Dell EMC launches a GPU-equipped machine learning server product targeting data center AI workloads.

Dell Technologies World unveiled the Dell EMC DSS 8440, a high-end machine learning server designed for data centers that can be configured with four, eight, or ten Nvidia Tesla V100 GPUs. This two-socket server features two of the new Xeon Scalable processors and is built for machine learning applications and other demanding workloads. Each Tesla is capable of more than 100 teraflops, making a fully equipped 10-GPU machine capable of one petaflop of processing power. Dell claims the DSS 8440 is almost on par with the performance of the DGX-1, which is also Tesla-powered.

The DSS 8440 is not Dell EMC's entry-level offering. The company also provides the 740 and 7425 servers, which support up to three GPUs, and the 4140, which supports up to four GPU cards. The DSS 8440 can accommodate up to ten 2.5-inch devices, translating to up to 32 terabytes of NVMe storage. It features a high-performance switched PCIe fabric for rapid I/O, allowing it to work with accelerators, storage, and network cards from other vendors.

Machine learning involves two distinctly different workloads: training and inference. The initial release of the DSS 8440 targets complex training workloads, such as image recognition, facial recognition, and natural language translation, which represent the computationally intensive phase of machine learning. Training a model is done by iteration, running massive amounts of data through a weighted, multi-layered algorithm thousands of times, comparing results to a specifically targeted outcome, and iteratively adjusting the model weights until achieving a "trained" model. For example, training image recognition to distinguish a cat from a car would require thousands of training images. Once the model is trained, inference—determining whether a given image is a cat, a car, or neither—is much easier and requires significantly less processing power.

Dell EMC is not finished developing the DSS 8440. The company has partnered with Graphcore, a start-up accelerator company, to develop machine learning-specific, graph-based technology for inference workloads. Future versions of the DSS 8440 will come with Graphcore processors, though no release date has been announced.