"Coyote Cove" does not feature HyperThreading, something the Intel CEO wants to see restored on the company's P-cores, but "Nova Lake" will make up for it with a numerical increase in P-core counts.

I think Tan doesn't know what he's talking about, or at least doesn't realize how gamers are interpreting this.

I don't think hyperthreading is coming back to consumer P cores. Intel has always said that the "Rapids" Xeon chips will have P cores with hyperthreading, but hasn't said "Lake" chips will. Why would they? For Meteor Lake, the scheduler:

1. Tries the LPE cores
2. If big, moves to the P cores
3. If backgrounded or P cores are busy, moves to E cores
4. If all E cores are busy, moves to virtual P cores (hyperthreading)

In this paradigm, virtual E cores would make more sense to the scheduler than virtual P cores.

dyonoctisAMD's old spun-off foundries are irrelevant; they don't qualify them for getting money from the chip act, whose main purpose is to boost manufacturing sovereignty. Both the US and the EU made that act after seeing how bad relying on other faraway countries could be during the pandemic

Oh, really?

You took a single comment I made out of context.

Here you go: www.techpowerup.com/forums/threads/intel-core-ultra-series-4-nova-lake-processors-launch-in-late-2026.345539/post-5662396

dyonoctisgetting money from the chip act, whose main purpose is to boost manufacturing sovereignty

I really wish it was, because Intel received $8.9B, but delayed one of their main fabs being built in Ohio to 2030, and now they're also deciding to focus away from the consumer market to prioritize AI instead.

Hecate91I really wish it was, because Intel received $8.9B, but delayed one of their main fabs being built in Ohio to 2030, and now they're also deciding to focus away from the consumer market to prioritize AI instead.

There is a silver lining. The cryptocrazes, the lockdown server boom, and the AI boom mean a lot of money is flowing into chip development, including R&D for more advanced nodes. On the other side of the boom when attention turns back to consumer, the foundry tech will be more advanced than it would have been otherwise.

igormpNo, they had proper AVX2 support, and have had since Gracemont (Alder Lake). They lacked AVX-512, which is still not present in any current E-core offering. Lion cove did have AVX-512 at first but Intel disabled it due to the disparity with the E-cores.

They have always had OOE since Silvermont in the Atom era.

Not really. Since Meteor Lake the LPE and E-cores have the exact same µarch (Crestmont). The difference was the node they're fabbed at, with the E-cores usually on the compute tile along with P-cores on a better node while the LPE-cores get shunned to the SoC tile with worse node and slower clocks, and the lack of L3 on the LPE cores. The only exception to that was Arrow Lake mobile, which re-used the SoC from MTL, so the E-cores were Skymont but the LPE-cores were Crestmont
However, Panther Lake will share the µarch, with both E and LPE-cores being Darkmont and both living in the compute tile. The difference is just that the LPE ones will clock lower and won't have an L3.
I assume Nova Lake will be similar to MTL, using Arctic wolf cores for both E and LPE, but with the LPE-cores being shunned to the SoC tile with lower clocks and no L3.

Ok, that's nice to know, thanks.

rattlehead99What benefit do you get from having more than 8P cores that is not a productive/professional workload?
Also their new E-Cores are on par with the last gen P-Cores in terms of IPC, and the clocks seem to be going up as well and now E-Cores will have the same instructions as the P-Cores.
If anything the P-Cores seem to be there just to top benchmark charts.

OK... that will be cool. e-cores that are not brain-dead would be a really really good thing. P-cores with a bunch of stuff on them that i never or rarely use is only ok because they run faster than the crippled e-cores. There is absolutely no reason to have slow e-cores especially if you have a few LP-cores.

ThomasKOh, really?

You took a single comment I made out of context.

Here you go: www.techpowerup.com/forums/threads/intel-core-ultra-series-4-nova-lake-processors-launch-in-late-2026.345539/post-5662396

I did not. The answer that you gave to FeelinFroggy in the context of Intel getting 5.7 billion from the chips act didn't refute his argument that if AMD was involved in manufacturing, they would have gotten (more) gov funds as well. Case in point, Global Foundries got 1.5 billion. Even Texas Instruments got 1.6 billion. The chip act wasn't a thing before 2022.

