Rocket lake leaks

[Grim5]

Yeah looks trash wonder why they even bothered going to the trouble and expense of backporting unless they wanted to hedge bets in case Alderlake was delayed as that should be hopefully much better.

Some good prices now for the 10th gen parts recently though.

Yeah Intel tried to convince us it found a way to backport it's 10nm architectures, turns out they were full of crap. The traces are too long on 14nm, resulting in high latency = poor gaming performance.
Rocket Lake is a 10nm architecture and should have been built on 10nm, not 14nm+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

As for 10th gen, heck yeah great prices, buy it before its out of stock - Intel will want to get rid of 10th gen asap

 

[pc-guy]

Power Consumption: Hot Hot HOT - Intel Core i7-11700K Review: Blasting Off with Rocket Lake (anandtech.com)

Goodness, 104degrees on an avx512 workload with almost 300 watts on the 11700.
No need for home heating.

in all fairness that 300w is under AVX512. i dont know if anything uses that at all.

11700k is an improvement over 9900k in terms of AVX2 load and greatly better than the 10900k which sucks down 330w under the same load i believe (albeit more cores). but it is still a whopping 60% over 5800X. that is pretty bad.

https://images.anandtech.com/graphs/graph16535/121878.png

**Do Not Hotlink Images**

 

[humbug]

Don't forget that Rocket Lake was originally designed for their 10nm process, but had to be backported to 14nm. This explains the core count and clock sped regression, but it's entirely possible that the move to 14nm has just made the interconnect traces too long to operate at the speeds they were designed for. It's not the design that's the issue, it's the hatchet job moving to a chunkier node than it was intended for.

But we'll never know for sure because Rocket Lake/Cypress Cove is not getting iterated, its just getting binned off in favour of Golden Cove for Alder Lake and Sapphire Rapids.

From the other thread...

On the high Inter core latency of Rocket Lake, i've put this theory out before but i think its worth saying again.

Intel may have ditched the Ring Bus with Rocket Lake and that would explain the high Intercore latency, Sky Lake-X, Intel's HEDT line had a "Mesh" core interconnect because the CPU's were just too big for the Ring Bus, it doesn't work on a CPU of that size and if you remember Sky Lake -X was also down on performance in games compared with the 9900K.

Just when AMD have thier Intel core Latency to around 17ns Intel have had to push theirs up to around 30ns and with that the IPC gain is eaten up in gaming.

I don't think Intel were expecting AMD to improve their Intercore Latency. Certainly not by this much. AMD are up over 40% in some games vs Zen 2.

 

[Destination]

in all fairness that 300w is under AVX512. i dont know if anything uses that at all.

11700k is an improvement over 9900k in terms of AVX2 load and greatly better than the 10900k which sucks down 330w under the same load i believe (albeit more cores). but it is still a whopping 60% over 5800X. that is pretty bad.

https://images.anandtech.com/graphs/graph16535/121878.png[][/QUOTE]

True but it is nearly double the power of the ryzen 5800 in AVX2 loads. Double the wattage for poorer perf, that's absolutely bloody awful.

 

[LePhuronn]

as BIG.little...is totally new to x86

It's not totally new, Intel have already done it with Lakefield. What let Lakefield down was, as usual, the Windows scheduler in bouncing single-threaded workloads around the 4 Atom cores instead of sitting on the single big core like it was designed to do. Intel's numbers looked quite favourable when the CPU was scheduled correctly, which presumably gave them the confidence to push their big.little+Foveros concept into an actual new Core generation.

And that's still the big question mark over Alder Lake: can the OS scheduler (Windows or otherwise) properly differentiate between big and little and assign workloads accordingly?

 

[Dontrocktheboat]

in all fairness that 300w is under AVX512. i dont know if anything uses that at all.

11700k is an improvement over 9900k in terms of AVX2 load and greatly better than the 10900k which sucks down 330w under the same load i believe (albeit more cores). but it is still a whopping 60% over 5800X. that is pretty bad.

https://images.anandtech.com/graphs/graph16535/121878.png[/QUOTE]

Intel is becoming a bit of a joke lately. 14nm++++++++++++++++++ manufacturing process, is clearly not fit for purpose anymore.

 

[SKILL]

It's not totally new, Intel have already done it with Lakefield. What let Lakefield down was, as usual, the Windows scheduler in bouncing single-threaded workloads around the 4 Atom cores instead of sitting on the single big core like it was designed to do. Intel's numbers looked quite favourable when the CPU was scheduled correctly, which presumably gave them the confidence to push their big.little+Foveros concept into an actual new Core generation.

And that's still the big question mark over Alder Lake: can the OS scheduler (Windows or otherwise) properly differentiate between big and little and assign workloads accordingly?

I would suspect there will still be times when the scheduler does the wrong thing... It's taken a long time for the Linux scheduler to be updated for big.little support, MS aren't generally as responsive (for kinda obvious reasons really). It took them ages to update the scheduler to be better with the Bulldozer architecture which is a similar task really.

