According to reports, Micron is exploring solutions to develop highly durable NAND flash memory modules to study how to bring NAND flash memory closer to GPUs to handle workloads that do not require the high latency and high bandwidth of traditional HBM or conventional DRAM modules. Currently known as “near-GPU NAND,” this architecture is designed to balance density and durability. This solution allows GPUs to access NAND flash pools with lower density than traditional TLC or QLC NAND, but the focus is on improving I/O speed, bandwidth, and read speed. The technology will allow GPUs to directly access potential storage pools of hundreds of gigabytes, and LLM will no longer be limited by memory.

According to reports, Micron is exploring solutions to develop highly durable NAND flash memory modules to study how to bring NAND flash memory closer to GPUs to handle workloads that do not require the high latency and high bandwidth of traditional HBM or conventional DRAM modules. Currently known as “near-GPU NAND,” this architecture is designed to balance density and durability. This solution allows GPUs to access NAND flash pools with lower density than traditional TLC or QLC NAND, but the focus is on improving I/O speed, bandwidth, and read speed. The technology will allow GPUs to directly access potential storage pools of hundreds of gigabytes, and LLM will no longer be limited by memory.