• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

14th Gen "Raptor Lake Refresh"

Soldato
Joined
31 Oct 2002
Posts
9,863
Good stage of development to start this thread!

N8n2Hfo.png

14th Gen Core “Meteor Lake” has been powered-on, on track to launch in 2023​


Source: https://videocardz.com/newz/14th-gen-core-meteor-lake-has-been-powered-on-on-track-to-launch-in-2023

Pat Gelsinger said:
“Intel 4 Meteor Lake has now successfully booted Windows, Chrome, and Linux. The speed at which the team was able to achieve this milestone is a significant sign of the health of both Meteor Lake and our Intel 4 process technology.”

— Pat Gelsinger, Intel CEO


13th gen Raptor Lake (also a new architecture) should launch in a few months, though won't have too long until 14th gen is knocking on the door!

We all have to be thankful to AMD, for pushing Intel to execute and advance so quickly. 2022 and 2023 will be very interesting!
 
Last edited:
Man of Honour
Joined
22 Jun 2006
Posts
11,656
Big milestone for new process & new architecture :)

I guess I can understand that :D But, what I'd like to hear Pat say is "we can produce these at huge volume, with great power consumption and performance", since they had 10nm working for ages, they just couldn't make any (for desktop).
 
Soldato
OP
Joined
31 Oct 2002
Posts
9,863
I guess I can understand that :D But, what I'd like to hear Pat say is "we can produce these at huge volume, with great power consumption and performance", since they had 10nm working for ages, they just couldn't make any (for desktop).

Intel's '7' process (10nm) was an improvement in power consumption, over 14nm+++, though as you said it was delayed a very long time.

Hopefully Intel's '4' process (7nm), that Meteor Lake is built on, is a further improvement. It's not been delayed like it's predecessor was, though that'smostly all we know for now.

Excited to see Intel's 7nm vs TSMC's 5nm. Though I suspect Intel having access to huge amount of wafers that it produces itself, will win it the supply (and market share) battle, judging by the rising prices/demand commanded by TSMC's leading node. Apple are their #1 customer after all.
 
Soldato
Joined
30 Jun 2019
Posts
7,875
Looking good, this is the one to watch. the Ratio of E-Cores to P-Cores looks like it will be 1:1, if the diagram in the slide is anything to go by.

Intel seems to think that making efficient use of the physical die space should be their priority, I wonder if I'm misunderstanding their logic here?

I am disappointed that the design includes E-Cores tbh. It means people will be paying for silicon that they won't use much / if at all on desktop PCs.
Intel have yet to demonstrate E-Core effectiveness on anything other than mobile devices.

Maybe the E-Cores will seem like a side show, if there are 4 or 8 Tiles for the top CPU products.

You have to wonder how much the design will be influenced by the tiled design of Sapphire Rapids server processors.
 
Last edited:
Associate
Joined
4 Oct 2019
Posts
2,169
Location
Belfast
There will always be a new product in the pipeline.

It takes billions of dollars to create a fab to product CPUs - you can't do that unless you've got a pretty damn good idea as to how you're going to pay it off in the next 6+ years.

Nice to see that Intel are moving forward with new products in the 7nm process but you can't really compare one process to another. A really ****** 5nm CPU may not outperform a 14nm++++ process.

Hopefully AMD gets the memo and we get a nice price to performance competition for our money.
 
Soldato
Joined
30 Jun 2019
Posts
7,875
Zen 4 is going to have models with extra cache (V-Cache) next year, so Meteor Lake is gonna have a tough fight on it's hands, especially in games.
 
Soldato
Joined
30 Jun 2019
Posts
7,875
True. For one thing, AMD's 5800X with V-cache has only shown performance improvements in games so far.

Meteor Lake or Zen 4, I think both will easily be fast enough for current games and games released in 2023. Alder Lake CPUs like the 12700K and 12900K are already coping well with even the most intensive games (Battlefield 2042 with 128 players comes to mind).

Transistor density of intel's 7nm EUV process is targeted to be greater than Zen 4's TSMC 5nm EUV process, so that should count in Intel's favour. According to Wikichip, this will be "around 202-250 million transistors per square millimeter". source here: https://en.wikichip.org/wiki/7_nm_lithography_process#P1276

We shoudn't really be paying much attention to nanometer transistor sizes of CPUs anymore, it's bound to be misleading.

We need more information about Meteor Lake's readiness for desktop products really...
 
Last edited:
Soldato
Joined
6 Feb 2019
Posts
17,595
Meteor lake is ahead of schedule well done.

I wonder if next year Zen 5 will be 3nm or 5nm.

If AMD is still on 5nm in 2023 then Intel can have the advanced given Intel 4 is equivalent to TSMC's 4nm.
If AMD is on 3nm then they have the advantage but they also stuck, TSMC is not taping out 2nm until 2026 so AMD will have 3-4 years where its stuck on 3nm.

So in all likely hood, barring some huge screw up Intel will reach 2nm well before AMD can
 
Soldato
OP
Joined
31 Oct 2002
Posts
9,863
Meteor lake is ahead of schedule well done.

I wonder if next year Zen 5 will be 3nm or 5nm.

If AMD is still on 5nm in 2023 then Intel can have the advanced given Intel 4 is equivalent to TSMC's 4nm.
If AMD is on 3nm then they have the advantage but they also stuck, TSMC is not taping out 2nm until 2026 so AMD will have 3-4 years where its stuck on 3nm.

So in all likely hood, barring some huge screw up Intel will reach 2nm well before AMD can

All very true. I think the most important factor for AMD will be how many wafers they'll be able to secure from TSMC on 5nm/3nm and at what price.

Intel's invested heavily in additional fabs, this could mean Intel are able to produce many more wafers (and thus CPU's) at a lower cost than TSMC. Or perhaps TSMC will have much better yields than Intel, making them far cheaper.

Can't wait to see what happens!
 
Last edited:
Soldato
Joined
30 Jun 2019
Posts
7,875
Was I incorrect to think that Meteor Lake is likely to have more than 8 P-Cores?

I think Intel is keeping quiet on the details at the moment, at least with Raptor Lake still to launch.
 
Soldato
Joined
6 Feb 2019
Posts
17,595
Is 'tiled' architecture the equivalent of chiplets?

Yes. Intel's approach has the benefit that each chiplet can be manufactured by a different factory on a different process node and still work together - for example one factory could make big cpu cores on 5nm and another does little cpu cores on 7nm and another makes the iGPU on 4nm and so forth

As far as I can tell though Intel doesn't plan to outsource meteor lake and the whole thing will be on 4nm. I think the chiplet design is just in case Intel runs into problems in the future then it's super easy to outsource other companies to build the cpu
 
Soldato
Joined
11 Oct 2009
Posts
16,591
Location
Greater London
Yes. Intel's approach has the benefit that each chiplet can be manufactured by a different factory on a different process node and still work together - for example one factory could make big cpu cores on 5nm and another does little cpu cores on 7nm and another makes the iGPU on 4nm and so forth

Correct me if I'm wrong but doesn't AMD's approach also use different foundries? The cores and v-cache made with TSMC 7nm, and the IO die made with GF's 12nm.

Main advantage really are the different dies are literally next to each other, which should give it the advantage of bandwidth and latency.
 
Associate
Joined
2 Sep 2016
Posts
911
Correct me if I'm wrong but doesn't AMD's approach also use different foundries? The cores and v-cache made with TSMC 7nm, and the IO die made with GF's 12nm.

Main advantage really are the different dies are literally next to each other, which should give it the advantage of bandwidth and latency.

Correct grim should know better
 
Back
Top Bottom