The Intel Execution Thread [2021]

[swaaye]

So you assertion is that no one should bother buying anything higher than a 4c8t ~4.4ghz CPU because at 1440p over all benchmarked games it doesn't make any difference? Yeah let's just go back to 2015 and let Intel continue stagnating the CPU market since there's no benefit.

For games both companies have lots of CPUs that are on essentially the same level. The 1080p chart in that review is similar. Is that really stagnation in the gaming case or is it the old 'multithreading has its limits and IPC is hitting a wall' thing? It's sweet, sweet competition making everyone be quite competitive on pricing.

I think AMD getting past the Bulldozer trash and also pushing core count way up was all very smart and the only way to go. Intel responded as they do to AMD competition. Unfortunately their process problems are a thing and that probably makes it look worse than it would otherwise be over there. It seems like their 10nm plan was very aggressive and overly optimistic.

 

[Rootax]

Fully expect ADL to lose to Zen3 while launching against Zen4 at this point. Intel seem to be unable to make good CPU these days.

What ? They have a process problem, but the architectures they're making seems fine to me.

 

[techuse]

For games both companies have lots of CPUs that are on essentially the same level. The 1080p chart in that review is similar. Is that really stagnation in the gaming case or is it the old 'multithreading has its limits and IPC is hitting a wall' thing? It's sweet, sweet competition making everyone be quite competitive on pricing.

I think AMD getting past the Bulldozer trash and also pushing core count way up was all very smart and the only way to go. Intel responded as they do to AMD competition. Unfortunately their process problems are a thing and that probably makes it look worse than it would otherwise be over there. It seems like their 10nm plan was very aggressive and overly optimistic.

For high refresh rate gaming and keeping minimum framerates up CPU matters a lot. There are also very CPU heavy titles/areas that TPU doesn't test.

 

[Kaotik]

Intel Core i7-11700K Review: Blasting Off with Rocket Lake

That's way worse than anyone was expecting. Wtf is Intel doing these days?

HardwareLuxx's results are pretty much in line with what we've gotten with our store-bought 11700K
https://www.hardwareluxx.de/index.p...-lake-s-als-core-i7-11700k-im-vorab-test.html
So there seems to be something wrong with Anandtechs results, they're worse than they should be. Most likely whatever motherboard/bios-combo they're using isn't boosting the CPU right or something, or at least I can't figure any other reason for 5% discrepancy in single core Cinebench R20 results (no, it's not sensitive to memory speed/settings)

 

[DegustatoR]

HardwareLuxx's results are pretty much in line with what we've gotten with our store-bought 11700K
https://www.hardwareluxx.de/index.p...-lake-s-als-core-i7-11700k-im-vorab-test.html
So there seems to be something wrong with Anandtechs results, they're worse than they should be. Most likely whatever motherboard/bios-combo they're using isn't boosting the CPU right or something, or at least I can't figure any other reason for 5% discrepancy in single core Cinebench R20 results (no, it's not sensitive to memory speed/settings)

There were some rumblings about them running the RAM at 2:1 ratio for some reason.
Will be interesting to see if the results will be better at official launch.
RKL as a product make sense only if it will be on par with Zen3 in parts with the same number of cores and for that it needs to add some 10-20% of per core performance to CML predecessors.
Because if it won't then there's CML already which is cheaper to produce and is likely less hot / power hungry.

 

[Kaotik]

There were some rumblings about them running the RAM at 2:1 ratio for some reason.

Yeah, but it still doesn't explain it, I couldn't find results for previous gen Intels, but for Ryzen 3900X the differences between single channel DDR4-3200 to dual channel DDR4-2400 to DC DDR4-4000 and everything in between fit in the same test within 10 points of each other, with no logical order suggesting run-to-run variation rather than anything else affecting the results.
And the "some reason" is that they run the CPU according to it's specs, only i9-11900K(F) support DDR4-3200 1:1, the rest are 2:1 3200 and 1:1 2933, but that wasn't apparently public information yet (it got out in the last set of Rocket Lake slides to leak)

Will be interesting to see if the results will be better at official launch.
RKL as a product make sense only if it will be on par with Zen3 in parts with the same number of cores and for that it needs to add some 10-20% of per core performance to CML predecessors.
Because if it won't then there's CML already which is cheaper to produce and is likely less hot / power hungry.

