Supermicro recently introduced an ARM-based server, the ARS-210M-NR i Servethehome service took it on the chest (opens in a new tab)loading it with four Nvidia A16 cards.
The test server was equipped with a 128-core Ampere Altra Max Arm processor running at 3GHz, with 16 DDR4 memory modules (512GB total), two SPF28 ports supporting 25GbE, 16 2.5-inch bays, and a pair of 2kW PSUs , all neatly assembled in a tool-free housing.
The icing on the cake is the quartet of Ampere-based Nvidia A16 graphics cards. Each card ships with 64GB of GDDR6 ECC memory and four GPUs that sit on a full-height, dual-slot, passive-cooled PCB. The A16 is not the most powerful data center card Nvidia has; this would be A100; however, it hits the perfect spot for service providers looking to get the maximum number of concurrent users on a single board.
But can he run…
In terms of pure performance, each of these GPUs has half as many CUDA cores as the Geforce RTX 3050, so yes, it should be able to run the legendary Crysis game without too much trouble – so theoretically you could have 16 iterations of one of the most demanding games ever ever run simultaneously – just create 16 virtual machines and load 16 copies crisis.
Pricing details have not been provided, but considering that The A16 costs around $3,000 and fully loaded server with similar credentials it costs around $10,000, you are expected to spend north of $22,000 for such a system.
Power in the cloud
Who would need such a debauchery of power? First, cloud gaming is now mainstream (even though Google bought out earlier this year when it was put on hold Stadiums) and it is the ARS-210M-NR that makes it a reality.
Beyond that, however, it is about VDI (Video Desktop Infrastructure) for businesses and enterprises. Growth hybrid work acted as a catalyst for the adoption of Virtual Working station (and Virtual PC) as work at home has become a reality for millions of users. And only one of these servers can accommodate multiple virtual machines (VMs); recent test of a cloud computing company Showed Nutanix that two Nvidia A16 graphics cards can handle 128 virtual machines, which means that four cards should handle 256 machines.
Of course, you may need to implement a more powerful processor than the Altra Max and more memory (Nutanix used 1.5TB of RAM and two Intel Xeon Gold 6354 processors, each with 18 physical cores). Increasing SaaS support for virtual GPUs in a wide range of applications (e.g Photoshop Or Chrome) explains why dense servers with very large numbers of GPUs are becoming a common sight.
Other applications, e.g hosting can take advantage of a large number of CPU cores, therefore VPS (virtual private servers) they are now so affordable with a price gap dedicated servers AND pure metal solutions remain high.
Servethehome has launched a 2U server Ubuntu and while the experience wasn’t as seamless as it could have been, it was certainly worth the effort. Arm is slowly turning into a formidable adversary for both Intel Xeon and AMD Epyc, and Amazon is preparing the fourth generation of its Graviton family of processors, there’s never been a better time to try Arm.