Parts/Memory/HBM2
Memory · HBM2 · 2016
HBM2
The first HBM generation to reach volume production outside a single vendor. Used in P100, V100 and A64FX.
Why this part
The generation that took stacked memory from a single vendor's experiment to an industry standard three foundries could build to. AMD's Fiji GPU used the prior HBM1 in 2015 in a Hynix-only supply chain; HBM2 is what let NVIDIA, Fujitsu and later AMD again all specify the same memory type from competing suppliers rather than depend on one. P100 and V100 both used it, and A64FX's entire architecture, no DDR interface at all, only became viable once HBM2 could supply enough capacity per stack to hold a real working set rather than a fraction of one.
Systems using this part
No systems in the current dataset record this part. It is carried in the taxonomy so the graph is complete, and so edges can attach to it as coverage grows.