Intel "Raptor Lake Refresh" Confirmed with 14th Gen Core Model Numbering

[btarunr]

Intel is giving its "Raptor Lake" client processor architecture an update in the second half of 2023, called simply "Raptor Lake Refresh." When we first heard about this development back in December 2022, there were two theories on how Intel could name these processors. The first one suggested that it would take the same path as "Coffee Lake Refresh" (9th Gen Core), and give "Raptor Lake Refresh" a whole new generational number scheme (14th Gen Core); while the other held that it would try to carve out new processor model numbers within the 13th Gen, like it did with "Haswell Refresh" (4th Gen Core).

Channel vendors in Taiwan are being communicated that the desktop "Raptor Lake Refresh-S" and mainstream notebook "Raptor Lake Refresh-H" will be the final generation of Core i processors (to retain the current nomenclature), and will be slotted as 14th Gen Core. The top desktop SKU could be named "Core i9-14900K," for example. The ultraportable "Raptor Lake Refresh-U" will be the first generation with the new nomenclature Core branding, while "Meteor Lake" will receive the Core Ultra branding, as it has next-generation CPU cores, iGPU, and an on-package AI accelerator.



As for what separates "Raptor Lake" apart from "Raptor Lake Refresh," we have a theory. Intel could implement the final refinements of its Intel 7 foundry node for the series; and generationally improve CPU core counts and cache sizes. The non-K 13th Gen Core i5 processors, for example, use the older generation die with 1.25 MB L2 cache per P-core, and 2 MB shared L2 cache per E-core cluster. Perhaps the 14th Gen Core i5 could be based on the newer silicon with 2 MB L2 cache per P-core, and 4 MB shared L2 cache per E-core cluster. We're not quite sure how Intel will segment the Core i7 and Core i9 SKUs. For 9th Gen "Coffee Lake Refresh," Intel had bumped up CPU core counts with the introduction of a new 8-core monolithic die, which allowed it to dial up CPU core counts for its Core i9 and Core i7 SKUs. The 10th Gen "Comet Lake" saw it introduce a 10-core die and the final refinements of the 14 nm process, to dial up CPU core/thread counts across the board.

View at TechPowerUp Main Site | Source

 

[Nanochip]

Technically rocket lake was the final iteration in 14nm (a backport of a 10nm design).

 

[N/A]

Technically rocket lake was the final iteration in 14nm (a backport of a 10nm design).

this is raptor lake now.
rocket is on socket 1200

 

[A&P211]

14th will just be a small ghz increase. 10nm++

 

[kondamin]

I knew it wasn’t going to be ground breaking, but I’m still disappointed they are wasting an entire generation on this minor of a tweak.

 

[Leavenfish]

I knew it wasn’t going to be ground breaking, but I’m still disappointed they are wasting an entire generation on this minor of a tweak.

The only thing they are 'wasting' is the term"14th Gen"...it's just a number/alpha...don't lose sleep over it.

 

[Taisho]

14th will just be a small ghz increase. 10nm++

Cache increase for the 400 model is very likely. 13400 was bad enough. Got myself a 12400 (new, integrated GPU, H0 stepping) for 138 euros. I can boost it to basically even 5 GHz in times of need on my AsRock B760 Riptide. 13400 is a complete disaster, lagging miles behind the 13600. Ridiculous price, no boost with external clock generator.

 

[Six_Times]

So, if I understand correctly, the only way to know if the cpu is meteor lake based architecture before you buy is if the laptop or retail (S) box says "ULTRA 7" is that correct?

 

[chrcoluk]

New voltage regulator not mentioned? Thats a pretty big feature if it gets enabled in the refresh.

 

[Nanochip]

this is raptor lake now.
rocket is on socket 1200

I know but the article says
“The 10th Gen "Comet Lake" saw it introduce a 10-core die and the final refinements of the 14 nm process, to dial up CPU core/thread counts across the board.”

my point is simply that rocket lake 11th gen was the final 14nm design.

 

[bobbybluz]

Still using socket LGA 1700?

 

[pressing on]

Still using socket LGA 1700?

Yes, with BIOS updates probably for all 600 and 700 series motherboards like the Raptor Lake update. An upgrade on the H610, the H710 has been rumoured.

So, if I understand correctly, the only way to know if the cpu is meteor lake based architecture before you buy is if the laptop or retail (S) box says "ULTRA 7" is that correct?

Yes, which was probably the whole reason for the change in name. It could also read ULTRA 5 or ULTRA 9 but there don't seem to be any ULTRA 3 devices planned.

