SRAM is not very dense.
replyIMO the next big shift in AI hardware is when some of the upcoming logic-compatible memory types mature. Inference inherently likes many small local pools of memory, with their own interface each. This is impractical with currently available memory other than sram.
It's SRAM. It's the equivalent of 44 GB of CPU cache.
replyExactly, which is why it’s so incredibly fast. And expensive.
replytraditional SRAM memory has 6 transistors per bit. Dense SRAM can have two transistors.
replyWafers can have trillions of transistors, 10-16 times more than the 4 trillion of the WSE-3.
1 Terabyte of SRAM per wafer will be the upper limit for a few years [1].
SRAM doesn't shrink as well as other types of circuitry do when process nodes advance.
replyYou mean it is unimaginably massive.
reply