Discussion Intel current and future Lakes & Rapids thread

[A///]

Yep. Thats a bit of a worry.

I'm sure it'll be fine. It's hopefully being announced next week and hopefully I can get my hands on an MTL laptop by end of December. I expect sales to be quite brisk due to the romp pomp mtl has had for a year or two now.

 

[TESKATLIPOKA]

I'm sure it'll be fine. It's hopefully being announced next week and hopefully I can get my hands on an MTL laptop by end of December. I expect sales to be quite brisk due to the romp pomp mtl has had for a year or two now.

I would rather wait for reviews to see how good MTL is, just to be sure, but the leaks are kinda promising.

This does look very nice for MT performance If the last number is for all-core turbo, but the question is at what power?
They use H and that is used for 45W as base power.

This Geekbench 6 screenshot shows 3.4GHz for Core Ultra 165H. Geekbench

Not bad considering i9-13950H has 2.6GHz as base. That would mean 3.4/2.6= 1.31 or 31% higher base clock, not sure about E-cores.

 

[coercitiv]

Not bad considering i9-13950H has 2.6GHz as base. That would mean 3.4/2.6= 1.31 or 31% higher base clock, not sure about E-cores.

Keep in mind you're comparing 8+16 with 6+8+2. So you get ~30% higher base clocks on the P cores but the older CPU has ~30% more P cores to begin with. It's still a good trade-off for consumer workloads, but not as surprising.

 

[TESKATLIPOKA]

Keep in mind you're comparing 8+16 with 6+8+2. So you get ~30% higher base clocks on the P cores but the older CPU has ~30% more P cores to begin with. It's still a good trade-off for consumer workloads, but not as surprising.

I wasn't talking about i9-13950HX, which has only 2.2GHz base clock at 55W.

I wrote i9-13950H, which should have been i9-13905H a 6P+8E CPU with 2.6GHz base clock at 45W, but Anandtech article had a typo.

edit:
If I compare i9-13900HK(45W) vs i7-1370P(28W), both of them 6P+8E CPUs, then the only difference is 200MHz in boost for both types of cores.
But base clock difference is 37% for P-cores and 36% for E-cores.
It looks like MTL could be set to 28W, but have base clock as a 45W Raptor or better.

P.S. i9-13900HK, i9-13905H and i9-13900H have the same specs.

 

[A///]

I would rather wait for reviews to see how good MTL is, just to be sure, but the leaks are kinda promising.

of course but the intel laptop I've got is long in the tooth. I could go with amd but there's simply more choice with intel and mtl sounds good. i'll wait for reviews but even if it's slightly slower than rpl laptops it's still worth it not being a little heater on the old lap. I suppose the m3 macbook will be available in summer and that'll be my early crimbo present then.

this all leans on whether I can get my hands on a laptop this year with mtl. every thread of my being tells me it'll be sought after and be produced in limited numbers until 2024 when it'll be readily available

 

[TESKATLIPOKA]

of course but the intel laptop I've got is long in the tooth. I could go with amd but there's simply more choice with intel and mtl sounds good. i'll wait for reviews but even if it's slightly slower than rpl laptops it's still worth it not being a little heater on the old lap. I suppose the m3 macbook will be available in summer and that'll be my early crimbo present then.

this all leans on whether I can get my hands on a laptop this year with mtl. every thread of my being tells me it'll be sought after and be produced in limited numbers until 2024 when it'll be readily available

My Legion(i5-9300H + GTX 1650M) also needs a replacement, but I can wait one more year.

My problem with MTL is that IGP won't be powerful enough to be considered as an upgrade for me, so that leaves me with dGPU, but unless I buy a more expensive laptop with RT 4080 everything is limited to 8GB Vram and that's a serious no-go zone for me.
I personally would prefer a compact one under 15-inch, but then RTX 4080 is impossible.
I will wait for Strix Halo, maybe there will be a compact offering with 48GB RAM.
If not, then other CPU+dGPU combo or I will build a desktop If It will provide a much better value for my money.

 

[SiliconFly]

I wasn't talking about i9-13950HX, which has only 2.2GHz base clock at 55W.

I wrote i9-13950H, which should have been i9-13905H a 6P+8E CPU with 2.6GHz base clock at 45W, but Anandtech article had a typo.

edit:
If I compare i9-13900HK(45W) vs i7-1370P(28W), both of them 6P+8E CPUs, then the only difference is 200MHz in boost for both types of cores.
But base clock difference is 37% for P-cores and 36% for E-cores.
It looks like MTL could be set to 28W, but have base clock as a 45W Raptor or better.

P.S. i9-13900HK, i9-13905H and i9-13900H have the same specs.

