Tuesday, April 28th 2020

Intel's First 7nm Client Microarchitecture is "Meteor Lake"

Intel's first client-segment processor microarchitecture built on its own 7 nm silicon fabrication process will be codenamed "Meteor Lake." The codename began surfacing in driver files and technical documents, one of which was screengrabbed and leaked to the web by Komachi Ensaka. Not much else is known about it, except that it succeeds the 10 nm++ "Alder Lake," an ambitious attempt by Intel to replicate Arm big.LITTLE heterogenous core technology on the x86 architecture, by combining a number of high-power cores with high-efficiency cores on a single piece of silicon. Intel "Lakefield," headed toward mass-production within this year, is the first such heterogenous core.

Older reports throughout 2019-20 speculate "Meteor Lake" (known at the time only by its name), could come out at a time when Intel monetizes its "Golden Cove" high-performance CPU core. It's quite likely that like "Alder Lake," it could be a heterogenous chip targeting several client form-factors, mobile and desktop. The company could leverage its 7 nm process - claimed to rival TSMC 5 nm-class in transistor density - in turning up core-counts over "Alder Lake." We'll learn more about "Meteor Lake" as we crawl toward its 2022 launch window, if it still holds up.
Source: Komachi Ensaka (Twitter)
Add your own comment

47 Comments on Intel's First 7nm Client Microarchitecture is "Meteor Lake"

#26
Vayra86
theoneandonlymrkThey could say they had a quantum processor that does 4x what any CPU and GPU combo can do and some, and I literally wouldn't care, shame they used to excite, now though :shadedshu:tut.
Well... they do and its leading, too. Part courtesy of Dutch research too :) Pretty amazing stuff they have set up, experiments of the tech working etc.

www.intel.com/content/www/us/en/research/quantum-computing.html

qutech.nl/

Its not fast though :D

Maybe that is Intel's big master plan. First they introduce 'bugs' and security issues with existing hardware, then they feed us bandaids that don't really fix much but do kill performance, and now they proceed with yet another stream of tiny improvements on the ever growing turd and every time just before the finish line, oops postponed or cancelled or 'roadmapped'. The end goal: we get so fed up with binary we will demand the ultra low performance of Quantum... :p
Posted on Reply
#27
dicktracy
Great job mods by showing your bias yet again.
Posted on Reply
#28
londiste
AnarchoPrimitivit's especially confusing when they have one name for the chipset/platform and another for the CPU generation, for example, how "Sapphire Rapids" CPU micro architecture will be on the "Eagle Stream" platform which will be followed by "Granite Rapids" and preceeded by "Tiger lake" who had an additional successor to "Sapphire Rapids" called "Alder lake"......
Unlike Banded Kestrel, Castle Peak, Colfax, Great Horned Owl, Snowy Owl or Whitehaven? :)

Generally, codenames themselves are not that important, the SKUs on sale will have different names anyway. If you want to follow the news on in-production and planned hardware, having some idea of the codenames just comes with the territory.

Some of the confusion in Intel lineup is because their server, desktop and mobile architectures do not align. The rest is due to all the 14nm updates they have been doing due to both 10nm being MIA as well as hardware fixes to security issues.
- Server: Skylake > Cascade Lake > Cooper Lake > Ice Lake > Sapphire Rapids > Granite Rapids > Diamond Rapids
- Desktop: Skylake > Kaby Lake > Coffee Lake > Coffee Lake Refresh > Comet Lake > Rocket Lake > Alder Lake > Meteor Lake
- Mobile: Skylake > Kaby Lake > Cannon Lake\Coffee Lake\Whiskey Lake > Ice Lake\Comet Lake > Tiger Lake\Rocket Lake > Alder Lake > Meteor Lake

The core architecture codenames have come into play for two reasons - all the 10nm drama, as well as Intel's apparent plan to do heterogeneous CPUs, where architecture of specific cores becomes important. Before this, code codename is/was usually the same as the CPU one.
- Mainline cores: Skylake > Palm Cove (in Cannon Lake) > Sunny Cove (in Ice Lake) > Willow Cove (Tiger Lake, Sapphire Rapids) > Golden Cove (probably Alder Lake and Granite Rapids), Ocean Cove (probably in Meteor Lake and Diamond Rapids).
There are other cores that are rumored to come into play soon, notably Tremont cores from Atom line into Alder Lake is the most known example.

