Intel Nova Lake (16th gen) on next gen platform/socket (LGA-XXXX)

[g67575]

Intel talking about the ‘innovation’ of putting the CPU on top of a cache die in 2023

Intel has no plans to use a 3D V-cache-like technology on the company’s Meteor Lake processors, which start rolling out in December. “But in our roadmap, you’re going to see this idea of the 3D silicon, where we’ll have cache on one die, right? And then we’ll have CPU compute on a stacked die, on top of it,” Gelsinger said.

That sounds different than AMD, which has been using the chip-stacking tech to place added memory on top of its CPU dies. According to Gelsinger, Intel wants to do the opposite and stack the CPU on top of the memory.

Although Intel’s CEO was light on details, he said the company plans on connecting the chip dies vertically using its EMIB and Foveros processes, allowing the silicon to communicate together in a single package.

Intel: Our 3D V-Cache Tech Will Take a Different Approach Than AMD

The company plans to use the chip-stacking tech for Intel processors and offer it to third-party customers contracting with Intel's foundry business.

uk.pcmag.com

Bad luck, Intel.

From what they said about Arrow Lake before launch, I think they acknowledged the advantage of AMD’s vcache. They must know that they are in deep trouble, without their own version of the tech, to compete. I think we could see it on roadmaps soon, as a refresh, or as part of the same series as Arrow Lake.

EMIB and Foveros are technologies already used in Arrow Lake, as discussed here:
https://www.muycomputer.com/2024/10...k-a-vista-de-silicio-todo-lo-que-debes-saber/

 

[SpellowHouse]

Well, DUH! Arrowlake Refresh would be like fitting new carpets in the Titanic.

It's not a break from tradition, it's the realisation that they released a piece of ****.

 

[SpellowHouse]

I don’t see why they would abandon their new LGA 1851 socket any time soon.

What would be the point? They have a decent DDR5 platform now.

Should be good until DDR6 becomes a thing.

AM5 will be alive until 2027, and still using DDR5.

So, I don’t think there will be much demand for DDR6 until late 2027 or 2028.

Maybe DDR6 will be used on workstation boards/ servers for a while, where the bandwidth might be more in demand.

Memory controller tech needs to catch up first, or we’ll be running at gear 4 or 8 in the future

The rumour mill rarely gets it right on socket longevity.

I don't know for sure, and I am not well read on the matter, but what I have seen indicates that LGA 1851 may change for 16th gen.

 

[Dave2150]

Nothing set in stone yet, though the Intel roadmap and focus for the next CPU line up suggests against there being any substantial updates on LGA1851 with the Arrow Lake successor for the same platform cancelled and the features for the next focus almost certainly requiring a new socket.

It is set in stone - Nova Lake physically cannot work on LGA-1851. These new generations and architectures are set in stone years ahead of launch.

 

[Dave2150]

Well, DUH! Arrowlake Refresh would be like fitting new carpets in the Titanic.

It's not a break from tradition, it's the realisation that they released a piece of ****.

I agree, Arrow Lake refresh will be pointless - but Intel will be forced to do it as it's going to be too long a wait for Nova Lake in 2026.

If Arrow Lake refresh (+200Mhz perhaps) gets them 1% closer to Zen5 X3D in benchmarks, they'll do it.

 

[g67575]

I think if they can solve the latency issues for the 16th gen, it would make a significant difference. Most reviewers commented on the performance regression in games.

Combine that with Intel’s own version of vcache, and I think they’d have a worthy series of CPUs.

 

[g67575]

It is set in stone - Nova Lake physically cannot work on LGA-1851.

How do you know for sure future generations will use a new type of CPU socket?

 

[adamsleath]

they dont know.
most rumours mention it is likely novalake and razer wtf-lake will be a new socket.

amd mentioned they'll support AM5 socket until 2027 or somesuch. for those who want to flip cpu's every generation. lolz

 

[pvp-forlife]

they dont know.
most rumours mention it is likely novalake and razer wtf-lake will be a new socket.

Intel not saying anything to their community about the longevity of 1851 is slimy and they are deliberately causing confusion and miss-trust which will back-fire on them big time

 

[adamsleath]

Slime-lake. just add it to the AMD propaganda files.

 

[KompuKare]

Since Intel now have years of history of extremely short-lived sockets, I think we can automatically assume 1851 is ARL only and Nova Lake will be on a new socket.

LGA 1851 was supposed to have been Meteor Lake and Arrow Lake but since MTL was cancelled on desktop therefore it is a one generation socket. No matter how expensive LGA 1851 motherboard are.

