Dear Apple: if you’re going to put an i9 in a MacBook Pro, you need to: 1. have a 150 W power adaptor; 2. make the MacBook Pro an inch thick; 3. have at least two heatpipes each for the GPU and the CPU, and 4. let the fans ramp up early. If not, you’ll have a throttling mess like you do now.
Literally everything that’s wrong with the MacBook Pro with the i9 (and I daresay the i7 too) is in the title.
Let me give some technical background before I address each point, one by one.
The Intel hexa-core, duodeca-thread Coffee Lake mobile CPUs all draw around 70-100 W at their maximum **all-core clock rate**, which range from [3.9 GHz on the 8750H](https://en.wikichip.org/wiki/intel/core_i7/i7-8750h), to [4.3 GHz on the 8950HK](https://en.wikichip.org/wiki/intel/core_i9/i9-8950hk), and [up to 4.5 GHz with Thermal Velocity Boost](https://en.wikichip.org/wiki/intel/thermal_velocity_boost).
Now, first off, the TDP of these CPUs is defined at the base clock, which are a paltry ~2.2—2.9 GHz (I presume the 8950HK is prepared from the best bins, which means those CPUs can achieve 700 MHz greater clocks at roughly equal power draw).
There is a second, and arguably more important nuance to Intel CPU power control, and that is the power limits enforced by the firmware and OS, frequently called ‘PL2’ and ‘PL1’, or ‘Short Term Power Limit’ and ‘Long Term Power Limit’. An [AnandTech article here](https://www.anandtech.com/show/13544/why-intel-processors-draw-more-power-than-expected-tdp-turbo) discusses these power limits (albeit on desktops) in some detail. On custom-made desktops, these can be set arbitrarily high as mentioned in the article (emphasis mine):
> This lets them set PL2 to 4096W and Tau to something very large, such as 65535, or -1 (infinity, depending on the BIOS setup). This means **the CPU will run in its turbo modes all day and all week**, just as long as it doesn’t hit thermal limits.
However, notebooks are smaller, have less thermal headroom, and more importantly, run off flammable batteries that cannot output large bursts of power willy-nilly without suffering electrochemical damage. Therefore, these power limits have to be set to something reasonable so that the notebook stays within spec. If these power limits are reached by the CPU, then there is *power limit throttling*, and the CPU is fixed to run at a certain package power draw. The resultant clock rate can therefore vary depending on the number of cores loaded and the load itself. For instance, one might see the highest clocks in games, slightly lower clocks whilst software ray-tracing or benchmarking, lower clocks still during software transcoding, and lowest clocks in torture tests like Prime95.
Nevertheless, all five 6C/12T CFL CPUs have fairly similar power profiles under load, and at 60W (typically the PL1 limit on many Windows gaming/workstation notebooks), one might see clocks of 3.4-3.8 GHz depending on the bin; at 90W (typically the PL2 limit), one generally exceeds 4 GHz. The 8950HK can draw [in excess of 100 W](http://forum.notebookreview.com/threads/overclocking-the-aorus-x9-dt-v8-w-i9-8950hk-to-the-max.820711/) while running unlocked and at full bore.
Then there comes the idea of thermal throttling. There are two mechanisms. First, there is SpeedStep, which reduces the CPU multiplier as the distance to [TjMAX](https://www.intel.sg/content/www/xa/en/support/articles/000005597/processors.html) temperature decreases, i.e. the CPU approaches its maximum safe operating temperature, which is set by Intel to be 100 °C. This can be offset by manufacturers by up to 15 °C, so certain notebooks might begin throttling even as they hit 85 °C (which are fairly typical load temperatures for notebooks). This multiplier is reduced until the CPU drops below the throttling temperature, and an equilibrium is reached.
Secondly, there’s PROCHOT throttling, which is a penultimate last-ditch measure that cuts the CPU multiplier to an extremely low value (typically ×8 or ×9, for an effective 800-900 MHz core clock), and this happens if the CPU’s temperature *exceeds* the temperature value above, specified by TjMAX. Any higher temperature, and the CPU forces the system to shut down instantly, as a safety measure.
Now that the technical detail is done, let me address the four points one by one.
1. The MacBook Pros are equipped with a **[power adaptor rated to deliver only 87 W](https://www.apple.com/macbook-pro/specs/)**. Furthermore, the USB-C specification denotes maximum power throughout of 100 W through one port. Exactly how is the i9 in the MacBook Pro supposed to achieve power draws of 60-90 W, if the power adaptor cannot deliver the power needed for the CPU alone under load, never mind the whole notebook? Therefore we have our first mechanism: **power limit throttling**, because the CPUs simply *cannot draw enough power*. If the solution by Apple (as was proposed in a similar discussion earlier) is to let the battery discharge while this occurs, then it is a poor band-aid fix to a problem that *shouldn’t exist in the first place.* Larger power adaptors rated 150-200 W have been miniaturised so successfully that the 180 W adaptor powering my current notebook, is actually *both* smaller *and* lighter than the 120 W adaptor I had for a notebook, four years ago. I don’t see why Apple can’t do it. Oh, wait, they can’t, because they ditched MagSafe.
2. Thickening the MacBooks is a no-brainer. The thicker your notebook, the more volume you have for thermal power dissipation; the more volume you have for larger heatsinks to keep your CPU and GPU cool.
3. The MacBook Pro, going by the [latest teardown](https://www.youtube.com/watch?v=IA9VqG4j5kk), has a grand total of *one* heatpipe for *both* the CPU and the GPU. [The VRMs for either aren’t even cooled *at all*.](https://i.imgur.com/RYx5WxN.jpg) Compare to rival notebooks like the [Asus Zephyrus M](https://www.notebookcheck.net/fileadmin/_processed_/8/3/csm_gm501_21_4a46170ace.jpg) (which, oddly enough, is not much thicker than the MacBook Pro), the [Aorus X5 v8](https://www.notebookcheck.net/fileadmin/_processed_/c/6/csm_hardware_323230d127.jpg), or, for a professional workstation, the [Dell Precision 7530](https://www.laptopmain.com/wp-content/uploads/2018/09/Dell-Precision-7530-Workstation-Disassembly-9-696×470.jpg). Even the [Dell XPS 15](https://i0.wp.com/laptopmedia.com/wp-content/uploads/2017/03/IMG_20170227_190933.jpg)—which is itself a throttling mess with the i9, though to less of an extent than the MacBook Pro—has two heatpipes each. Speaking of fans, where the heck are the fan intakes? The tiny slits on the edges of the notebooks? Seriously, Apple?
4. Fans provide airflow. Airflow means heat from the heatsinks is dissipated to the environment quickly. This in turn means heat from the CPU and GPU is dissipated to the environment quickly. There’s no simpler way to put this.
Point 1 leads to **power limit throttling**. Points 2, 3, and 4 lead to **thermal throttling**. Either way, the MacBook Pro is never going to achieve its rated performance.
After all this is said and done, I thoroughly don’t expect Apple to turn back on their ways and suddenly release a MacBook Pro with dimensions from 2011, though that’d be a dream come true. This is more of a warning to potential buyers that they are almost certainly ***not*** getting their money’s worth when buying the MacBook Pro with the i9, and possibly even the i7s (if value-for-money ever amounted to something with Apple products in the first place). There are notebooks where the i9 belongs in, and the MacBook Pro is **not it**.