NAND is very, very slow relative to RAM, so you'd pay a huge performance penalty there. But maybe more importantly my impression is that memory contents mutate pretty heavily during inference (you're not just storing the fixed weights), so I'd be pretty concerned about NAND wear. Mutating a single bit on a NAND chip a million times over just results in a large pile of dead NAND chips.

No it's not slow - a single NAND chip in SSDs offers >1GB of bandwidth - inside the chip there are 100+ wafers actually holding the data, but in SSDs only one of them is active when reading/writing.

You could probably make special NAND chips where all of them can be active at the same time, which means you could get 100GB+ bandwidth out of a single chip.

This would be useless for data storage scenarios, but very useful when you have huge amounts of static data you need to read quickly.

That bandwidth is for the whole GPU, which has 6 chips. But anyways, what I'm proposing isn't for the high-end and training, but for making inference cheap.

And I was somehat conservative with the numbers, a modern budget SSD with a single NAND can do more than 5GB/s read speed.

That bandwidth is for the whole GPU, which has 6 mermoy chips. But anyways, what I'm proposing isn't for the high-end and training, but for making inference cheap.

And I was somehat conservative with the numbers, a modern budget SSD with a single NAND can do more than 5GB/s read speed.