Verne Arase

Having a primary driver with a supported OS is important to me - even though I'm retired now.
 
The only thing the M1 Max brings you (over the M1 Pro) is more GPU cores and double the encoding blocks in the media engine (as well as double the ProRes decoders), and double the memory bandwidth for the CPU and IP blocks.
 
If you're considering the 2021 MacBook Pro, the 16" does bring with it much better thermal headroom along with the screen real estate, better speakers, and a big gorgeous display (along with the increased weight, of course).
 
I find it easier to spring for Apple hardware since I finally put some coin into Apple stock in 2014 when I realized how much my Apple habit was costing me and the missus told me I should put my money into something I believe in - I've owned Apple continuously since then except for a brief stint where I jumped out at the start of the pandemic when Apple fell from around 210 to 150 due to production halt, with an aggregate appreciation of around 981% for a yearly appreciation of about 151% per annum. The wife gives me much less flack about buying Apple hardware nowadays 😊.
 
The 2021 MacBooks have had the best increase in performance I've ever seen over a two year period - though it looks like high end x86 is getting better too. Only difference is that on the x86 side, like always you're getting a portable desktop with a humongous power supply, whereas with the MacBooks you get performance and portability (and the ability to run for a long time on battery). With the x86 models, your performance dives off a cliff when you're running on battery - something that channels like LTT sort of gloss over.

The M1 Max performs better in the 16" MacBook Pro - the 14" is clock locked due to thermal constraints.
 
If you must get a M1 Max with a 14", I wouldn't go for more than the 24 core GPU.
 
Firefox shouldn't have the ability to drive the CPU to 100% on multiple cores and set the machine non-dispatchable - probably a bug. I'd see if it still happens with 12.1. These machines are pretty new and they're probably still working the hairballs out of 'em. The latest release of Handbrake (1.5.1 as of this post) allows the M1 variants to use the media engine by specifying use of the VideoToolbox.
 
Remember, these puppies are using a brand new architecture and there are likely to be a few bumps in the road (though I haven't experience any beachballs). I've removed BitDefender from my Monterey Macs as there seems to be something amiss. BitDefender insisted on quarantining a bunch of Win files on my Crossover installation.

Tried to leave this information on a YouTube comment, but you appear to be blocking my posts.
 
The latest snapshot build of Handbrake seems to fix the Media Engine access via the VideoToolbox API - try it at:
 
https://github.com/HandBrake/handbrake-snapshots/releases/tag/mac
 
My testing has shown it improves a 1080p transcode of a 1:48:09 MKV RIP to 1080p HEVC from 45 minutes to 9 minutes.
 
You make the point that putting additional silicon support of IP blocks like the Media Engine is not the way to go, but when you've already produced eight wide high performance cores, going nine wide isn't going to yield much benefit. I'm sure the Apple Silicon team raised a toast when they could get eight instructions to execute simultaneously. There is always a point of diminishing returns.
 
You have to realize that Apple is not a traditional CPU or GPU manufacturer, and their silicon design team does not follow the tried and true path of those silicon developers. Apple's Silicon team does not strive to make the fastest CPU or GPU, but rather optimization of the product pipeline and reduction of pain points for their users.
 
When one route is blocked, they'll take another to try to optimize their product's performance.
 
What is a GPU other than an IP block attempting to optimize graphical workflows?
 
Any way, the latest Handbrake snapshot build appears (at least in my testing) to produce a 5x speed improvement on the M1 Max in my 2021 MacBook Pro 16", M1 Max with 32 GPU cores, 32 GB RAM, 2 TB SSD configuration.
 
I at this time have no idea when commit ce52b4d755a2799a4801f256622b8e8191e71220 will make it into generally available versions of Handbrake.

Tried to leave this information on a YouTube comment, but you appear to be blocking my posts.
 
The latest snapshot build of Handbrake seems to fix the Media Engine access via the VideoToolbox API - try it at:
 
https://github.com/HandBrake/handbrake-snapshots/releases/tag/mac
 
My testing has shown it improves a 1080p transcode of a 1:48:09 MKV RIP to 1080p HEVC from 45 minutes to 9 minutes.
 
You make the point that putting additional silicon support of IP blocks like the Media Engine is not the way to go, but when you've already produced eight wide high performance cores, going nine wide isn't going to yield much benefit. I'm sure the Apple Silicon team raised a toast when they could get eight instructions to execute simultaneously. There is always a point of diminishing returns.
 
