Polywell Computers servers
Polywell Computers has extensive experience in building server systems of various classes. At present the company is focused on high-performance rack systems in 1U, 2U and 4U form factors using Intel Xeon and AMD EPYC processors.
They can be divided into 3 groups:
- General purpose servers
- Multi Node Servers
- Servers with GPU
General purpose servers are designed for use as file, cloud, virtual, compute, database servers. They are characterized by the use of modern server processors Xeon and EPYC, a significant amount of supported RAM and disk drives in 2.5″ form factor with hot-swappable. The latter can have both SATA interface and the fastest NVMe interface for SSDs. There are also two slots for SATA-DOM modules. There is a significant set of network interfaces – 10 GbE Ethernet, SFP+, IPMI 2.0, OCP 3.0.
Multi-node servers are a combination of 4 servers in one chassis form factor 2U. This is useful when you need to save rack space. Shared power supplies (2 2600W power supplies installed) also save space and power to power and cool the servers. This layout is especially useful for compute or fault-tolerant clusters or for virtualization. The most popular solutions are from Microsoft, Stratus, Citrix and VMware. All of them involve duplication of hardware resources, so such server layouts by Polywell Computers are especially beneficial in such applications. Each of the four nodes supports latest Intel or AMD processors, up to 4TB of memory, 6 disk drives in 2.5″ form factor, 1 OCP 3.0 slot for network modules, one IPMI 2.0 (Intelligent Platform Management Interface) module.
GPU servers are designed for high performance computing (HPC), building GPU computing clusters, artificial intelligence and deep machine learning. Models are available in 2U form factor with 4 Nvidia® HGX A100 40GB/80GB GPUs, NVLink, NVSwitch, and in 4U form factor with 8 Nvidia® HGX A100 40GB/80GB GPUs, NVLink, NVSwitch. They also support up to 8TB (12TB) of memory, up to four (6) 2.5″ drives, two x 10GbE network adapters or one OCP 3.0 slot for network modules depending on the form factor. Through the use of Polywell Computers servers with GPUs, AI acceleration up to 20 times and HPC acceleration up to 2.5 times compared to previous generations of GPUs can be achieved.
All Polywell Computers server systems are assembled from industrial grade components with high reliability. MTBF (mean time between failures) is 100,000 hours (almost 11.5 years).