I'd put my money on retail launch results pretty much mirroring HWLuxx's and our numbers.

 

[DegustatoR]

Yeah, but it still doesn't explain it,

Some microcode and/or BIOS fuckery likely.

I'd put my money on retail launch results pretty much mirroring HWLuxx's and our numbers.

This would put RKL as a direct competitor to 5600X and 5800X which could only be a good thing nowadays considering that Intel has a separate and substantial production capacity.

 

[DegustatoR]

Seem to be a confirmation of Anandtech data mostly.
A complete waste of effort.
No idea why Intel thought that this is a product worth launching.

 

[Malo]

Rather embarrassing again for Intel. When your previous generation is as fast or faster then your release is just inertia, something that is a "new" product with a more recent product number. Meaningful only for some OEMs.

 

[DegustatoR]

Rather embarrassing again for Intel. When your previous generation is as fast or faster then your release is just inertia, something that is a "new" product with a more recent product number. Meaningful only for some OEMs.

With RKL being some 25% larger than CML for similarly performing parts it is way worse than that. The only slightly redeeming side of it is the PCIE4 support I guess. But still this seems like a waste on all parts considering that ADL will supposedly come in less than a year from now.

 

[Kaotik]

With RKL being some 25% larger than CML for similarly performing parts it is way worse than that. The only slightly redeeming side of it is the PCIE4 support I guess. But still this seems like a waste on all parts considering that ADL will supposedly come in less than a year from now.

And AVX-512 finally on consumer desktop platform. Though, given what a mess AVX-512 is with pretty much every product supporting different subsets of instructions, not sure how much it matters.
PCIe 4.0 support was supposed to come with CML already, but they apparently couldn't get it working stable on it.

 

[Deleted member 13524]

Rather embarrassing again for Intel. When your previous generation is as fast or faster then your release is just inertia, something that is a "new" product with a more recent product number. Meaningful only for some OEMs.

I don't know.. if they have actual stock on their 6-8 core parts then it seems pretty meaningful to me, all else considered.

And AVX-512 finally on consumer desktop platform. Though, given what a mess AVX-512 is with pretty much every product supporting different subsets of instructions, not sure how much it matters.

Honest question: who/what really needs AVX512?
I know that up until half a dozen years ago some people thought it would revolutionize the world of computing and make GPUs irrelevant for real time rendering, though I guess that was before we knew it would become a power hog and require massive frequency throttling to work.

 

[DegustatoR]

https://videocardz.com/newz/intels-first-7nm-client-architecture-meteor-lake-to-launch-in-2023

 

[Bondrewd]

https://www.anandtech.com/show/1657...ndry-services-ibm-collaboration-return-of-idf

Pat G fucks and you should hear him talk, too.

 

[DegustatoR]

• Intel to Revive ‘Tick-Tock’ Model, Unquestioned CPU Leadership Performance in 2024/2025
• Intel’s x86 Designs No Longer Limited to Intel on Intel: IP Blocks for Foundry, Cores on TSMC

 

[DegustatoR]

RKL reviews are out and it's still a weak sauce: https://videocardz.com/107476/intel-11th-gen-core-rocket-lake-s-review-roundup

 

[Rootax]

Yep. I just think they wanted to be on the front page again, and if RKL are available (vs the zen 3 situation), they will sell very well.

 

[DegustatoR]

I mean, 11400 to 11700 can be fine products depending on the prices. They are able to compete with similarly "cored" Ryzens, including those based on Zen3.
11900K doesn't make any sense whatsoever.

 

[Bondrewd]

and if RKL are available (vs the zen 3 situation)

The issue at large is RKL not competing with actually unavailable higher end Z3 SKUs.
Even the best RKL is a bad 5800X.

I mean, 11400 to 11700 can be fine products depending on the prices.

No they're all shit, buy discounted CML while it lasts and spend whatever money you save on fancy DRAMs (which CML can actually run, yea) or shell out for Z3.

 

[DegustatoR]

No they're all shit

No, they are not. 11600K is more often on par with 5600X than you'd think while being cheaper for example. These are good products - if they'll be available at these prices.