View attachment 33180
Full M4 (orange) analysis in the thumbnail above (click to enlarge) based on NotebookCheck CB R24 power and performance data reveals a few interesting insights. But to start with some caveats, unfortunately NotebookCheck did not publish CB R24 power data for the base M4. Using CB R24 performance and CB R23 performance and power data for M4 Pro and M4 Max, I estimated the power for the M4 under CB R24. Further, I subtract idle power for the devices, so my efficiency numbers will differ slightly from NotebookCheck's.
For ST data, we see very similar efficiency numbers for the base M4 as for the base M3. While the M4 does use more power, it also gains a commensurate amount of performance. As the SOC becomes larger however, the power needs go up without any increase in performance and ST efficiency goes down substantially. This likely due to powering the additional fabric, bus, and RAM of the larger SOCs. Using CB R23 power data and comparing to RAM capacity, bus, and bandwidth we do see a big jump when the bus is 256b or larger and the RAM capacity is 48MB or larger, but smaller increases after that.
View attachment 33182
For MT data, the base M4 has a more similar profile to the Snapdragons than previous M3 - manufactured on a better node, but only 4 performance cores and 6 efficiency cores versus 12 performance cores. Still it represents a substantial power increase over the base M3 but with a nearly linear increase in performance such that efficiency is still roughly the same. This increase in power has been noticed with larger fan noise under load for the Mini along with its smaller chassis.
Speaking of linear increases in power and performance, the 12, 14, and 16-core (8P+4E, 10P+4E, 12P+4E) Brava and Brava-chop dies have nearly identical MT efficiency as 2 P cores are added at each step up. Interestingly the 12-core model uses similar power to the old M2 Pro/Max CPU, but with a new node and new architecture, it nets 36% more performance for an increase of 34% in efficiency. The 14-core Pro and 16-core Max stand alone in the current mobile performance landscape. For the Max NotebookCheck had to break out desktop chips or an old overclocked Intel gaming laptop to compare to for performance with the latter in particular only able to match the Max's CPU performance drawing well over 300W even after idle. Arrow Lake and Zen 5 desktops fared little better, scoring around 1800-2400pts and getting less than half the efficiency of the Max chip even after substantial idle draws were subtracted out.
Upcoming Arrow Lake and Strix Halo laptop CPUs should be more efficient than their desktop counterparts, but may not match the M4 Max in performance - in particular for the Strix Halo which will top out at 16 cores, only 50% more than the Strix Point CPU used above. It may have higher clocks (though may also be across two dies), but we can already see the HX 370 plateau at around 1200 with 12 cores. While not true for ST tasks, for MT purposes, all the Zen 5c cores are practically P-cores (the biggest change between "c" and non-"c" cores is that the former are missing the circuitry needed for high clocks). Thus it seems unlikely to that Strix Halo chips will offer enough of a jump in performance to quite match the Max's score. They are much more likely to equal and then exceed the 14-core Pro's performance, but almost certainly at higher power draws - like above 90-110W or so. Arrow Lake laptop chips, even if more efficient than their desktop counterparts will also likely require a large number of cores and high power draws to match the M4 Max's MT performance.
Notebookcheck has tested the MacBook Pro 16 2024 with the M4 Max, which has 16 CPU cores and 40 GPU cores.
www.notebookcheck.net
Notebookcheck has tested the brand-new Apple MacBook Pro 14 with the M4 Pro, 48 GB RAM, Thunderbolt 5 and a matte nano-texture display.
www.notebookcheck.net
Extensive tests and benchmarks of the MacBook Pro 14 with the M4 Pro 12-core SoC
www.notebookcheck.net
Notebookcheck has tested the new Apple Mac Mini with the M4 SoC, 16 GB RAM and a 256-GB SSD.
www.notebookcheck.net
Notebookcheck has tested the new Apple MacBook Pro 16 with the M4 Pro SoC, 48 GB RAM and a matte nano-texture glass panel.
www.notebookcheck.net
Apple has presented its new MacBook Pro models featuring the new M4 processor generation and we have taken a look at the new M4 Pro SoC. Aside from its pure performance, we are interested in how efficient it is. Has Apple managed to further extend its lead over Qualcomm, AMD and Intel?
www.notebookcheck.net
Notebookcheck takes a look at the new AMD Zen 5 Ryzen AI 9 HX 370 processor and compares it with the Intel Core Ultra, Apple M3 Pro and Qualcomm Snapdragon X Elite.
www.notebookcheck.net
Chip expected to feature up to 16 Zen 5 cores, Radeon 8600S iGPU, and 40 Compute Units.
www.tomshardware.com
EDIT: the 12-core M4 Pro has the same configuration as the M2 Pro (8+4). Previously I wrote that it had 2 more P-cores in its explanation for its superior performance. It does not. That means it's superior performance at the same power is all down to improved node and architecture.
www.macrumors.com
- also the reasoning seems weird "to focus on the server AI chip" - I would've thought it was the server AI chip? So is the AI chip different or can Apple not make enough to fulfill demand and sell to customers? or is The Information full of it?
Still should be able to hit M4 Pro levels in CB R24, but unless it follows the Desktop chip design on N4X rather than the mobile design on N4P (in which case it’ll use a lot of power), I doubt it’ll hit anywhere close to the Max. The TDP is rumored to be 120W (which means more than that in practice and I’m presuming that’s all usable for the CPU, ie the considerable GPU is separate) which is a fair bit more than the official 54W TDP of Strix Point (HX 370, which actually uses more than 
