QuoteQuad-Channel
Wrong, According to the big AIs / LLMs over at lmarena.ai, LPDDR5X uses 16 bits per channel (never 32 bits), and if it's 16 bits per channel, then "Quad-Channel" would mean that the bandwidth is:
41.6 GB/s = 4 channels * 16 bits per channel * 5200 MT/s / 1000 / 8
41.6 GB/s is obviously incorrect, as AIDA64 memory test measures 70 GB/s.
As a matter of fact, it's 128-bit, one can also state that it's octa-channel (8 * 16 bits).
Desktop DDR5 RAM has 64 bits per channel, so it's different vs LPDDR5X.
If you'd just state that it's 128-bit, then it would be absolutely clear.
83.2 GB/s = 128*5200/1000/8.
The theoretical 83 GB/s, minus the overhead etc., perfectly align with the measured 70 GB/s.
Quote- from 16GB to 32GB RAM (graphic softwares and a LOT of multitasking)
Now we can see how all 16 GB RAM laptops/devices (have) become e-waste with the soldered RAM, accelerated by OS's own LLM AI feature requirements, too.