors.  At least this site apparently does allow feedback.
2.  Cpu-z is a questionable benchmark and one should research the history and testing methods before giving it too much credibility.  There are other more widely used benchmarks such as Geekbench and Cinebench.
3.  In general it is a good idea to use a reliable benchmark over making the mistake of relying on overclock speeds.  First of all overclock are simple short bursts of extra adrenalin and not sustained performance.  Secondly it is not clock alone but clock x IPC (Intructions Per Clock) that measure work performed.  And even then there are specialized subset of instructions that can accelerate specialized work flows.  And there are many architectural factors as well.  So the best way to get a truer picture is to use a well constructed unbiased benchmark.  Unfortunately history shows that Intel financed the sabotage of AMD code numerous times even as late as 2019 although it started much earlier.  Then there were instances where Intel paid benchmark companies to skew result in their favor.  And we must also remember Intel's contra revenue practices.  In short one must understand the technology changes in the last 5 years along with Intel's nefarious business practices.

[vertigo]

1.  Techradar has a similar article on the same subject by Bill Thomas except that that site does not allow any form of feedback.  I strongly advise my readers not to pay any attention to any sites that do not permit readers to voice disagreement with authors.  At least this site apparently does allow feedback.
2.  Cpu-z is a questionable benchmark and one should research the history and testing methods before giving it too much credibility.  There are other more widely used benchmarks such as Geekbench and Cinebench.
3.  In general it is a good idea to use a reliable benchmark over making the mistake of relying on overclock speeds.  First of all overclock are simple short bursts of extra adrenalin and not sustained performance.  Secondly it is not clock alone but clock x IPC (Intructions Per Clock) that measure work performed.  And even then there are specialized subset of instructions that can accelerate specialized work flows.  And there are many architectural factors as well.  So the best way to get a truer picture is to use a well constructed unbiased benchmark.  Unfortunately history shows that Intel financed the sabotage of AMD code numerous times even as late as 2019 although it started much earlier.  Then there were instances where Intel paid benchmark companies to skew result in their favor.  And we must also remember Intel's contra revenue practices.  In short one must understand the technology changes in the last 5 years along with Intel's nefarious business practices.

100% agree about sites with no comment sections, and I either avoid them or spend very little time on them.

Also agree completely about Intel's anti-competitive business practices, which is just one of many reasons I prefer AMD, and now that AMD has closed the gap between them, I would personally pay more (to an extent) for an AMD chip than a similarly performing Intel one. But AMD is of course not perfect, either. I went with Intel over AMD despite it costing more when I upgraded my desktop a few years ago because AMD had a bug related to virtualization that they weren't fixing, and I have no idea if they've resolved it yet. And AMD and Nvidia have (allegedly) conspired to price fix GPUs.

[hfm]

Well, seriously? For the past 5 years since Sky-lake they say the same thing - "oh, this generation is just a stop-gap to the next one which will be amazing!" Like, 5 generations of saying that is astonishing at how bad intel reacts to Ryzen and how bad the situation goes within themselves.
Am I the only one who noticed this trend with intel? Saying every year that this one is just a stop-gap to the next one and that's why it's not that good?

Their real problem has been manufacturing. They've really fallen a lot behind their plans. If they struggle to match TSMC, it's difficult to compete with AMD. It used to be the other way around with TSMC chasing Intel. That has nothing to do with Ryzen. AMD just got lucky that they launched an offensive when Intel is struggling so much with getting their designs manufactured as planned. Processors are designed counting on the ability to deliver, to manufacture the design. You can't just take the design and make it on an older node. Their manufacturing problems mean they can't manufacture what they designed, what they planned to release. Instead, they have to improvise improvements of older designs or try to back-port the design to an older node, but that's potentially very difficult and won't be as good. Under these conditions, it's difficult to show good progress.

Intel seems less keen on pushing the core count, at least for consumer processors. If Ryzen for you means pushing quite a lot of cores, it looks like Intel isn't that interested. Only market can decide which way is preferable (it's not a question of what is better - which is workload specific - but what gets bought). An 8 core processor shouldn't be able to compete with a 16 core processor intended for the same segment (just as a 4 core processor shouldn't be able to compete with an 8 core processor). The coming hybrid desktop processors might suggest that Intel might be interested in going for more cores (perhaps that market has spoken), but might be struggling with manufacturing capacity and yields which make it not viable.

It seems like it isn't that Intel is less keen to focus on core count, it's that their manufacturing and architecture doesn't lend itself to executing on it. Can you imagine the TDP of something competing with 5950x in multi core? This single digit lead in single core perf is meaningless in all but edge cases.