Even I think its gonna be around 28W. But god knows what Intels gonna do.

 

[TESKATLIPOKA]

Even I think its gonna be around 28W. But god knows what Intels gonna do.

Intel or OEMs?
We are talking about 3 different configurations:
6P+8E+2e <- bigger CPU tile
4P+8E+2e <- a cutdown of the big CPU tile
2P+8E+2e <- smaller CPU tile
Considering Intel uses H in the name, I would expect 45W.

Intel also plans Raptor Lake Refresh, so probably that will be used for the >55W machines.
I don't think Intel will be able to supply enough MTL chips to be truthful.

edit: Maybe this is still valid.

 

[mikk]

There is no P branding anymore, it's all H and U. We can't see from the H label if it's 45W or 28 SKU. Ultra 9 should be 45W however.

 

[TESKATLIPOKA]

There is no P branding anymore, it's all H and U. We can't see from the H label if it's 45W or 28 SKU. Ultra 9 should be 45W however.

Maybe default TDP will stay as 15W and 45W.
then configurable TDP range will include 28W.

 

[lightisgood]

Another Intel Meteor Lake-P CPU spotted, Core Ultra 7 1002H features 16 cores and 22 threads - VideoCardz.com

Another Core Ultra CPU spotted Intel preparing its Meteor Lake CPU series to launch in the coming months. The newly discovered Core Ultra 7 1002H is not the first Meteor Lake mobile chip to surface. The formerly found SKU called Core Ultra 7 1003H quickly sparked a debate about eventual Intel...

videocardz.com

Core Ultra 7 1002H (ES) is probably 28W processor.

 

[Geddagod]

I would rather wait for reviews to see how good MTL is, just to be sure, but the leaks are kinda promising.

This does look very nice for MT performance If the last number is for all-core turbo, but the question is at what power?
They use H and that is used for 45W as base power.

This Geekbench 6 screenshot shows 3.4GHz for Core Ultra 165H. Geekbench

Not bad considering i9-13950H has 2.6GHz as base. That would mean 3.4/2.6= 1.31 or 31% higher base clock, not sure about E-cores.

At a 45 watt power limit, the 12900h has it's p cores at 2.8 GHz, e cores at 2.3 GHz. 35% higher P-core frequency and 20% higher E core frequency all core seems like a lot tbh, and a bit hard to believe.

 

[TESKATLIPOKA]

At a 45 watt power limit, the 12900h has it's p cores at 2.8 GHz, e cores at 2.3 GHz. 35% higher P-core frequency and 20% higher E core frequency all core seems like a lot tbh, and a bit hard to believe.

Those clocks are all-core turbos? Base clock for 12900H are 2.5GHz for P-cores and 1.8GHz for E-cores.
I don't know how you got 35% higher P-core frequency, because 4.3 or 4.5GHz vs 2.8GHz is 54-61% higher.

 

[Geddagod]

Those clocks are all-core turbos?

Ye, jarrods tech, CBR23 IIRC

Base clock for 12900H are 2.5GHz for P-cores and 1.8GHz for E-cores.

Base clock doesn't always reflect all core turbos

I don't know how you got 35% higher P-core frequency, because 4.3 or 4.5GHz vs 2.8GHz is 54-61% higher.

Oh I think I may have read his post wrong. I thought 3.8/2.8 was the P core/ E-core all core frequencies, but I think he meant 4.5 was the P core all core boost clock, 3.8 GHz was the E-core boost clock (single core) and 2.8 Ghz is the E-core all core frequencies.
Ye that makes these numbers seem even more improbable lol, I fully expect the power consumption of these MTL chips to be >45 watts. Maybe 80.

 

[TESKATLIPOKA]

Ye, jarrods tech, CBR23 IIRC

Good to know.

Base clock doesn't always reflect all core turbos

I know. I use base clocks because that's basically guaranteed clockspeed.

Oh I think I may have read his post wrong. I thought 3.8/2.8 was the P core/ E-core all core frequencies, but I think he meant 4.5 was the P core all core boost clock, 3.8 GHz was the E-core boost clock (single core) and 2.8 Ghz is the E-core all core frequencies.
Ye that makes these numbers seem even more improbable lol, I fully expect the power consumption of these MTL chips to be >45 watts. Maybe 80.

Yes, It is very high. I will stay with base clocks. Looks more believable.

 

[dullard]

OneRaichu has Geekbench 6 14900KF benchmarks: https://browser.geekbench.com/v6/cpu/2600997

I couldn't find an exact duplicate system with the 13900KF for a proper comparison. But does look like maybe 1% to 20% gains depending on the test. Reasonably similar system: https://browser.geekbench.com/v6/cpu/compare/2600997?baseline=2591765 In that comparison, the single core score went up 7.9% and multi-core went up 6.3%.

