Contribute
Register

October 18th 2021 Apple Event: M1 Pro/Max MacBook Pros

Status
Not open for further replies.
There is also the rumor of a more compact Mac Pro. By the way, Snazzy Labs just released a video (yes, another glowing impressions video) that concludes with this:

View attachment 531356
There is also the rumor of a more compact Mac Pro. By the way, Snazzy Labs just released a video (yes, another glowing impressions video) that concludes with this:

View attachment 531356

Not sure if the OS and programs can scale so well to use all those cores… sometimes the old 5.3ghz boost is preferable… but maybe Monterey is helping in using all those parallel resources.
 
There is also the rumor of a more compact Mac Pro. By the way, Snazzy Labs just released a video (yes, another glowing impressions video) that concludes with this:
Not sure I could withstand 20 minutes of YouTube silliness by a hipster trying to look young and cool…

M1 Max is 432 mm2. Twice that would be slightly larger than Xeon Scalable (about 700 mm2), EPYC (750 mm2) or the largest GPU dies. Four times that as a monolithic design, at over 1300 mm2, would take a manufacturing miracle. I doubt that Apple and TSMC can do that.
Having two, or four, M1 dies on a motherboard, which each die attached to its own memory, would mean that the memory architecture is not uniform. I know about NUMA; I'm not sure how NUUMA (Non-Uniform Unified Memory Architecture) would work.
Jade is probably something different than two or four M1 Max stitched together.
 
Not sure I could withstand 20 minutes of YouTube silliness by a hipster trying to look young and cool…
Snazzy Labs is one of my favorites on YouTube. His content is predominantly more educational than marketing fluff. This is an impressions video, however. He is typically clean shaven, but his look does not detract from his message.

M1 Max is 432 mm2. Twice that would be slightly larger than Xeon Scalable (about 700 mm2), EPYC (750 mm2) or the largest GPU dies. Four times that as a monolithic design, at over 1300 mm2, would take a manufacturing miracle. I doubt that Apple and TSMC can do that.
Having two, or four, M1 dies on a motherboard, which each die attached to its own memory, would mean that the memory architecture is not uniform. I know about NUMA; I'm not sure how NUUMA (Non-Uniform Unified Memory Architecture) would work.
Jade is probably something different than two or four M1 Max stitched together.
M1 uses discrete memory chiplets connected to the main computing complex in a single SoC package. AMD’s Ryzen architecture has shown that multiple core complexes can be tightly integrated into a single SoC. Apple Silicon will most likely follow a similar path, as will Intel. Whether the core complexes are interconnected on the same X/Y plane or stacked along the Z axis (e.g. AMD’s stacked V-cache coming soon), the future of scaling, at least on the consumer front, appears to be along these lines (i.e., not multiple sockets).
 
Last edited:
They probably see the writing on the wall.
They probably do...
...which reminds me of this. Anyone remember that?

pentium2snail.jpg



Now updated for 2021

corei9snail.jpg
 
They probably do...
...which reminds me of this. Anyone remember that?

View attachment 531373


Now updated for 2021

View attachment 531402

There's no question that the loss of X86/X64 compatibility is a disadvantage. It's unfortunate that, even to this day, so many games are Windows only. Hopefully, there will be enough adoption of Apple Silicon by users that game publishers are forced to follow the money and bring their titles to macOS.

Perhaps the current GPU shortages and insane prices will help people see that the M1 Pro/Max represent a fairly decent value proposition when cost of GPU is factored in.
 
Not sure I could withstand 20 minutes of YouTube silliness by a hipster trying to look young and cool…

M1 Max is 432 mm2. Twice that would be slightly larger than Xeon Scalable (about 700 mm2), EPYC (750 mm2) or the largest GPU dies. Four times that as a monolithic design, at over 1300 mm2, would take a manufacturing miracle. I doubt that Apple and TSMC can do that.
Having two, or four, M1 dies on a motherboard, which each die attached to its own memory, would mean that the memory architecture is not uniform. I know about NUMA; I'm not sure how NUUMA (Non-Uniform Unified Memory Architecture) would work.
Jade is probably something different than two or four M1 Max stitched together.

Maybe the desktop versions of Apple Silicon will go to 3nm...

Also, 1300mm^2 is nothing compared to...
Screen Shot 2021-10-20 at 7.17.39 AM.png

Source:https://www.hardwaretimes.com/leake...max-tdp-of-700w-and-zen-4-chiplet-dimensions/
 
Last edited:
I thought to share this interesting piece of info I managed to read earlier today.
It is a table that shows the actual power efficiency of the M1 chip vs the new M1 Pro and M1 Max in normal use.

m1efficiency.png


It seems the 1st M1 chip is actually incredibly efficient compared to the two newer chips!
 
I thought to share this interesting piece of info I managed to read earlier today.
It is a table that shows the actual power efficiency of the M1 chip vs the new M1 Pro and M1 Max in normal use.

View attachment 531438

It seems the 1st M1 chip is actually incredibly efficient compared to the two newer chips!
This should add a little perspective! :)

Chip Transis- tors CPU Cores GPU Cores Media Engines Memory Memory Type
Thunder- bolt
Dis- plays
NVMe Speed
M1 16 billion 4P, 4E 7 or 8 No ProRes 8GB or 16GB LPDDR4X 4266MT/s
TB 3​
1​
PCIe 3​
M1 Pro 33.7 billion 6P, 2E
8P, 2E
14 or 16 1 ProRes 16GB or 32GB LPDDR5 6400MT/s
TB 4
2​
PCIe 4
M1 Max 57 billion 8P, 2E 24 or 32 2 ProRes 32GB or 64GB LPDDR5 6400MT/s
TB 4
4​
PCIe 4
 
Last edited:
The need versus want argument continues! Fortunately, Luke introduces a compromise between the two, justify. Alas, I agree with everything he says!

 
Last edited:
Status
Not open for further replies.
Back
Top