Discussion Intel current and future Lakes & Rapids thread

[igor_kavinski]

Intel marketing materials confirm upcoming 6.0 GHz 24-core, 5.6 GHz 20-core and 5.3 GHz 14-core CPUs - VideoCardz.com

Intel teases upcoming desktop CPU upgrade Marketing materials confirm configs of three 14th Gen Core models. The first official marketing materials for Raptor Lake Refresh to leak are not very detailed, but they do confirm the configuration of the new series. The slides show the basic specs for...

videocardz.com

First ever non-KS 6.0 GHz CPU.

How high is Intel gonna fly with a KS? 6.3 GHz?

 

[inf64]

Intel confirmed that P cores in MTL are essentially the same as in Raptor Cove:

Intel Says Meteor Lake & Raptor Lake CPUs Have Very Similar P-Core & E-Core Architectures, Intel 4 Brings Efficiency Improvements

Intel has confirmed that the P-Cores & E-Cores on upcoming Meteor Lake & existing Raptor Lake CPUs are very architecturally similar.

wccftech.com

 

[inf64]

After the latest confirmation from intel's VP about the P core in MTL, I think that Arrow Lake will bring a similar ~20% IPC gain vs Raptor Cove (as they did in the past several generations of new cores). ~18-20% seems to be a target that intel had ever since Sunny Cove:

Sunny Cove/Cypress Cove ~18%/19% more IPC
Golden Cove ~19% more IPC
Lion Cove ~20% more IPC?

 

[coercitiv]

After the latest confirmation from intel's VP about the P core in MTL, I think that Arrow Lake will bring a similar ~20% IPC gain vs Raptor Cove (as they did in the past several generations of new cores). ~18-20% seems to be a target that intel had ever since Sunny Cove:

Depends on whether they skipped a core design or not, we still don't know their original plans for Intel 4.

That being said, aiming for a roughly similar achievable target with each iteration is good practice, if you want more you better iterate faster... not harder.

 

[Geddagod]

After the latest confirmation from intel's VP about the P core in MTL, I think that Arrow Lake will bring a similar ~20% IPC gain vs Raptor Cove (as they did in the past several generations of new cores). ~18-20% seems to be a target that intel had ever since Sunny Cove:

Sunny Cove/Cypress Cove ~18%/19% more IPC
Golden Cove ~19% more IPC
Lion Cove ~20% more IPC?

Wouldn't be surprised if they aimed for a larger IPC increase tbh.
Intel 7 to Intel 20A/ TSMC n3 is esentially 2 node jumps worth of logic density gains. Even using Intel 4 would be esentially 2 nodes worth of logic density jumps, because Intel uses HP/UHP in their cores.

Depends on whether they skipped a core design or not, we still don't know their original plans for Intel 4.

Ye, which was why I was speculating that LNC might have been used in redefined GNR.

 

[JoeRambo]

(5M 20c 3.91GHz + 2.61GHz, 3.3GHz IMC, 4x 2.5MB + 4MB L2, 2x 8MB L3)

i think it has two 2xP core clusters with 5MB L2 each + four 4xE core clusters with 4MB of L2 cache each. L3 is reported in a strange way tho, probably problems with parsing info from CPUID. I'd expect this product to have 24MB of L3.

 

[Tigerick]

(5M 20c 3.91GHz + 2.61GHz, 3.3GHz IMC, 4x 2.5MB + 4MB L2, 2x 8MB L3)

i think it has two 2xP core clusters with 5MB L2 each + four 4xE core clusters with 4MB of L2 cache each. L3 is reported in a strange way tho, probably problems with parsing info from CPUID. I'd expect this product to have 24MB of L3.

Actually if based on 3MB per core, LNL should have 15MB L3 cache total....so pretty near to 16MB reported

 

[Henry swagger]

Intel reportedly planning Arrow Lake Refresh featuring 8P+32E cores for 2025 debut - VideoCardz.com

Arrow Lake Refresh with 40 cores Intel’s latest client CPU roadmap has been presented by Moore’s Law is Dead. The video reiterates some rumors and solidifies the latest information about Intel’s plans for the next three years. However, the further we look into rumored and official roadmaps, the...

videocardz.com

Arrow and nova lake use N3b.. hope n3 clocks high i thnk they can hit 6ghz with it if N5 can hit 5.7ghz on zen 4

 

[Hulk]

Intel marketing materials confirm upcoming 6.0 GHz 24-core, 5.6 GHz 20-core and 5.3 GHz 14-core CPUs - VideoCardz.com

Intel teases upcoming desktop CPU upgrade Marketing materials confirm configs of three 14th Gen Core models. The first official marketing materials for Raptor Lake Refresh to leak are not very detailed, but they do confirm the configuration of the new series. The slides show the basic specs for...

videocardz.com

First ever non-KS 6.0 GHz CPU.