You have to realize that Apple is not a traditional CPU or GPU manufacturer, and their silicon design team does not follow the tried and true path of those silicon developers. Apple's Silicon team does not strive to make the fastest CPU or GPU, but rather optimization of the product pipeline and reduction of pain points for their users.
 
When one route is blocked, they'll take another to try to optimize their product's performance.
 
What is a GPU other than an IP block attempting to optimize graphical workflows?
 
Any way, the latest Handbrake snapshot build appears (at least in my testing) to produce a 5x speed improvement on the M1 Max in my 2021 MacBook Pro 16", M1 Max with 32 GPU cores, 32 GB RAM, 2 TB SSD configuration.
 
I at this time have no idea when commit ce52b4d755a2799a4801f256622b8e8191e71220 will make it into generally available versions of Handbrake.

Tried to leave this information on a YouTube comment, but you appear to be blocking my posts.
 
The latest snapshot build of Handbrake seems to fix the Media Engine access via the VideoToolbox API - try it at:
 
https://github.com/HandBrake/handbrake-snapshots/releases/tag/mac
 
My testing has shown it improves a 1080p transcode of a 1:48:09 MKV RIP to 1080p HEVC from 45 minutes to 9 minutes.
 
You make the point that putting additional silicon support of IP blocks like the Media Engine is not the way to go, but when you've already produced eight wide high performance cores, going nine wide isn't going to yield much benefit. I'm sure the Apple Silicon team raised a toast when they could get eight instructions to execute simultaneously. There is always a point of diminishing returns.
 
You have to realize that Apple is not a traditional CPU or GPU manufacturer, and their silicon design team does not follow the tried and true path of those silicon developers. Apple's Silicon team does not strive to make the fastest CPU or GPU, but rather optimization of the product pipeline and reduction of pain points for their users.
 
When one route is blocked, they'll take another to try to optimize their product's performance.
 
What is a GPU other than an IP block attempting to optimize graphical workflows?
 
Any way, the latest Handbrake snapshot build appears (at least in my testing) to produce a 5x speed improvement on the M1 Max in my 2021 MacBook Pro 16", M1 Max with 32 GPU cores, 32 GB RAM, 2 TB SSD configuration.
 
I at this time have no idea when commit ce52b4d755a2799a4801f256622b8e8191e71220 will make it into generally available versions of Handbrake.

Wow ... what a thoughtful, insightful post.
 
I especially liked your detailed analysis and meticulous evidence chain.

Let's see: A11 Bionic SoC has a CPU with two high powered cores, each 25% faster than the A10, and four high efficiency cores, each 70% faster (and twice as many). They have an improved M11 motion co-processor which means better gyros and motion sensing, a new neural engine capable of 600 billion AI ops/sec, a new GPU 30% faster and using (like the above) less energy. They have an improved ISP which has hardware noise reduction, and a video processor which splits the image into 2 million tiles and monitors for motion, edge detection and the like. They have a new processor controller which dispatches the CPU cores from light use all the way up to full blazing speed - which, unlike the A10 which could only dispatch the two high power _or_ the two high efficiency cores - can dispatch all cores.
 
All models now have Qi wireless charging and fast USB-C charging as long as you've got a USB-C to lightning cable and a USB-C power source. The primary wide angle camera is a f/1.8 OIS camera with deeper pixels and shoots in the DCI-P3 color space.
 
The 8 has 2GB and a 4.7" 750p screen with a True Tone DCI-P3 color IPS display, 3D touch, and HEIF and HEVC support (half the space at better quality), slow-mo at 240 fps @1080p, up to 60 fps @4K, IP67, 25% louder speakers with deeper base, and TouchID.
 
The plus has everything above and 3GB, a 2x f/2.8 telephoto (non-stabilized), a bigger battery, a 5.5" 1080p display, and portrait mode and portrait lighting on the rear cameras.
 
The X has everything above and a 5.8" 2436x1135 True Tone DCI-P3 OLED HDR display, a f/2.4 OIS 2x telephoto, a larger battery (I think), no TouchID and a TrueDepth FaceID front-facing camera (which is essentially a miniaturized Xbox Kinect). This phone offers portrait mode and portrait lighting on both front and rear cameras using stereoscopic depth on the rear camera and TrueDepth on the front camera.
 
Whew!

Considering the snark they used announcing the Kirin 970, that better be some hot SoC.