Intel 14th Gen Core Lineup Confirmed to be Meteor Lake CPU Range

[R0H1T]

No it will be a new socket.

The purpose of this leak is basically just to confirm that even if their is a raptor lake refresh (which is the current heavy rumor) it won't carry the 14th gen name.

Or that MTL isn't coming to desktops, at least according to some rumors. TGL & first gen 7nm (don't remember the name) was never a desktop thing either so there's precedent in there.

 

[Hxx]

So this is an LGA1700 part?

It will work on B660?

there are 2 series of chips discussed in this article. RL refresh coming out in Q3 will be on LGA 1700 and will presumably support b660 although nothing has been officially confirmed. Meteor Lake coming out next year will be on a brand new socket and chipset and will not be backwards compatible.

 

[AusWolf]

That's not likely to happen. Rather, like Apple, more chips, such as memory, will be unified into the CPU package.
The latest bus standards have so much wiring that if the North Bridge were separated now, the PCB would not fit into the laptop chassis.

I know. I was only joking.

The iGPUs up to the 13th gen are DG1 (Xe-LP) and Meteor lake is DG2 (Xe-LPG) based. mTL GT2 is basically Arc A370M with the tensor core (XMX) removed, and according to Techpowerup's GPU database, is twice as powerful as the 13 th gen iGPUs.

Intel Iris Xe Graphics 96EU Mobile
Pixel Rate: 33.60 GPixel/s
Texture Rate: 67.20 GTexel/s

Intel Arc A370M
Pixel Rate: 65.60 GPixel/s
Texture Rate: 131.2 GTexel/s

Ah, so by "Alchemist iGPU", they mean it's getting full RT and/or XMX support now? That's the only architectural difference between Alchemist and 11-12-13th gen Xe, as far as I know.

 

[hs4]

I know. I was only joking.


Ah, so by "Alchemist iGPU", they mean it's getting full RT and/or XMX support now? That's the only architectural difference between Alchemist and 11-12-13th gen Xe, as far as I know.

There was a problem with my post. To clarify, the 14th gen iGPU (Xe-LPG) has RT and not XMX. The change from Intel7 to N5 is expected to improve the iso-power clock, and the separation of tiles will decouple it from the CPU's L3/ring bus system, resulting optimization on memory sharing.


The Xe version numbers are confusing.

The 11-12-13th gen iGPUs are:
Internal code DG1, Xe-LP, "0th gen ARC", dGPU is named Xe MAX.

The 14th gen iGPU is.
Internal code DG2, Xe-LPG, 1st gen ARC, dGPU is namedAlchemist.

The Lunar lake iGPU is:
Internal code DG3, Xe2-LPG, 2nd gen ARC, dGPU is named Battlemage.

 

[AusWolf]

There was a problem with my post. To clarify, the 14th gen iGPU (Xe-LPG) has RT and not XMX. The change from Intel7 to N5 is expected to improve the iso-power clock, and the separation of tiles will decouple it from the CPU's L3/ring bus system, resulting optimization on memory sharing.


The Xe version numbers are confusing.

The 11-12-13th gen iGPUs are:
Internal code DG1, Xe-LP, "0th gen ARC", dGPU is named Xe MAX.

The 14th gen iGPU is.
Internal code DG2, Xe-LPG, 1st gen ARC, dGPU is namedAlchemist.

The Lunar lake iGPU is:
Internal code DG3, Xe2-LPG, 2nd gen ARC, dGPU is named Battlemage.

Apologies, my post might have been confusing as well.

What I mean is, I don't know about any architectural difference between Xe-LP and Xe-LPG apart from RT and XMX. The rest of the chip is a scaled-up version of Xe-LP, isn't it?

 

[hs4]

Apologies, my post might have been confusing as well.

What I mean is, I don't know about any architectural difference between Xe-LP and Xe-LPG apart from RT and XMX. The rest of the chip is a scaled-up version of Xe-LP, isn't it?

Yes. Theoretical Performance per clock per EU of DG1 and DG2 is exactly the same. The differences are following:

- Now supports RT.
- Unslice is updated, e.g. now it handles AV1 format.
- 1.5-2.0x performance is expected at the same power due to 1.3x EU count and higher clocks with the better node.

 

[agent_x007]

Haswell-E is 4th gen and Broadwell-E is 5th gen (despite the -5000 and -6000 nomenclature for these chips), and neither design has integrated graphics, which the media SDK seems to entirely rely upon. That's why only the mainstream desktop Core and Xeon E3 segments are serviced, they do have an iGPU.

Oh, so this isn't for Arc/Alchemist in general, but only for iGPUs ?

 

[watzupken]

This sounds like its going to be great,
If i understand this correctly, its confirmed that 14th Gen intel processors will be Meteor Lake, 5nm/6nm process
11th gen intel was 14nm
12th gen intel was 10nm
This next generation finna be Fast AF Boii

As far as I recall, the replacement of Raptor Lake should be based on Intel 7nm, which is not to be confused with Intel 7 which is essentially Intel’s 10nm+++++. I actually lost track of the number of + behind. The first Intel 10nm product was actually the laptop only Cannon Lake, which was replaced by Ice Lake, and subsequently Tiger and Alder Lake.