Alder Lake will be interesting of course, and certainly on the mobile side of things, but for the desktop I don't see it doing so well. If you've got a 16c/24t (one of the bigger ones leaked so far?) then unless the 8 'big' cores are significantly faster than Zen3 then in several tasks it will struggle to even beat that, obviously I'm talking the multithreaded tasks here. In single threaded it stands a good chance but given the leaks on Zen4 performance (29% IPC increase?) that doesn't bode well.

I say doesn't bode well, amuses me to see Intel do so badly, but as a consumer it's not good...

 

[LePhuronn]

MS didn't bother updating the scheduler for Bulldozer because they didn't see the point. Little old AMD and their rubbish CPUs with little marketshare. It literally took Zen 2's near obliteration of Intel for Windows to get a decent scheduler update. I guarantee if Alder Lake shows anywhere near its potential MS will be all over it, the Intel connection runs very deep.

 

[humbug]

MS didn't bother updating the scheduler for Bulldozer because they didn't see the point. Little old AMD and their rubbish CPUs with little marketshare. It literally took Zen 2's near obliteration of Intel for Windows to get a decent scheduler update. I guarantee if Alder Lake shows anywhere near its potential MS will be all over it, the Intel connection runs very deep.

Some Scheduling might have helped it in Single thread loads but Bulldozer was always bad anyway, 8 threads scored about 1250 points in R20, i get 640 with a single core and about 16,000 with all 8, an 8 core Zen 3 is over 12X faster, that's hilarious.

 

[Martini1991]

MS didn't bother updating the scheduler for Bulldozer because they didn't see the point. Little old AMD and their rubbish CPUs with little marketshare. It literally took Zen 2's near obliteration of Intel for Windows to get a decent scheduler update. I guarantee if Alder Lake shows anywhere near its potential MS will be all over it, the Intel connection runs very deep.

Bulldozer was crap though lol.

 

[SKILL]

I'm pretty sure, but really can't be bothered to check, that there was substantial work done by MS/AMD on the scheduler that brought decent increases to the performance from release onwards. But yes, that effort was always going to be limited due to the MS/Intel 'Wintel' association and lack of money from AMD.

And I'm not saying Bulldozer was good here, just another, and the last desktop/Windows chip that required updates to the scheduler to maximise performance. Bulldozer was an interesting (and 'viable') concept, poorly executed and utterly terribly marketed. It was never 8 core, it was essentially 4-core capable of 8-threads really only in integer workloads

Comparing it to Zen3 is lolworthy, especially in a FP benchmark, you're basically comparing a 4-core architecture from 9 years ago to an 8-core + SMT architecture from today...

 

[Grim5]

MS didn't bother updating the scheduler for Bulldozer because they didn't see the point. Little old AMD and their rubbish CPUs with little marketshare. It literally took Zen 2's near obliteration of Intel for Windows to get a decent scheduler update. I guarantee if Alder Lake shows anywhere near its potential MS will be all over it, the Intel connection runs very deep.

Updating your scheduler for more of the same cores is much much easier than updating your scheduler when some of the cores are completely different to others - and even then it took Microsoft several patches and many months to get its scheduler finally working for high core counts

 

[LePhuronn]

Bulldozer was always bad anyway

Bulldozer was crap though lol.

Um...

Little old AMD and their rubbish CPUs

Yeah, I said that

Updating your scheduler for more of the same cores is much much easier than updating your scheduler when some of the cores are completely different to others - and even then it took Microsoft several patches and many months to get its scheduler finally working for high core counts

True indeed, but I think there's more incentive for MS to put the work in because of the long-standing Intel association. Even with the success of Zen and the console connection, would MS put the effort in if it were AMD doing big.little x86?

 

[Zeeflyboy]

Even with the success of Zen and the console connection, would MS put the effort in if it were AMD doing big.little x86?

Surely purely a question of market share/user base... AMD was all but irrelevant when Windows 10 launched.

Intel still a hugely dominant force in windows install base, but AMD certainly winning back share. If it were AMD coming out with big.little in the next year or two, then yes I’d expect windows would be likely to accommodate it. Were Zen1 to have been big.little on the other hand, not a chance.

Still not convinced by the whole big.little idea outside of power sensitive mobile however... will be interesting to see how it works out on desktop. Personally I’d rather just have all big probably

 

[Dave2150]

Going to be interesting to see what you can expect with just out of the box performance vs 10th Gen, and not spending umpteen hours messing around with it, just what the 99% will see when they buy one.
Is it going to be slower than both 10th Gen and 5xxx Ryzen? That would be the important question, I mean if you need to buy special RAM and tweak it for hours who'd do that.

Anandtech's 11700k results look bugged to me. Likely beta/pre-release bios not helping, or something else going on they're not aware of. They also tested on a questionable air cooler for some reason.