How high is Intel gonna fly with a KS? 6.3 GHz?

Looking at the 14th Gen desktop from the bottom up it's pretty much 13th gen until you get to the top of the stack.

The 14600K gets a 200MHz frequency bump to 5.3GHz. No problem assuming yields are good. 14700K gets another Gracemont cluster and a small frequency bump. Again the 13th Gen could be binned accordingly. Even the 14900K appears to simply be a 13900KS.

So we are left with two possibilities.

First, Raptor Refresh is simply a bunch of higher binned 13th generation parts. As for a 14900KS it wouldn't be too hard for Intel to find some silicon with 2 golden cores that can hit 6.2GHz and call is "KS."

Second possibility is the 14th Gen is actually a refinement of Intel 7 and is allowing better clocks. Meaning 13th gen was "binned out" for increasing clocks.

I personally only care about all-core max clocks, one or two core max frequency is only good for advertising. If a 14900K can do 5.8GHz all-core at the power the 13900K could do 5.4GHz then that would be something I guess. Outside of the extra Gracemont cluster on the 14700K the Raptor Refresh in my mind not earning the designation for a new gen. Devil's Canyon = Haswell Refresh = new generation?

I'm hoping I'm wrong on all this and Intel will show us a new gen designation is worth more than 0.2GHz increase in clocks?

 

[igor_kavinski]

I'm hoping I'm wrong on all this and Intel will show us a new gen designation is worth more than 0.2GHz increase in clocks?

I'm getting the 14900KS if it can do 6 GHz all core. Should be a decent CPU to tide me over and take the place of my 12700K until the rumored 2026 16+32+4 Nova Lake (assuming AMD decides not to play the Core Wars and keeps pretending that most people have no use for more than 16 cores).

 

[Doug S]

Intel reportedly planning Arrow Lake Refresh featuring 8P+32E cores for 2025 debut - VideoCardz.com

Arrow Lake Refresh with 40 cores Intel’s latest client CPU roadmap has been presented by Moore’s Law is Dead. The video reiterates some rumors and solidifies the latest information about Intel’s plans for the next three years. However, the further we look into rumored and official roadmaps, the...

videocardz.com

Arrow and nova lake use N3b.. hope n3 clocks high i thnk they can hit 6ghz with it if N5 can hit 5.7ghz on zen 4

There's no way they are using N3B in a product coming out a year from now, it would use N3E. I am willing to bet Intel never uses N3B, and that Apple is quite likely the only company we will ever know for sure used N3B (there may be some bitcoin ASIC that uses it)

 

[Kepler_L2]

There's no way they are using N3B in a product coming out a year from now, it would use N3E. I am willing to bet Intel never uses N3B, and that Apple is quite likely the only company we will ever know for sure used N3B (there may be some bitcoin ASIC that uses it)

ARL was designed years ago, N3B was the only option at the time.

 

[Mopetar]

Didn't Intel sell their N3B wafers to Apple?

I really can't see them having a successful product with it considering how the new iPhone SoC has turned out. Maybe some of that is Apple having to push voltages higher to get more functioning chips, but I don't think Internet would be content having heavily bin chips on such an expensive node and I don't think they could get the higher clocks they want from it.

 

[H433x0n]

Didn't Intel sell their N3B wafers to Apple?

I doubt it. It'd be great if they could but I'm not sure how that'd work considering there's no viable alternative since changing midstream would require porting to the core to a new SDK.

I really can't see them having a successful product with it considering how the new iPhone SoC has turned out. Maybe some of that is Apple having to push voltages higher to get more functioning chips, but I don't think Internet would be content having heavily bin chips on such an expensive node and I don't think they could get the higher clocks they want from it.

I agree, the only hope is that N3B yields improve over the course of a year. I'm not sure how that's possible unless they bring in new equipment such as NXE:3800E and modify the process by removing layers. I don't see yield improving much with the existing tools since they've been at it for 12-18 months at this point.

