46 ms·
The M1 is using LPDDR4X-4266. It's possible that packages bigger than 8GB just don't exist (8GB was all that was announced just a few years ago, and the RAM ann
by kllrnohj 6y ago
The M1 is using LPDDR4X-4266. It's possible that packages bigger than 8GB just don't exist (8GB was all that was announced just a few years ago, and the RAM announcements usually come far ahead of shipping usages https://www.anandtech.com/show/11021/sk-hynix-announces-8-gb-lpddr4x4266-dram-packages https://www.anandtech.com/show/11021/sk-hynix-announces-8-gb... ).
That'd then be 1 module = 8GB, and 2 modules in a daisy chain or T-topology gets you 16GB, and there just isn't anything more to be done without a bigger memory controller (more channels) or bigger DRAM modules.
- hajile 6y agoSamsung claims to have very large chips (96Gb) in mass production. It really seems like a reason to upgrade next year when the CPU isn't getting much faster. https://www.samsung.com/semiconductor/dram/lpddr4x/ https://www.samsung.com/semiconductor/dram/lpddr4x/
- kllrnohj 6y agoI think you've got a bits vs. bytes confusion there. Samsung is only claiming up to 12GB modules. 96Gb(its) = 12GB(ytes). So that'd get you +50% max RAM on the M1's, which would be a nice offering, but still won't be enough to reach the 32GB+ of the outgoing Intel lineup.
- hajile 6y agoI knew (thus the lowercase b) That's likely the largest that LDDR4x is ever going to go because the tech is outdated. If they had a 24GB option, you'd hear a lot fewer complaints (and the cost difference isn't massive). They announced 16GB (128Gb) LPDDR5 chips were in mass production in Q3 of last year. These both use less power and have better performance than LPDDR4x. Both Intel and Qualcomm have support for LPDDR5 in their chips that launched last year. No excuse for Apple here.