The NVIDIA® Tesla® V100 is the most advanced data center GPU ever built to accelerate graphics, HPC, and AI. To enable significantly faster GPU-to-GPU communication, NVIDIA offers the Tesla V100 in an SXM2 form factor that leverages NVIDIA® NVLink™, a high-bandwidth, energy-efficient connection. The SXM2 design enables GPUs to provide speeds that are faster than the standard PCIe bus, delivering maximum performance potential.
When you want the speed and power of an NVIDIA SXM2 server, Thinkmate provides a wide range of server computers and storage solutions that can be customized 100% to meet the needs of your computing environment and the demands of your high-performance workload.
PCI Express (PCIe) data buses have been the standard in high-speed serial computer expansion, providing lower latency and higher data transfer rates that improve a system's overall throughput. Devices connected via PCIe have dedicated point-to-point connections, meaning they don't have to compete for bandwidth with other devices.
While PCIe is adequate for many systems, they don't provide sufficient bandwidth for high-end GPUs and multi-GPU systems moving vast amounts of data in massively parallel functions. When connected via a PCIe bus, the systems can quickly become overloaded, resulting in bottlenecks.
To resolve this issue for its GPU accelerators, NVIDIA introduced an SXM2 connector that leverages NVIDIA NVLink for ultrafast connections between CPUs and GPUs, or for GPU-to-GPU communication. An SXM2 server can reach peak data rates of 300 gigabytes per second a 10-fold increase over the speeds available with PCIe buses. Whereas PCIe buses share a central hub for communication, an SXM2 / NVlink server uses a mesh that provides multiple paths, allowing the highest bandwidth data interchange between GPUs. This increase in raw compute power improves data flow and total system throughput to enhance the performance of workloads.
Thinkmate provides a broad range of customized server, storage and workstation solutions for customers in business, government, and education. Our solutions combine outstanding reliability with exceptional customer service and ultra-competitive pricing, making us the #1 white box solutions provider for more than 25 years.
Our line of GPU accelerated, enterprise-class supercomputing servers offers performance and quality that are optimized for the most computationally intensive applications. With up to four GPUs in a 1U form factor, our GPX servers are ideal for high-performance computing environments and mission-critical applications in a broad range of industries.
Offering the performance of up to 100 CPUs in a single GPU, our GPX servers with NVIDIA Tesla V100 and SXM2 enable data scientists, researchers, and engineers to tackle challenges that were once thought impossible. Our SXM2 servers are available in form factors from 1U to 4U with 1-10 GPUs, 4-24 drive bays, and 1-4 processors.
Thinkmate offers a world-class configurator with more customizable options than any other system builder on the web. Offering transparent pricing and 100% customizability, our machines can be configured to meet the needs of any computing environment, and if you can't find what you need within our extensive selection of servers, our solution architects will custom build a machine to your exact specifications.
To configure an SXM2 server, simply select one of our base model GPX servers based on your ideal platform, coprocessors, form factor, maximum RAM, type of drives, and other features. Next, customize your SXM2 server by adding components or swapping out features to configure a server that meets your exact specifications. There's no mystery when it comes to pricing a server from Thinkmate – the price of every component is clearly marked, and you'll see the total configured price of the machine at the top of the page as you make your selections. You can build your machine choosing from a wide range of customizable options including processors, memory, M.2 drives, hard drives and solid-state drives, GPU accelerators, controller cards, network cards, operating systems, software, and more.
An SXM2 server is an NVIDIA GPU-accelerated server built with the SXM2 form factor, leveraging NVIDIA NVLink to achieve the highest bandwidth data interchange between GPUs.
An SXM2 server provides higher compute power, enhances dataflow, and maximizes total system throughput to increase performance for a variety of workloads.