Caring about Eagle Stream is a niche thing - it is a specific server platform that will be under Sapphire Rapids and Granite Rapids. If I remember correctly - this being a Whitley successor it should be the 2-socket platform.
Posted on Reply
#29
AsRock
TPU addict
Caring1It's going to crash and burn?
Just as bad as the PS4 AMD\SONY naming, Volcanic Islands.
Posted on Reply
#30
Tomorrow
londistePretty sure AMD has absolutely no interest in coming back to foundries especially considering how GlobalFoundries are doing these days. Even more so when we talk about cutting edge nodes. TSMC, Samsung and Intel are putting over $10 billion into foundry R&D each year.
???

5nm+ refers to TSMC's custom node for AMD in this context. I don't know where you got the idea to bring up GF.
londisteIntel has said 7nm has planned density twice of their 10nm. Their 10nm is about the same density as TSMC's N7. TSMC's N5 is estimated about twice the density of N7.
This puts Intel's 7nm and TSMC's N5 very close to one another in terms of density and due to physical characteristics, also in terms of voltage and power.
Im not so sure. Intel had to increase the pitch size to make their 10nm production viable. Besides it literally does not matter how dense Intel's node supposedly is if they can't produce it in volume. Something TSMC does not have a problem with. So even if Intel's 10nm is as dense as TSMC's 7nm the fact remains that TSMC has products in all segments: supercomputers, servers, desktop and mobile. Intel's 10nm is only in mobile thus far and even there it's nothing great.
Posted on Reply
#31
londiste
TomorrowIm not so sure. Intel had to increase the pitch size to make their 10nm production viable. Besides it literally does not matter how dense Intel's node supposedly is if they can't produce it in volume. Something TSMC does not have a problem with. So even if Intel's 10nm is as dense as TSMC's 7nm the fact remains that TSMC has products in all segments: supercomputers, servers, desktop and mobile. Intel's 10nm is only in mobile thus far and even there it's nothing great.
We do not know all too well, really. Intel states the same ~100MTr/mm^2 as TSMC does for the process. For both, this is with a high density library - for mobile SoCs and whatnot. Anything in desktop CPU space is definitely with a high performance library that is usually 30% less dense - in this case about 65-60 MTr/mm^2. That should be about where Cannon Lake was. Ice Lake is reported to be less dense but numbers are hard to come by - guesstimates are somewhere in 50-55MTr/mm^2 range.

Edit:
For comparison on TSMC side, Zen2 CCD is 3.9 billion transistors on 74mm^2 - roughly 52MTr/mm^2.
Tomorrow5nm+ refers to TSMC's custom node for AMD in this context. I don't know where you got the idea to bring up GF.
I am willing to bet that is not a thing. This is just clickbait derived from lost in translation. Foundries do work with customers to help them best optimize the production but custom node is not what this "5nm+" is.
Posted on Reply
#32
IceShroom
londisteIntel has said 7nm has planned density twice of their 10nm. Their 10nm is about the same density as TSMC's N7. TSMC's N5 is estimated about twice the density of N7.
This puts Intel's 7nm and TSMC's N5 very close to one another in terms of density and due to physical characteristics, also in terms of voltage and power.
That is on paper. What is the real density of intel's 10nm on real product??
Posted on Reply
#33
londiste
IceShroomThat is on paper. What is the real density of intel's 10nm on real product??
Exact enough transistor counts are not known for Intel CPUs. Estimates put Cannon Lake to around 60MTr/mm^2 and Ice Lake to around 50MTr/mm^2.
Posted on Reply
#34
ARF
londisteExact enough transistor counts are not known for Intel CPUs. Estimates put Cannon Lake to around 60MTr/mm^2 and Ice Lake to around 50MTr/mm^2.
300 million 10nm transistors
Coffee Lake is estimated to house 217 million 14nm transistors per core and Keller has stated that Ice Lake contains 300 million 10nm transistors.
www.pcgamesn.com/intel/next-gen-cpu-architecture-significantly-bigger-than-sunny-cove

Someone to measure the die size - pixels counting or real die size measurement.




