With the rise of Artificial Intelligence, the demand for server accelerator chips designed to process different models and technologies has multiplied, and this has made HBM memories take on much greater importance and their development is growing by leaps and bounds.
Micron, one of the leading manufacturers of these memories, has announced its new generation of 12-layer or “height” HBM3E chips with a capacity of 36 GB each. They achieve bandwidths of 1.2 TB/s thanks to a transmission speed of 9.2 GB/s per each of their connection pins.
This represents a 50% increase in capacity over other 8-layer HBM3E chips. Expanding the capacity allows for loading into memory larger models such as LLaMa 2 and its 70 billion parameters.
Micron’s 12-layer HBM3E memory chips are manufactured at TSMC after joining the Taiwanese manufacturer’s “3DFabric Alliance.” They will be integrated into GPUs and accelerators specific to artificial intelligence and other uses. This type of memory can be integrated directly into SoCs using 3D vertical stacking technology (hence the term “layers” or “heights”) and achieve high density in a small space.
End of Article. Tell us something in the Comments!
Antonio Delgado
Computer Engineer by training, editor and hardware analyst at Geeknetic since 2011. I love to dissect everything that passes through my hands, especially the latest hardware that we receive here to do reviews. In my free time I tinker with 3D printers, drones and other gadgets. For anything you need, here I am.
Add Comment