To me, I am less optimistic about Meteor Lake because the way that Intel is limiting this only to laptop kind of sound similar to the failed Cannon and Ice Lake processors. It is mentioned that Meteor Lake is limited to mobile/ laptop. So if you are looking for a desktop version, then I think you may be disappointed. The other red flag is the regression in number of P-cores. Intel when releasing Alder and Raptor Lake is clearly aware that 8 P-cores is the sweet spot, and they can “fortify” their multithreaded performance through the spamming of E-cores. But yet the number of P-cores regressed. Perhaps they are trying to limit power draw for the mobile space, but considering existing Raptor Lake mobile processors are drawing record amount of power for a laptop, it is surprising why Intel will bother about dropping core count to “keep power in check”. Unless there is a real power or density crisis with their 7nm that is forcing them to take this design approach. It is all my speculation, but I think we will know soon.

 

[Minus Infinity]

Or that MTL isn't coming to desktops, at least according to some rumors. TGL & first gen 7nm (don't remember the name) was never a desktop thing either so there's precedent in there.

It now appears ML for desktop will only be for i3 and i5. Arrow Lake will target i7 and i9, but that's 12-18 months away.

As far as I recall, the replacement of Raptor Lake should be based on Intel 7nm, which is not to be confused with Intel 7 which is essentially Intel’s 10nm+++++. I actually lost track of the number of + behind. The first Intel 10nm product was actually the laptop only Cannon Lake, which was replaced by Ice Lake, and subsequently Tiger and Alder Lake.

Yes ML is on Intel 4 which is their rebranded 7nm and Arrow Lake is on Intel 3, a more refined version of Intel 4.

 

[hs4]

As far as I recall, the replacement of Raptor Lake should be based on Intel 7nm, which is not to be confused with Intel 7 which is essentially Intel’s 10nm+++++. I actually lost track of the number of + behind. The first Intel 10nm product was actually the laptop only Cannon Lake, which was replaced by Ice Lake, and subsequently Tiger and Alder Lake.

To me, I am less optimistic about Meteor Lake because the way that Intel is limiting this only to laptop kind of sound similar to the failed Cannon and Ice Lake processors. It is mentioned that Meteor Lake is limited to mobile/ laptop. So if you are looking for a desktop version, then I think you may be disappointed. The other red flag is the regression in number of P-cores. Intel when releasing Alder and Raptor Lake is clearly aware that 8 P-cores is the sweet spot, and they can “fortify” their multithreaded performance through the spamming of E-cores. But yet the number of P-cores regressed. Perhaps they are trying to limit power draw for the mobile space, but considering existing Raptor Lake mobile processors are drawing record amount of power for a laptop, it is surprising why Intel will bother about dropping core count to “keep power in check”. Unless there is a real power or density crisis with their 7nm that is forcing them to take this design approach. It is all my speculation, but I think we will know soon.

When comparing the 4-core 1065G7 (Acer Swift 3⁑) and 1135G7 (Dell Inspiron 13⁑) at the same power consumption, the iso-power clock is improved by about 10%, and 17% higher performance including the IPC difference (7%).

Comparing the same 6-core 11400H⁑ (10nm Super Fin) and 12400⁂ (Intel 7), the iso-power clock is improved by about 10-15%, and including the IPC difference (18%), the latter has 30 % better performance.

Comparing 12900K⁂ and 13700K⁂ with the same 8P+8E configuration, the iso-power performance is improved by 10%.

Ice lake's 10nm and Raptor lake's Intel 7 have improved the iso-power clock by about 35%. It corresponds to the 1.5th generation progress (roughly equivalent to the performance difference between N7 and N4), considering that the iso-power clock improvement for an average node generation is around 20%. Strictly speaking, the initial 10nm was too unfinished.

⁑ notebook check
⁂Blog measuring current at EPS12V

 

[AusWolf]

When comparing the 4-core 1065G7 (Acer Swift 3⁑) and 1135G7 (Dell Inspiron 13⁑) at the same power consumption, the iso-power clock is improved by about 10%, and 17% higher performance including the IPC difference (7%).

Comparing the same 6-core 11400H⁑ (10nm Super Fin) and 12400⁂ (Intel 7), the iso-power clock is improved by about 10-15%, and including the IPC difference (18%), the latter has 30 % better performance.

Comparing 12900K⁂ and 13700K⁂ with the same 8P+8E configuration, the iso-power performance is improved by 10%.

Ice lake's 10nm and Raptor lake's Intel 7 have improved the iso-power clock by about 35%. It corresponds to the 1.5th generation progress (roughly equivalent to the performance difference between N7 and N4), considering that the iso-power clock improvement for an average node generation is around 20%. Strictly speaking, the initial 10nm was too unfinished.

⁑ notebook check
⁂Blog measuring current at EPS12V

I love the way you include annotations in your post to indicate your sources.