ScaleMatrix and NVIDIA introduce distributed AI/HPC appliances that operate independently without traditional data center infrastructure.

At SC19, the supercomputing conference in Denver, data center provider ScaleMatrix introduced AI Anywhere, appliances that can deliver up to 13 petaFLOPS of performance with built-in cooling, requiring only "a roof, floor space, and a place to plug the appliance in" to turn on "an enterprise-class data center capable of supporting a significant artificial intelligence or high performance computing workload," according to Chris Orlando, ScaleMatrix's co-founder and CEO. The products address a critical barrier: most high performance computing systems required for machine learning and AI workloads cannot be located in a typical data center without major facility modifications, as GPU-intensive HPC systems bring rack density up to about 30kW per rack, at least five times higher than the average data center load of 3kW to 5kW per rack.

AI Anywhere was developed through a three-way collaboration between ScaleMatrix, chipmaker Nvidia, and Microway, a provider of computer clusters and servers for HPC and AI. The solution comes in two single-rack versions, each employing one of Nvidia's DGX supercomputer models. One model contains 13 DGX-1 units, delivering a payload of 13 petaFLOPS, while the other contains four DGX-2 systems, delivering 8 petaFLOPS. Both units adhere to DGX-POD reference architecture designs and include the full NVIDIA DGX software stack, deep learning and AI framework containers, NetApp ONTAP storage, and Mellanox switching.

The innovation enabling these plug-and-play supercomputers is ScaleMatrix's proprietary closed-loop chilled water-assisted forced-air cooling system. With the DGX-1 version consuming 42kW and the DGX-2 version running at 43kW, the appliances employ a hybrid air-and-water approach using in-rack sensors to direct cool air where needed. As Orlando explained, "Where the water comes in and out and does the thermal exchange, that area is sealed off from the rest of the cabinet, so we're bringing all the efficiency of water cooling to the cabinet without introducing any of the risk." The design can handle much higher densities—DDC, a ScaleMatrix subsidiary, recently introduced a rack capable of handling up to 85kW using the same cooling system.

According to Tony Paikeday, Nvidia's director of product marketing of AI and deep learning, AI Anywhere addresses a persistent customer need. "Customers are deploying larger and larger infrastructures to either tangle with more complex AI problems like natural language processing, or they're doing a consolidation play of trying to take stranded AI platform investments, kind of like 'Shadow AI,' that are sprawling across their enterprise and bring them under one roof," he said. The key barrier has been that many enterprises hesitate to invest further CapEx in data center infrastructure. "So this is now a perfect way to remove that last-mile barrier of how to get this kind of computing power into their hands," Paikeday noted. The appliances will be marketed under the AI Anywhere brand, with component partnerships from Nvidia, NetApp, Mellanox, Microway, and DDC each retaining their own branding on the individual infrastructure elements.