MLCommons has ranked Huawei’s new OceanStor A800 AI Storage first worldwide in its annual performance test after evaluating 13 leading vendors based on the maximum number of GPUs supported by an AI storage system.

The OceanStor A800 AI Storage solution:

·         met data throughput requirements of 255 GPUs using just a single storage system.

·         achieved GPU utilization above 90 per cent

·         achieved a bandwidth of 679 GB/s with a single controller enclosure — ten times greater than conventional storage systems provides 100 TB/s–level bandwidth through scale-out expansion, reducing the read/write time of checkpoint data from ten minutes to just seconds and enabling resumable training is under 15 minutes. This minimizes GPU wait times, improves end-to-end computing power utilization by over 30 per cent, and enhances the training efficiency of large AI models.