Before that, it's important to remember that Intel is also contracted by the government, being the only US-based bleeding-edge foundry. Global Foundries also got various federal funds and contracts for the military. The government gives money to companies that can manufacture goods for them locally. AMD got gov funds to build a US-based AI datacenter for the Department of Energy. They don't care much about the contribution those companies make to consumers. They can keep using China, Korea, and Taiwan-based stuff for all they care, but the military? They prefer local stuff.

I'm not debating about whether or not Intel ethically deserved those funds; the US gov do not care, they want power and sovereignty above all.

dyonoctisI did not. The answer that you gave to FeelinFroggy in the context of Intel getting 5.7 billion from the chips act didn't refute his argument that if AMD was involved in manufacturing, they would have gotten (more) gov funds as well. Case in point, Global Foundries got 1.5 billion. Even Texas Instruments got 1.6 billion. The chip act wasn't a thing before 2022.

Before that, it's important to remember that Intel is also contracted by the government, being the only US-based bleeding-edge foundry. Global Foundries also got various federal funds and contracts for the military. The government gives money to companies that can manufacture goods for them locally. AMD got gov funds to build a US-based AI datacenter for the Department of Energy. They don't care much about the contribution those companies make to consumers. They can keep using China, Korea, and Taiwan-based stuff for all they care, but the military? They prefer local stuff.

I'm not debating about whether or not Intel ethically deserved those funds; the US gov do not care, they want power and sovereignty above all.

All of that has absolutely nothing to do with what I said beforehand. Your answer is out of context, once again. And I couldn't care less how much GloFo got from all of that.

What I said is that Amd didn't get a bailout from government WHILE it was struggling and OWNED its own fabs. In contrast to what's currently happing to Intel, which not only got the biggest share of the pie from the chips act, for obvious reasons, as much as government trust, in eventually converting these subsidies into 10% equity ownership, while boosting its stock price, and making that very same government Intel's biggest shareholder.

That's why I said Intel is essentially Amd from the past, apart from government help. Intel is struggling now and Amd was back then. It's a point of comparison and a meme, what's so hard to comprehend about that? You're grasping at straws.

32 useless cinebench accelerator e-cores? This doesn’t sound optimistic.

ThomasKAll of that has absolutely nothing to do with what I said beforehand. Your answer is out of context, once again. And I couldn't care less how much GloFo got from all of that.

What I said is that Amd didn't get a bailout from government WHILE it was struggling and OWNED its own fabs. In contrast to what's currently happing to Intel, which not only got the biggest share of the pie from the chips act, for obvious reasons, as much as government trust, in eventually converting these subsidies into 10% equity ownership, while boosting its stock price, and making that very same government Intel's biggest shareholder.

That's why I said Intel is essentially Amd from the past, apart from government help. Intel is struggling now and Amd was back then. It's a point of comparison and a meme, what's so hard to comprehend about that? You're grasping at straws.

I'm a bit autistic, and can get hung up on details like that, haha. My reasoning went beyond the meme: I got the "moar core" comparison, but was bothered about the bailout part because the eras seem too different to be compared. Looking at the reaction of other people, I'm not the only one who read that as a means to showcase the unfairness of one situation vs the other.

Chips are far more ubiquitous, Glo fo bailed on bleeding edge, and the sole US bleeding edge maker was about to go down and leave the US orphan and heavily dependant from Taiwan. That's more dire than circa 2009 when the best US foundry was healthy and running full steam; the overall stakes were lower, and the operating cost less daunting. Sure, there might be a bit of market manipulation, but in the 2020's they also barely had a choice but to save that local manufacturing sinking ship.

Another intel, another mobo..

Seems though that lga 1700 will keep going alongside lga 1954.
1851 is doomed to oblivion, like some bad windows iteration..

rattlehead99What benefit do you get from having more than 8P cores that is not a productive/professional workload?
Also their new E-Cores are on par with the last gen P-Cores in terms of IPC, and the clocks seem to be going up as well and now E-Cores will have the same instructions as the P-Cores.
If anything the P-Cores seem to be there just to top benchmark charts.

The latest game engines can easily use up to 32 CPU cores with significant performance improvements. We all know how laggy the UE5 engine is. However the more stronger CPU cores, the fastest is runing.

rattlehead99What benefit do you get from having more than 8P cores that is not a productive/professional workload?
Also their new E-Cores are on par with the last gen P-Cores in terms of IPC, and the clocks seem to be going up as well and now E-Cores will have the same instructions as the P-Cores.
If anything the P-Cores seem to be there just to top benchmark charts.

