Intel "Meteor Lake" Chips Already Being Built at the Arizona Fab

[Richards]

Lol Apple's m1 max destroys most of the electricity pig desktops un the market for less power... the m2 will destroy a 3090 or 4090 using 50watts lol

I think you have a real problem grasping a what desktop represents. Higher power consumption is necessary to achieve certain level of performance. You may argue if the power used by components is higher than expected or not but it all comes down to components. You claim desktops have too high power consumption but yet this is the only scenario where you can actually do this without getting your system constrained. You cant compare a laptop to a desktop because these are two different segments. Saying that desktops use too much power because laptops dont is foolish. Desktops can use more power and have better performance because you can cool those easier than in a confined space laptop gives you. That is why laptop are lower power consuming machines thus less powerful. Desktops dont have terrible power consumption than laptops. They have higher power capabilities which prevents laptops having this due to heat and cooling constraints. Laptops are compact and build with power saving in mind giving certain level of performance.
Now, a processor may have a ridiculous or terrible power consumption to achieve a certain level of performance in today's standard but that is basically it. You can balance it though to get as much as possible from a CPU within a reasonable power consumption range.
People prefer laptops over desktops. That is the most vague and ridiculous thing one can say really.

 

[ratirt]

If I get that answer, all I can say is, cool, go for it. (So?)

It's like saying "I don't need a sports car, I just go shopping once a week." Well, don't buy a sports car, then.

Talking about a legit argument hehe

Lol Apple's m1 max destroys most of the electricity pig desktops un the market for less power... the m2 will destroy a 3090 or 4090 using 50watts lol

Twist away bro twist away. It would seem you are very good at it

 

[AusWolf]

Lol Apple's m1 max destroys most of the electricity pig desktops un the market for less power... the m2 will destroy a 3090 or 4090 using 50watts lol

You do realize that Apple has an entirely different market segment on its own, don't you?

In other words, unless you show me where I can buy a desktop M1 CPU with a compatible motherboard that runs Windows, your point is totally irrelevant.

 

[TheoneandonlyMrK]

Lol Apple's m1 max destroys most of the electricity pig desktops un the market for less power... the m2 will destroy a 3090 or 4090 using 50watts lol

And it's total shite, show the M2 destroying the 3090-4090(in a vague metric not yet mentioned but soon to be cherry picked by you), you can't it isn't out so as a point , worthless.

 

[bug]

Great to see Intel is only 7 years behind AMD on chiplets/glue.

As if chiplets were an end-goal on their own...

 

[enzolt]

Intel's idea to mitigate these problems is bunch of EMIB and dies with mirrored layouts. It should - in theory - get close enough to monolithic architecturally. With EMIB the connections are quite a bit more efficient than over PCB and should allow for wide/fast connections between dies/chiplets/tiles.

The really interesting details on Sapphire Rapids (for example whether memory will be UMA or NUMA) are so far quite scarce but we will see soon enough.
Everything new is well-forgotten old

In Nephilim666 defense, theyre probably still a baby when the Pentium D was out.

 

[qubit]

What happened to the 13th gen? They're not pandering to people's superstitions about the lucky number 13 are they?

 

[sillyconjunkie]

Isn't Sapphire Rapids similar to first gen EPYC (Naples) ?
- Each Die(Tile) is a complete SoC on its own
- Separated memory controller on each Die(Tile)
- High latency on accessing data located in the memory on the opposite corner.

Kinda. Connectivity between dies/tiles is different.

EPYC, TR, etc use substrate traces to connect the dies. Higher latency communication between dies and the reason why EPYC/TR packages are so large. Each die/tile is an independent SoC.

Sapphire Rapids uses silicon interconnects between the tiles/dies. Lower latency communication between tiles and closer to one uniform SoC package than EPYC (design-wise).

 

[lexluthermiester]

Where exactly do you see any mention of 40 cores for a laptop CPU? Besides, you'd have 256 cores for desktop at that stage, so desktop would still be massively superior

Seems you missed some context. btarunr was pointing out the flaw in your comparative AMD jab at Intel concerning die "glue".

 

[Nuckles56]

Seems you missed some context. btarunr was pointing out the flaw in your comparative AMD jab at Intel concerning die "glue".

You're wrong there Lex, it was addressed at Richard's statement about a 40 core CPU for a laptop (192 EU GPU as well), which was never mentioned in the original story by btarunr

 

[lexluthermiester]

You're wrong there Lex, it was addressed at Richard's statement about a 40 core CPU for a laptop (192 EU GPU as well), which was never mentioned in the original story by btarunr

Let's review...

Nephilim666 said;

Great to see Intel is only 7 years behind AMD on chiplets/glue.

This statement very clear. Then btarunr responded...

Xeon Sapphire Rapids is essentially EPYC Rome (four chiplets), with better named glue.

View attachment 225708

..which directly address Nephilim666's comment.

Then Richards responded;

This is very advanced packaging... 40 cores+ 192 eu laptops will make desktops irrelevant very quickly

Which clearly alludes to the die interconnect. Then you commented;

Where exactly do you see any mention of 40 cores for a laptop CPU? Besides, you'd have 256 cores for desktop at that stage, so desktop would still be massively superior

And thus my comment.


So unless Richards was NOT responding to btarunr...