Memory bandwidth (200GB/s vs 400GB/s) on a M1X Macbook Pro: usage scenarios where it makes difference?

Solution 1:

You cannot answer your question "is it worth it" unless you know how much your time is worth.

Essentially, the memory bandwidth difference does not change which programs can be run on the computer. You can get the same results from each computer. The only difference is how fast you'll achieve them.

In general you can say that programs that benefit most from that increased memory bandwidth are those have a lot of accesses to the same memory over and over again.

I.e. a file decompression program that loads in 10 GB of data from disk, reads through it once, and then writes it out to disk again, won't see any improvement. The bottle neck is the (relatively) slower disk. Also note that file decompression programs usually do not work like this and will not occupy 10 GB of RAM.

However a program that analyses your data and does so by reading and writing through the same data many times over (without touching disk) - they can benefit.

A subjective view: I don't anticipate the difference in memory bandwidth alone will radically change the user experience of working with programs such as Google Chrome, Microsoft Excel or file decompression - as mentioned in your question. On the flip-side, you mention a $200 added cost over 7 years - that is less than 8 cents a day. If your time is worth $40 per hour that means that if you can save 8 seconds a day because of the upgrade, then it is "worth it".

Note that in addition to the increased memory bandwidth on the M1 Max, there's also a much larger system level cache. It is 48 MB on the M1 Max and 24 MB on the M1 Pro. For most work loads, I think this will matter much more than the increased memory bandwidth.

Solution 2:

Some insights from Anandtech analysis on M1 Max:

  1. CPU alone can max to 243 GB/sec, not more:

M1 Max isn’t able to fully saturate the SoC bandwidth from just the CPU side

Adding a fourth thread [DRAM bandwidth] lands us at 224GB/s and this appears to be the limit on the SoC fabric that the CPUs are able to achieve, as adding additional cores and threads beyond this point does not increase the bandwidth to DRAM at all. It’s only when the E-cores ... are added in, when the bandwidth is able to jump up again, to a maximum of 243GB/s.

  1. Finding how to fully load bandwidth with GPU is still an open question:

...in my testing, I’ve had extreme trouble to find workloads that would stress the GPU sufficiently to take advantage of the available bandwidth.

For actual 3D rendering and benchmarks, I haven’t seen the GPU use more than 90GB/s (measured via system performance counters). While I’m sure there’s some productivity workload out there where the GPU is able to stretch its legs, we haven’t been able to identify them yet.

  1. Bottom line is: only CPU + GPU + media engines [+other parts of the chip] together can possibly utilize the 400GB/sec fully:

Workloads that stress CPU, GPU, media engines all at the same time would be able to take advantage of the full system memory bandwidth, and allow the M1 Max to stretch its legs and differentiate itself more from the M1 Pro and other systems.