Is the scheduler any better now? With the way Win11 is going I figure its not.

dgianstefani16+32 = 48, not 64.

The IGPU of modern Intel CPUs is also incredibly useful, and the E cores add more MT more efficiently than a quarter of the number of P cores would.

I thought HT was coming back?

It's promising. Especially the big cache.
But why would Intel think they need 48 cores to compete with AMD's 24 cores? Especially when they should both be made on TSMC's latest process. AMD did it with Zen 2 because they had a process lead (so they could do it in smaller area) and because they had a ST performance deficit. I worry the latter reasoning may be behind Intel's decision here.
And, like how AMD's best selling parts are the 1 CCD variants, it'll probably be the 8+16 variant that is the most appealing.

UpgrayeddI thought HT was coming back?

Not in time for Nova Lake.

ThomasKAll of that has absolutely nothing to do with what I said beforehand. Your answer is out of context, once again. And I couldn't care less how much GloFo got from all of that.

What I said is that Amd didn't get a bailout from government WHILE it was struggling and OWNED its own fabs. In contrast to what's currently happing to Intel, which not only got the biggest share of the pie from the chips act, for obvious reasons, as much as government trust, in eventually converting these subsidies into 10% equity ownership, while boosting its stock price, and making that very same government Intel's biggest shareholder.

That's why I said Intel is essentially Amd from the past, apart from government help. Intel is struggling now and Amd was back then. It's a point of comparison and a meme, what's so hard to comprehend about that? You're grasping at straws.

Intel at its lowest has never been in the same dire straits that AMD found itself in the past, especially at its pre-Zen crash, when stock suffered a near total collapse becoming "dollar stock" by mid-2015 ($1.61 historic low). The company was largely mismanaged by its former CEOs, and while I may not agree with every decision Lisa Su takes (such as prioritizing the CPU business over graphics), she has done well by AMD's shareholders.

I also disagree with the lawsuit being what "saved" AMD. If anything, it's a pittance in the grand scheme of things. It was their reckless, poorly thought out acquisition of ATI. They paid out many times what ATI was worth in 2005, but hindsight is 20/20 - it largely kept AMD afloat thanks to IP deals (console revenue streams were essential at this time) when all of their other businesses, including their core business of designing processors, failed.

www.cbc.ca/news/science/amd-says-it-overpaid-for-ati-1.641782

If Zen hadn't worked out, AMD would have gone bankrupt and completely liquidated a few years ago. By the time they launched the original Ryzen, AMD was running on fumes and on its last stand, without having any major importance to the market at the time and no exceptional quality product to even justify government intervention. They bet the whole farm on it, and pulled it off. Their IP's would probably have ended up with Qualcomm, some maybe even with Intel. The AMD of the mid 2010's was a zombie of a company, nothing like it is today, and Lisa did turn that ship around.


Government interests were different 20 years ago, CEOs were different 20 years ago, the whole industry was different 20 years ago. Such comparisons are emotional, naïve and short-sighted, and of course AMD isn't getting bail money: they're not in need of it, and as a fabless chip design company, their real value and relevance to national security interests lie in their intellectual property and detention of technology.

phints32 useless cinebench accelerator e-cores? This doesn’t sound optimistic.

Useless cinebench accelerator Arctic Wolf cores with a full ISA, AVX-512 support that outperform Zen 4, yes. At least on paper.

JohHIt's promising. Especially the big cache.
But why would Intel think they need 48 cores to compete with AMD's 24 cores? Especially when they should both be made on TSMC's latest process. AMD did it with Zen 2 because they had a process lead (so they could do it in smaller area) and because they had a ST performance deficit. I worry the latter reasoning may be behind Intel's decision here.
And, like how AMD's best selling parts are the 1 CCD variants, it'll probably be the 8+16 variant that is the most appealing.

8+16 is going to be Core 3/5 tier, if leaks are to be believed. This may just be the case of another generational leap in core count, like when we went from 4 to 8, and eventually 16 core CPUs being common at the high end. Probably minimal impact at best for current day workloads.

UpgrayeddI thought HT was coming back?

Maybe not that fast, it was said in earlier rumors there wont be any HT until Titan Lake.

Who wants a 300W+ chip? Intel better limit these to 200W.
It is bad for the environment.