 

[Geddagod]

Arrow and nova lake use N3b.. hope n3 clocks high i thnk they can hit 6ghz with it if N5 can hit 5.7ghz on zen 4

The slide says Nova Lake uses N2P.

 

[DrMrLordX]

Didn't Intel sell their N3B wafers to Apple?

Here's the chain of events as best I understand it:

1). Intel delayed taking wafers from TSMC
2). TSMC maybe reduced foundry output for all fabs responsible for producing N3B (not really sure here)
3). Apple reduced their wafer request from 80-100k wpm to uh 50-60k wpm (my numbers may be a bit off but you get the idea)

The net effect is that Apple supposedly wound up with all the N3B coming from the research fab that had been retooled solely to provide Intel with N3-family wafers. Along with all the N3B coming from other TSMC fabs.

 

[Henry swagger]

Here's the chain of events as best I understand it:

1). Intel delayed taking wafers from TSMC
2). TSMC maybe reduced foundry output for all fabs responsible for producing N3B (not really sure here)
3). Apple reduced their wafer request from 80-100k wpm to uh 50-60k wpm (my numbers may be a bit off but you get the idea)

The net effect is that Apple supposedly wound up with all the N3B coming from the research fab that had been retooled solely to provide Intel with N3-family wafers. Along with all the N3B coming from other TSMC fabs.

I.m sure tsmc will have enough wafers for intel and apple

 

[DrMrLordX]

I.m sure tsmc will have enough wafers for intel and apple

That is not the point. When it came to N3B, Intel took a hard pass in the form of a delay. They want N3E.

 

[Henry swagger]

Alleged Intel Arrow Lake-S CPU Power Limits Revealed: 125W PL1, 177W PL2, 333W PL4 For 8+16 Core Unlocked SKU

Alleged power limits of Intel's Arrow Lake-S Desktop CPU featuring 8 P-Cores and 16 E-Cores have been revealed by Darkmont.

wccftech.com

Arrow power limits revealed.. tsmc N3 lookinh good 43% more efficient at pl2 than intel 7 node 😃💻

 

[Henry swagger]

Alleged Intel Arrow Lake-S CPU Power Limits Revealed: 125W PL1, 177W PL2, 333W PL4 For 8+16 Core Unlocked SKU

Alleged power limits of Intel's Arrow Lake-S Desktop CPU featuring 8 P-Cores and 16 E-Cores have been revealed by Darkmont.

wccftech.com

Arrow power limits revealed.. tsmc N3 lookinh good 43% more efficient at pl2 than intel 7 node 😃💻

 

[Geddagod]

Alleged Intel Arrow Lake-S CPU Power Limits Revealed: 125W PL1, 177W PL2, 333W PL4 For 8+16 Core Unlocked SKU

Alleged power limits of Intel's Arrow Lake-S Desktop CPU featuring 8 P-Cores and 16 E-Cores have been revealed by Darkmont.

wccftech.com

Arrow power limits revealed.. tsmc N3 lookinh good 43% more efficient at pl2 than intel 7 node 😃💻

Potentially just pretty poor power scaling there, you see a similar thing with the 7950x where power scaling is pretty much dead after 175-200 watts.
I wouldn't actually try extrapolating any hard numbers like you did off of it though, it's really just not worth it

 

[S'renne]

Potentially just pretty poor power scaling there, you see a similar thing with the 7950x where power scaling is pretty much dead after 175-200 watts.
I wouldn't actually try extrapolating any hard numbers like you did off of it though, it's really just not worth it

From the previous leaks it does sound like that instead so they restricted it for "efficiency"

 

[H433x0n]

Alleged Intel Arrow Lake-S CPU Power Limits Revealed: 125W PL1, 177W PL2, 333W PL4 For 8+16 Core Unlocked SKU

Alleged power limits of Intel's Arrow Lake-S Desktop CPU featuring 8 P-Cores and 16 E-Cores have been revealed by Darkmont.

wccftech.com

Arrow power limits revealed.. tsmc N3 lookinh good 43% more efficient at pl2 than intel 7 node 😃💻

Ah. If this turns out to be accurate this is what I’ve been worried about.

RIP overclocking or the ability to cool your CPU.

 

[Henry swagger]

Ah. If this turns out to be accurate this is what I’ve been worried about.

RIP overclocking or the ability to cool your CPU.

You worried about what ? .. intel cpu's are easier to cool than amd"s chiplets

 

[Markfw]

You worried about what ? .. intel cpu's are easier to cool than amd"s chiplets

I don't think so.