ScaleFlux bets smarter SSD controllers can break the AI memory bottleneck
AI factories run on graphics processing units that cost a fortune and work at blistering speed, yet those chips increasingly sit idle waiting for data. As the bottleneck moves from compute to memory and storage, NAND flash is being asked to do far more than hold files. That shift is putting data movement constraints at […] The post ScaleFlux bets smarter SSD controllers can break the AI memory…
ScaleFlux Inc. is addressing the AI memory bottleneck by developing smarter SSD controllers that can utilize NAND flash as active memory for AI workloads. According to ScaleFlux's vice president of products and marketing, Pichai Balaji, GPUs operate at speeds of 20 to 22 terabytes per second, but every tier of memory and storage below it adds significant delay.
This creates a bottleneck as data moves from the GPU into memory and storage. ScaleFlux aims to break this bottleneck by using NAND flash as a memory device, allowing it to perform key-value cache offload for inference workloads. The company's AI-optimized SSD platform targets flexible data placement streams and high drive writes per day, making it adaptable to the tight supply market for flash storage.
Brief written by urgent.news from SiliconANGLE's own syndicated text. Machine-written — may contain errors; check the original before relying on it.