Question: is this the same Raichu with past leaks?

 

[trivik12]

MTL details we should hopefully know next week. If the base clock speeds are this high, it should be very good and makes their future iterations like Intel 3/20A/18A even more promising. Big question is why ARL-S would be still on N3E?

 

[dullard]

MTL details we should hopefully know next week. If the base clock speeds are this high, it should be very good and makes their future iterations like Intel 3/20A/18A even more promising. Big question is why ARL-S would be still on N3E?

Here are a few possible reasons to use an external node (most likely it is a combination of these):

1) Risk reduction. When Intel was struggling due to severe node delays, they announced that they will go down multiple paths for future projects. That way their CPUs won't be delayed if the node is delayed. Using external fabs is one of the paths that Intel mentioned.

2) Fab wafers must be reserved well in advance. By the time Intel figured out that their nodes are doing okay, contracts were already signed for a specified number of wafers. If you have already paid for it, why waste it?

3) Intel only has a small number of EUV machines. They just don't have enough production capacity for Meteor Lake, and Arrow Lake, and the server CPUs, and their external fab sales, and the line time needed to actually do testing of nodes and new designs.

 

[Abwx]

Here are a few possible reasons to use an external node (most likely it is a combination of these):

1) Risk reduction. When Intel was struggling due to severe node delays, they announced that they will go down multiple paths for future projects. That way their CPUs won't be delayed if the node is delayed. Using external fabs is one of the paths that Intel mentioned.

2) Fab wafers must be reserved well in advance. By the time Intel figured out that their nodes are doing okay, contracts were already signed for a specified number of wafers. If you have already paid for it, why waste it?

3) Intel only has a small number of EUV machines. They just don't have enough production capacity for Meteor Lake, and Arrow Lake, and the server CPUs, and their external fab sales, and the line time needed to actually do testing of nodes and new designs.

4) External node has significantly better perf/watt than available in house processes...

 

[H433x0n]

I would rather wait for reviews to see how good MTL is, just to be sure, but the leaks are kinda promising.

This does look very nice for MT performance If the last number is for all-core turbo, but the question is at what power?
They use H and that is used for 45W as base power.

This Geekbench 6 screenshot shows 3.4GHz for Core Ultra 165H. Geekbench

Not bad considering i9-13950H has 2.6GHz as base. That would mean 3.4/2.6= 1.31 or 31% higher base clock, not sure about E-cores.

The numbers he put up are as follows-

P core boost frequencies:
1 Core / 2 Core / 4 Core / 6 Core

E core boost frequencies
1 Core / 8 Core

 

[TESKATLIPOKA]

The numbers he put up are as follows-

P core boost frequencies:
1 Core / 2 Core / 4 Core / 6 Core

E core boost frequencies
1 Core / 8 Core

I know, but 4.5GHz for all-core turbo is a gigantic improvement compared to Raptor or Alder Lake.
How much does a comparable i9-13900H manage at 45W? 3-3.2GHz? So basically 40-50% improvement.

 

[FangBLade]

I'm interested in how the 6P cores will perform in next-gen games. Communication between the P and E cores adds significant latency, and on top of that, there's a big difference in performance, so it's questionable how much benefit these E cores will have for games unless developers specifically optimize the game from the ground up for Intel's architecture.

 

[H433x0n]

I know, but 4.5GHz for all-core turbo is a gigantic improvement compared to Raptor or Alder Lake.
How much does a comparable i9-13900H manage at 45W? 3-3.2GHz? So basically 40-50% improvement.

It’s a massive node jump with many improvements over Intel 7 (mainly significantly less leakage current).

 

[TESKATLIPOKA]

It’s a massive node jump with many improvements over Intel 7 (mainly significantly less leakage current).

ASUS TUF Gaming A17 (FA707, 2023)​

8C16T 7940HS manages 4.73 GHz at 80W, not sure in what test. I think at 45W It would be <4.5GHz.
Yet a 6P MTL would manage 4.5GHz while also needing to feed 8-10 E-cores looks a bit too good.

Here is Raptor for comparison.

 

[SiliconFly]

4) External node has significantly better perf/watt than available in house processes...

20A is supposed to a better node with better PPA than N3

It’s a massive node jump with many improvements over Intel 7 (mainly significantly less leakage current).

Intel 7 to Intel 4 PPW gain is just 20% (but pat mentioned Intel 4 cell library is optimized for efficiency unlike Intel 7 which is heavily optimized for performance).

The massive node jump actually comes next year from Intel 4 to Intel 20A with a PPW gain of 36%

Not sure about ARL desktop parts, but ARL mobile parts will gain a lot from this jump.