Follow the latest coverage, related explainers and connected technology stories.
An analysis by The New Stack reveals that managing shared graphics processing units on Kubernetes does not provide the same level of visibility available for CPUs and memory, leaving platform teams to build measurement, cost-allocation, and control systems themselves. The problem is compounded by the fact that the GPU is the most expensive resource on the bill, while the accuracy of assigning compute consumption among workloads remains limited.
Dell Technologies has begun selling the Dell Pro Max with GB300 desktop workstation, which is based on the NVIDIA Grace Blackwell Ultra GB300 chip and delivers up to 2 petaflops of FP4 compute performance. The workstation targets model development and fine-tuning, data science, molecular simulation, and edge AI applications.