If Intel ever want to change they would really have to very clear and shout about the longevity of any such new sockets.

 

[g67575]

Since Intel now have years of history of extremely short-lived sockets, I think we can automatically assume 1851 is ARL only and Nova Lake will be on a new socket.

Why? LGA 1700 lasted 3 years / for 3 generations (even if they were a refresh of the 12th). They might even release more CPUs for this socket, especially because they know lots of people are still buying AM4 CPUs....

I think 3 years is a good bet, so end of 2027. The socket itself is a decent bump up in pin count from the previous socket.

 

[g67575]

I have a question about leaks, how come it's always the far off nebulous 'uber Lake' being predicted, but never the next generation. They didn't have a clue about Meteor Lake, or Lunar Lake...

If the rumour mill hears nothing, they just say it doesn't exist (like Arrow Lake refresh for example).

There's been quite a bit of contradictory information, from Intel (as well as leakers). For example, is the 20a process 'manufacturing ready', or is it completely cancelled?

Why does Intel think 18a is so brilliant, but 20a is something to be forgotten?

 

[Rroff]

It is set in stone - Nova Lake physically cannot work on LGA-1851. These new generations and architectures are set in stone years ahead of launch.

Intel has loads of projects on the go and sometimes repurposes designs for other lines, etc. over the last few years they've "banked" loads of R&D so as to trickle stuff out to customers, so I wouldn't be surprised by anything at this point.

 

[WARburton]

• Shorter LGA1851 Socket Lifespan: Arrow Lake might be the only CPU generation to use this socket, which could impact upgrade plans for users who invested in the new platform.

that's funny as hell, if they added HT back i may die from laughter

 

[WARburton]

grfx is where the punch is for Most games. there are exceptions.

yer and no, most competitive games still use 1080p at 300+FPS this is all CPU

 

[Rroff]

that's funny as hell, if they added HT back i may die from laughter

The chipset/BIOS still has HT support so entirely possible.

 

[Grim5]

I have a question about leaks, how come it's always the far off nebulous 'uber Lake' being predicted, but never the next generation. They didn't have a clue about Meteor Lake, or Lunar Lake...

If the rumour mill hears nothing, they just say it doesn't exist (like Arrow Lake refresh for example).

There's been quite a bit of contradictory information, from Intel (as well as leakers). For example, is the 20a process 'manufacturing ready', or is it completely cancelled?

Why does Intel think 18a is so brilliant, but 20a is something to be forgotten?

To be fair it seems like process node density changes on paper are not the only thing that affects how well a process node works for a product.

For example, we've been told that SRAM scaling is dead because there was no improvement in density or efficiency when attempting to manufacture SRAM on TSMC's 6nm, 5nm, 4nm and 3nm nodes.

But today TSMC announced that it's seeing a significant improvement in density and efficiency with SRAM production on its test 2nm node using GAA nanosheet transistors. They also say they've got the yields now up to 80% for 256Mb SRAM chips, which could be used as cache for CPU's and GPU's etc.

As a comparison, AMD is still using an older node for it's cache and currently on Ryzen 3D cache CPU's, the density is 14Mb/mm2. TSMC's new GAA 2nm SRAM has a density of 38Mb/mm2

 

[g67575]

I think whatever architecture is planned next (codename not really important), it will be on 18A-P, not the current version of the Intel 18A process:

Roadmap is vague as hell, so probably will be 2026 for the 18A-P CPU (desktop) series.

The 18A process will probably just be for mobile CPUs, maybe in late 2025.

Looking at comments on the 20A process, it seems they didn't manage to scale the 20A process very well, due to yield quality issues:

Focusing resources on Intel 18A also helps us optimize our engineering investments. When we set out to build Intel 20A, we anticipated lessons learned on Intel 20A yield quality would be part of the bridge to Intel 18A. But with current Intel 18A defect density already at D0 <0.40, the economics are right for us to make the transition now.

https://www.intel.com/content/www/us/en/newsroom/opinion/continued-momentum-intel-18a.html#gs.hhk9at

It's written in investor friendly language, so they just say it's about 'lessons learned'.

 

[adamsleath]

yer and no, most competitive games still use 1080p at 300+FPS this is all CPU

ill leave them to it then.

i found i could still slaughter the noobs at 60 fps XD

what matters more is server Ping , in my opinion.

==

Gaming PC Market Size & Share, Growth Trends 2036