Dell is making hardware and software upgrades to its PowerScale storage solutions to help companies train and deploy custom AI models.

Companies are excited about the big promises around generative AI, but hosting AI workloads is a big challenge. These are not our words, but those of Dell CTO John Roese. Just a few days after this statement, the company also presented the solution. Dell announces enhancements to its PowerScale storage platform to accelerate AI workloads, particularly generative AI.

Dell adds new all-flash storage systems based on the latest generation of Dell PowerEdge servers that are up to two times faster in streaming reads and writes. PowerScale also offers a new intelligent scale-out feature to improve individual node performance for better GPU utilization, resulting in faster memory processing for AI training, checkpointing and inference.

Hardware is one component, but software is at least as important. Improvements to integrated OneFS software give organizations access to faster AI model preparation, training, tuning, and interference.

HPC Champions

AMD announced its new HPC champions Instinct MI300X and MI300A this week to compete with Nvidia, and Dell is making no bones about offering these options in the PowerEdge. The PowerEdge

The solution features eight MI300X GPUs, 192 GB of 5.3 TB/s High Bandwidth Memory (HBM3) per GPU for a total coherent HBM3 capacity of 1.5 TB per server, and more than 21 petaflops of FP16 performance. All of this, in turn, to accelerate access to generative AI. Dell believes that generative AI should not necessarily run in the public cloud, but can also live on-premises or at the edge. The prerequisite for this is a sufficiently solid infrastructure.