We already know that stock 11900k out of the box will beat Zen 3 in gaming, I'm just looking forward to how far it can be pushed when overclocked:

Oh, and it's worth mentioning that you'll likely actually be able to buy a 11900k, since Intel have their own fabs to make huge quantities. Meanwhile, I've not once seen the 5900x in stock anywhere.

 

[deusexlumina]

Can we get a fresh shipment of kool-aid please?? I think we're about to run dry in a minute.

The release is a swing and a miss, but was always going to be. Intel are just treading water and riding the marketing machine until they get their act together and can be competitive again performance wise.

To argue anything else is just denial at this point. The processors are by and large technically inferior to the competition in almost every metric.

I hope they become competitive again soon. And I hope that they learn to be a bit more humble in the process (wishful thinking!).

 

[Minstadave]

Anandtech's 11700k results look bugged to me. Likely beta/pre-release bios not helping, or something else going on they're not aware of. They also tested on a questionable air cooler for some reason.

We already know that stock 11900k out of the box will beat Zen 3 in gaming, I'm just looking forward to how far it can be pushed when overclocked:

Oh, and it's worth mentioning that you'll likely actually be able to buy a 11900k, since Intel have their own fabs to make huge quantities. Meanwhile, I've not once seen the 5900x in stock anywhere.

Keep clutching at straws. Micro-code/BIOS isn't going to fix this and you know it. It's just not a well conceived CPU, that's obvious from the need to cut 2 cores off the 11900K just to get it feasibile.

The cooler was old but that doesn't get around the fact the CPU uses a vast amount of power for underwhelming results. Overclocking the 11900K is going to be challenging to say the least.

I wouldn't get your hopes up the 11900K is going to be winning many gaming crowns if all you're basing that on is Intel's own marketing slides either. The 11700K is soundly beaten by the 5800X unless this magical update changes things materially - which it won't.

Intel having its own fabs hasn't stopped them having serious production shortages in the last few years, but keep repeating this regardless.

 

[DeathStorm]

Oh, and it's worth mentioning that you'll likely actually be able to buy a 11900k, since Intel have their own fabs to make huge quantities. Meanwhile, I've not once seen the 5900x in stock anywhere.

Not to worry you can buy a 5800X instead as it's an outright better chip than the Intel, plenty of them available right now too.

 

[Grim5]

Anandtech's 11700k results look bugged to me. Likely beta/pre-release bios not helping, or something else going on they're not aware of. They also tested on a questionable air cooler for some reason.

We already know that stock 11900k out of the box will beat Zen 3 in gaming, I'm just looking forward to how far it can be pushed when overclocked:

Oh, and it's worth mentioning that you'll likely actually be able to buy a 11900k, since Intel have their own fabs to make huge quantities. Meanwhile, I've not once seen the 5900x in stock anywhere.

Are you seriously posting Intel's graphs as fact despite nearly every single graph they ever put out being lies? The latest being their claim that PCIE4 drives have 10% faster speeds on rocket lake compared to Zen 3, which ended up being completly wrong as benchmarks have shown (there is no difference at all)

And Hardware unboxed already confirmed 1) Anandtech's number are NOT an outlier for Rocket Lake and 2) The BIOS coming will NOT improve performance that is required to fix this

And the 5900x is not the competition for the 10900k, that's the 5800x which yes you can buy, is in stock and has been getting price reductions as well. The 5900x and 5950x destroys the 11900k, pointless to try and say the 11900k is somewhere competition to them

 

[Destination]

Anandtech's 11700k results look bugged to me. Likely beta/pre-release bios not helping, or something else going on they're not aware of. They also tested on a questionable air cooler for some reason.

We already know that stock 11900k out of the box will beat Zen 3 in gaming, I'm just looking forward to how far it can be pushed when overclocked:

Oh, and it's worth mentioning that you'll likely actually be able to buy a 11900k, since Intel have their own fabs to make huge quantities. Meanwhile, I've not once seen the 5900x in stock anywhere.

It will be interesting to see if the 11900 manages to beat the 5800, as clearly the 11700 thus far has been demonstrated not to, barely even managing to overtake it predecessor.
I think when it comes to the competition versus the 5900x if that is where it is aimed price wise, by then it looks like AMD will actually have stock, which might be simply the worst of timing for Intel.
Vastly cheaper and very available AMD motherboards combined with a processor that single thread goes toe to toe with the 11900, and multithread wipes it from the map, will make AMD the high end, productivity, gaming and indeed value checkpoint.
As I said, true competition will likely come from intel's previous generations, coupled with significant price cuts.
Unless you are specific use case, where you NEED storage speed from the main PCI slot, while offloading your GFX card to a secondary slot, whilst ignoring the M2 slots on the motherboard, and Ryan Shrouting the whole time, in that case you 'might' get some storage speed improvement.