Dr. DroUseless 32 cinebench accelerator Arctic Wolf cores with a full ISA, AVX-512 support that outperform Zen 4, yes. At least on paper.

I wont believe without seeing.
I would like to see they perform well.
But I wont hold my breath.

PS: I do rendering.

melkor_unlimitedI wont believe without seeing.
I would like to see they perform well.
But I wont hold my breath.

PS: I do rendering.

The 285K's Skymont E-cores are already a massive improvement over Gracemont. MT is an area where Intel doesn't have any difficulties, really.

The 16P+32E cores should be the maximum config, since its modular they could scale it down. Big Last level cache sounds interesting, so does Xe3 iGPU. Its a shame it needs new socket, typical Intel.

LGA 1954, and the mobile is FCBGA1964. Very similar. Does this mean that the PCH is finally no longer soldered to the motherboard?

Dr. DroThe 285K's Skymont E-cores are already a massive improvement over Gracemont. MT is an area where Intel doesn't have any difficulties, really.

I am not question performance of e-cores (or, at least, not so much). But mixing different cores with unpredictable impact on overall perfomance, latencies, drop-outs, etc. And microsoft scheduler does not help at all.

Offer me passive cooled ntb with 4 or 8 Arctic Wolves drawing <15W without Premium Price and I will take one. Or three.
For rendering I would prefer something that can run 24x7x365 without noise of starting Jet.

rattlehead99What benefit do you get from having more than 8P cores that is not a productive/professional workload?
Also their new E-Cores are on par with the last gen P-Cores in terms of IPC, and the clocks seem to be going up as well and now E-Cores will have the same instructions as the P-Cores.
If anything the P-Cores seem to be there just to top benchmark charts.

Are people gonna repeat this a hundred times to those who won't listen?

At this point people still believe current E cores are for background tasks despite Intel saying they have taken over the majority of workloads.

We can only pressure the big reviewers at launch to have some P core vs E core only and vs AMD normal and dense cores to lay the debate about actual performance of E cores to rest

MufasaVWe can only pressure the big reviewers at launch to have some P core vs E core only and vs AMD normal and dense cores to lay the debate about actual performance of E cores to rest

Pressure? Lol. You need a reviewer to be personally interested in this as it's very time consuming.
If you read the last paragraph in the Arrow Lake reviews you'll see that we were supposed to get it, but then we didn't.
www.techpowerup.com/review/intel-core-ultra-9-285k/33.html
I proposed this exact thing a while ago, and it did not seem to me that people are interested not even for academic purposes.

Sol_Badguy in [Request] Flagship CPU battle royale in various core configsHello. Firstly I know how there's never enough time, I won't dwell on it. So with that out of the way, I'm kindly requesting and hoping there's a chance this could be squeezed in the workflow/review s…

Public demand is definitely an incentive that can convince a reviewer that his effort will get the necessary views/clicks etc. to justify the effort, but if many consumers have negative sentiments towards Intel that are mainly stemming from the belief that the E-cores are weak (alongside the lackluster gaming performance) then the interest isn't really there, only the informed have that interest but the informed consumers are the minority. It's a vicious cycle, however it can be broken if such a review is made and the wrong beliefs are corrected by facts.

Maybe we'll get it for Nova Lake if Intel at the very least matches AMD. Well actually maybe not because if Nova Lake will have a warm reception then people will simply buy it and not stop to hate on it for being loaded with "weak" E-Cores.

melkor_unlimitedBut mixing different cores with unpredictable impact on overall perfomance, latencies, drop-outs, etc.

You have just described HT. Which also means mixing two types of cores: fast ones running one thread and slow ones running two, even if the "type of core" constantly changes.
As a replacement for hyperthreading, E cores are great.
Then again, they might have their own scheduling issues. If all four cores in a cluster are given hard work, what's the performance drop? Shared L2 seems like a bottleneck but if the cores execute threads belonging to same process, meaning shared memory, it might not be true.

melkor_unlimitedAnd microsoft scheduler does not help at all.

Yes, that's the baddest problem.
There's one thing I still don't understand: can a game direct the scheduler to use certain cores for certain threads or not? Without any additional system utilities, drivers, OS settings?

Sol_BadguyI proposed this exact thing a while ago, and it did not seem to me that people are interested not even for academic purposes.

Very few reviewers/analysts do testing just for science, unfortunately, and TPU is not one of them. W1zzard does testing with more practical value, such as memory scaling and PCIe scaling.