As others already said, there is not much evidence for anything. Apple is quite good at secrecy. However, we do have some evidence. We "know" that the new Mac chips are codenamed Jade Die. We know that Apple has been working on a more powerful GPU core (Lifuka) that is supposed to launch this year.
The code words "Jade" "Tonga" "Lifuka" seems to be more so consistently names for dies not "cores". Or core microarchteture names. Pretty decent chance that Lifuka could be a GPU focused chiplet. So more of the same cores with some cache and inter-die communication mechanism.
Or a die that is just waaaaaaaaay skewed toward GPU cores. ( If Apple is highly focused on monolithic dies to incrementally same power and 2D-3D packaging costs ).
We know that the prosumer chips must be faster than the entry-level chips. And finally, we know that Apple is already making chips on the N5P node. So at tis point we could really get anything.
"Must be faster". Required? the Mini , Air , iMac 24" , and lower end MBP 13" all have just one chip. ( back in the Intel variants that would be 9-12 different cpu speed variations. ). Apple has shifted this more so to picking the better fit container than buying peformance. At very least shifted ti off of geeky CPU benchmark scores being the 'guide'.
What is more necessary is that these be more competative with non-Mac competitors.
Already making doesn't necessarily mean shipping soon. If the "bake" time is 90-100 days than making N5P now means have a sizable inventory in late Aug early September. .... right around yearly iPhone time.
To have stuff shipping in July Apple would have had to start N5P production back a month or so back earlier in the Spring when it was still solidly in the "at risk" (not "High volume") status stage.
It is unlikely that Apple is going to pull N5P wafer starts from the iPhone SoC for Mac SoC to cause a slide in iPhone release. As long as the iPhone is on a strict 12 month release cycle they are probably going to get higher priority. [ Apple has shown zero rigidly hard commitment to 12 month cycles for the Mac . Folks trying to throw all the blame on Intel but MBA sliding for 2-3 years . MP 3, 6 years snooze cycles , Mini snooze cycle. etc. That wasn't all Intel. ]
N5P is very highly likely A15. The "don't know" part is if there is a relatively small 'side piece' of those wafers that got allocated to Mac SoCs.
They are world class in their bracket, true. But what about prosumer hardware? In single-threaded performance M1 can match fastest x86 cores out there, but that's about it. Intel just brought out Tiger Lake refresh with higher single-core turbo. They are bringing a new CPU architecture (Alder Lake) this fall. AMD just showed a chip with 3D-stacked 192MB cache the other day. New entry-level Nvidia and AMD GPUs are coming to the laptops,
So they are "winning" on single thread performance so need to add efatures to boost single thread and limit power (e.g., the obj_c_ dispatch prediction discussed earlier in the thread). That isn't going to help much. If the core count on 10-16 on the AMD/Intel side then the primary missing piece is "more area for more stuff " far more than fewer "more magical" cores.
The mainstream x86 market primarily leverage incrementally higher clockspeeds to charge higher pricing. If Apple is solidly on this "world best" iGPU kick then they have a another lever to charge higher prices on. (and already do in the MBP 16" and iMac 27" with dGPUs.). Customer needs a bigger GPU ... pay more. Need a bigger integrated screen ... pay more. More CPU cores ... well pay more for more GPU cores too. Need more RAM... pay more only from us as a source (and buy up front because pragmatically can't add more later).
In short, they have a huge 'hook' to generate more revenue in the "max integrated" aspects of the SoC design.
10 cores ( split 8 P 2 E ) probably quite well against most , if not all, of Alder Lake ( which maxes out at 8 "big"). Apple isn't particularly behind there is scale up the memory subsystem to keep those fed with data (and keep the hit rate approximately the same. )
It's not enough for Apple to produce a CPU that consumes much less power. It's great for a MacBook Air, not so much for a Pro.
Presuming MBPro here. Leaks so far point to Apple "half sizing" the Mac Pro with the transition. That is less power ( will result in less high power add in cards).
Certainly not on the iMac 24" which had its cooling capability crippled relative to the 21.5" model.
They have to show that their product is faster, better, more efficient. They need to be faster than anything that Intel or AMD can put in a laptop (or even in a compact desktop). Can they do it with Firestorm? I don't know.We have seen no evidence that Firestorm can reach more than 3.2ghz.
Cranking the clock higher isn't gong to help if can't keep the core fed. Apple probably isn't going to go after the desktop-replacement-luggable market. Mostly likely they are going to compete on "best performance solely on battery". Unplug you high end AMD laptop and the performance takes a hit. And Adler Lake at max configuration , at max compute is likely in a different zip code to battery consumption.
Apple didn't do that "move from desk to desk" laptop product before. Why would they be trying to do that now with M-series.
A new architecture, designed for performance, with more execution units, more cache, higher clocks, deeper execution window, that could do the trick though.
Like the current one isn't design for high single threaded performance. Higher clocks with LPDDR4 (or even DR5) memory isn't a good match. More cache and more execution units doesn't necessarily require a significant jump in core design. The "wider" is more likely to just do that to the memory subsystem on that same "LPDDR" path.
Pretty likely these same cores are heading the for the A15 with a different uncore on the die when Apple does do the N5P iteration.
Apple is quick unlikely to jump into the desktop, high end CPU overclocker competition. Largely eschewed overclocking support with INtel chips. Not likely they are putting windows in for thst with their own stuff. So those "top fuel dragster" single thread speed freak.... are going to be able to push more out of the x86 with edge case, exotic set ups.