 

[InVasMani]

I thought the earlier rumors were indicating they wouldn't bump up core count!?

I was always under the impression that Intel would bump core count, clock speed, memory support, and introduce that voltage regulator. The cache is another area where I think they could make changes, but I don't know if they will or not and just a bump to core count will add a bit more cache, but who knows if they'll try to increase the relative cache per core P core or per E core cluster. Perhaps another thing Intel will have adjusted is speedstep again to further refine it.

I still think if Intel did shared cache in P core clusters of 2 cores similar to the E core clusters with it's cache it would be a positive tweak. Doubling the HT thread count on P cores I could see as being a good direction potentially at least then the threat count on a P core would synchronize with the amount of cores in a E core and who knows maybe that works better with OS scheduler in turn.

I wonder if CPU cache's make use of PAM4 signaling or not!? Kind of weird implications if they don't already actively do so especially in the case of like X3D with a huge cache.

 

[nguyen]

should be kicking ass with 7200MT+ DDR5

 

[ratirt]

I wonder how many weak cores it will have. I guess this is the one thing that will change or maybe not.

 

[Gica]

Cache increase for the 400 model is very likely. 13400 was bad enough. Got myself a 12400 (new, integrated GPU, H0 stepping) for 138 euros. I can boost it to basically even 5 GHz in times of need on my AsRock B760 Riptide. 13400 is a complete disaster, lagging miles behind the 13600. Ridiculous price, no boost with external clock generator.

I don't really see why you would upgrade from 12400F to 13400F. It brings a noticeable plus in multitasking, but in gaming the difference is small even with the RTX 4090.
And I don't understand what is the point of the comparison in this topic. If you are satisfied with 12400 now, in 2-3 years you can upgrade to a more powerful.

 

[Mahboi]

Are they calling it Clever Girl Lake?

 

[R0H1T]

No but probably ~ Intel: we need all your subsidies & a spare kidney

 

[Taisho]

And I don't understand what is the point of the comparison in this topic. If you are satisfied with 12400 now, in 2-3 years you can upgrade to a more powerful.

The point is that if Intel doesn't come up with a real upgrade for the 400 model, instead of just increasing the two starting digits in product name, informed consumers will still be buying the 12400 until it goes out of stock.

 

[ikjadoon]

Alder Lake++ it is, then.

Intel will never get their uArch cadence right. They just can’t admit even Pat’s timelines were misleading optimistic.

Funnily enough that Arm releases 10%+ IPC increases every year and their perf / GHz is already so much higher.

 

[Gica]

The point is that if Intel doesn't come up with a real upgrade for the 400 model, instead of just increasing the two starting digits in product name, informed consumers will still be buying the 12400 until it goes out of stock.

13400 versus 12400
+ 200MHz P cores
+ 100MHz igp
+ 4 E cores

It is not worth upgrading from 12400 to 13400, but many may find 13400 interesting. In many multitasking applications, the 13400 is up to 30% faster.

 

[Dirt Chip]

So 14900k=13900ks?
14900ks=?+100w

 

[wheresmycar]

@btarunr, no mention of DLVR or was this never intended for the refresh? I understand this is a secondary reg to soften up on CPU VID/core power consumption.

If anything, if Intel can achieve 20-25% reduction in power and preserve or increase performance by some measure... that would be a big win! Hope the patent is not locked to mobile parts only, we "want" increased efficiency on the desktop. MORE efficiency, MORE cache, MORE clock freq and MORE memory^mhz = perfect combo for a 'refresh' and a welcomed addition!

 

[InVasMani]

@btarunr, no mention of DLVR or was this never intended for the refresh? I understand this is a secondary reg to soften up on CPU VID/core power consumption.

If anything, if Intel can achieve 20-25% reduction in power and preserve or increase performance by some measure... that would be a big win! Hope the patent is not locked to mobile parts only, we "want" increased efficiency on the desktop. MORE efficiency, MORE cache, MORE clock freq and MORE memory^mhz = perfect combo for a 'refresh' and a welcomed addition!

Perfect would be more cores as well, but even if they just do those other things it's still better than the current lineup. Just having more cache, and memory support with increased efficiency alone would be a general improvement. It would be great if BCLK OC were possible again as well, but seems rather doubtful.

 

[R0H1T]

More cache isn't happening without more cores, they'd have to redesign the chips for that. Think of 6xxx till 10xxx what changed besides higher cache on them?