[YetAnotherNick]

5200 mt/s is 83 GB/s assuming dual channel, which isn't much compared to 100 GB/s for base M2, and 800 GB/s for M2 ultra[1]

[1]: https://en.wikipedia.org/wiki/Apple_M2#:~:text=The%20SoC%20a....

[BearOso]

The higher end M2s just run more chips in parallel. The bandwidth is there to support the GPU.

The latency of LPDDR is higher than desktop RAM, so it relies on the cache to get around that, and the cache is about the same as most x86 CPUs. So there isn't much benefit in the real world.

[gumballindie]

That’s okay given the size. The mobo can handle 8000 mt/s, and I can swap sticks whenever I want. Even my gpu has a sweet 24gb ram - more than an average m2. And I can do more than one if I need to train anything.

[YetAnotherNick]

You are comparing one of the highest end RAM and graphics card to average M2. Don't you think it's bit disingenuous?

[gumballindie]

I am not entirely following. The m2 and m3 “scary fast” specs cost way more than my setup.

[justsomehnguy]

Don't you think it's bit disingenuous to compare only things where your favourite piece has the advantage?