Source: Google.
Posted on Reply
#35
IceShroom
londisteExact enough transistor counts are not known for Intel CPUs. Estimates put Cannon Lake to around 60MTr/mm^2 and Ice Lake to around 50MTr/mm^2.
Maybe even less, that is why Intel dont publish it.
Renoir has 62 MT/mm2 and it is a high proformence part.
Posted on Reply
#37
IceShroom
ARF300 million 10nm transistors
Coffee Lake is estimated to house 217 million 14nm transistors per core and Keller has stated that Ice Lake contains 300 million 10nm transistors.
www.pcgamesn.com/intel/next-gen-cpu-architecture-significantly-bigger-than-sunny-cove

Someone to measure the die size - pixels counting or real die size measurement.




Source: Google.
Those are some assumption on paper not real number.
The only known density 10nm Intel part is Lakefield and it is low power part. It's density is 49MT/mm2.
Lakefield -> 49MT/mm2
Renoir -> 62MT/mm2
A14/S865 -> Defenetly more than Renoir
Kirin 990 5G -> 90.9MT/mm2
www.anandtech.com/show/15099/the-huawei-mate-30-pro-review-top-hardware-without-google/2
Posted on Reply
#38
ARF
IceShroomThose are some assumption not real number.
The only known density 10nm Intel part is Lakefield and it is low power part. It's density is 49MT/mm2.
Core i5-L15G7 die size, source?

10nm parts:

Ice Lake Core i7-1068G7
Cannon Lake Core i3-8121U
Lakefield Core i5-L15G7
Posted on Reply
#40
londiste
Lakefield die size source is this:
Any sources for transistor counts?
Posted on Reply
#43
tygrus
The mention of cannons, meteors or craters remind me of disasters waiting to happen. Will they FAB plans crash and burn again?
Posted on Reply
#44
Prima.Vera
DDR5? PCI-E 4.0 ?? USB 4.0? 16 Cores for the i9 12900K one?
If not... DOA.
Posted on Reply
#45
Metroid
Here is hoping we will not wait thousands of years for the meteor to come to earth hehe, maybe intel intended to name meteor for the sake we will never see it to come to earth? and funny here I was hoping intel would deliver this time hehe

I guess we could say 10nm 2013 was the real meteor lake hehe, we are in 2020 and we are still waiting it hehe
Posted on Reply
#46
asher2420
edbe2022 it will be one player only.. AMD
AMD 5nm+
DDR5
PCi-eX 5
lmao
you obviously have no idea what is Intel and how much resources it has.
Posted on Reply
#47
ToxicTaZ
Prima.VeraDDR5? PCI-E 4.0 ?? USB 4.0? 16 Cores for the i9 12900K one?
If not... DOA.
14nm++
Intel Rocket Lake (H5 LGA 1200 socket PCIe 4.0 "8 Cores"

10nm+++
Intel Alder Lake (H6 LGA 1700 socket PCIe 4.0 "16 Cores" big.Little

7nm+
Intel Meteor Lake (H6 LGA 1700 socket PCIe 5.0

Intel H6 LGA 1700 socket will have PCIe 5.0 ready DDR5 and USB-4 and all the other goodies.... Like WiFi-6E 5G ........

Intel Meteor Lake will double Sky Lake IPC performance.
Posted on Reply
Add your own comment
Mar 31st, 2025 15:18 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts