SK hynix ships samples of its HBM4E memory: 16Gbps per pin, 48GB capacity per 12-layer stack | Infinium-tech

The AI ​​hardware doesn’t use regular DDR memory – it’s too slow. Instead, the preferred format is high bandwidth memory Or HBM. SK hynix has just announced that it has started delivering samples of the latest HBM4E to its customers.

HBM4E provides 16Gbps bandwidth per pin. For comparison, the previous HBM4 standard did 10Gbps per pin. Samsung began sampling its own HBM4E design a month ago and claims 14Gbps per pin.

The design currently being shipped to SK Hynix’s partners consists of 12 dies stacked on top of each other. This 12-layer design works out to 48GB of capacity – that’s per stack, of course, and most AI accelerator designs will use multiple stacks.

SK hynix claims that its HBM4E memory is 20% more power efficient than the previous HBM4. Additionally, it manufactures the stack using MR-MUF – Mass Reflow Molded Underfill – which uses protective liquid between the silicon dies to protect the circuits on them. The end result is 17% lower heat resistance than the old design, which will help with cooling.

SK hynix’s HBM4E memory does 16Gbps per pin, is 20% more power efficient

SK Hynix wrote in its press release, “The company was able to deliver 12-stack HBM4E samples on schedule thanks to its advanced HBM development and production expertise for HBM. We will work closely with partners to achieve mass production in a timely manner.”

Source