GoKawiilTom's Hardware Premium roundup: DGX Spark vs Mac Studio, AMD's MEXT halves DRAM needs
Tom's Hardware Premium's latest weekly digest benchmarks Nvidia's DGX Spark against Apple's Mac Studio for local AI workloads, and details AMD's new MEXT memory engine, which uses a transformer model and flash storage to deliver 99% of server performance on half the DRAM. AMD is currently running MEXT on 145 servers and plans to scale that to 2,000 next year. The issue also covers Jensen Huang's recent step-down from Nvidia's CEO role and an interview with Cognichip's Chief Product Officer about its AI chip-design tools.