Scan AI
Blistering Performance for AI Workflows and More Budget for GPUs
Today’s AI servers consume and analyse data at much higher rates than many traditional storage solutions can deliver, resulting in low GPU utilisation and dramatically extending training times decreasing productivity. PNY, NVIDIA’s global partner, has been developing solutions from the ground up for AI workloads and optimised for the NVIDIA DGX range of AI appliances. Delivering ultra-low latency and tremendous bandwidth at a price which allows more investment to be made on GPU resource and less on expensive, slower storage - ensuring your project’s funds are better spent and your team are more productive, by taking full advantage of the DGX compatibility.
Start Small - Scale Easily
With many new A.I. projects and inference solutions requiring only limited amounts of storage, the PNY 3S storage range starts from 30TB, while still delivering full performance.
With a 1U delivering up to 150TB and a 2U capable of housing a massive 360TB, starting small and scaling as needed is simple. And should a project require larger capacities, additional expansion units are available.
Simple Connectivity
Connections to either a single deep learning server or multiple servers couldn’t be simpler. The interfaces on PNY appliances allows for direct connection to a single server or utilising NVIDIA Mellanox low latency switching technology to connect to multiple servers and create a cluster.
Cluster Filesystem for multiple
DGX compatibility
As solutions evolve and multiple DGX’s are used, the storage requires a sharable filesystem, which traditional has been both complex and expensive to implement at high performance. PNY’s second generation appliances provide a simple yet blisteringly fast sharable solution. There is no need for multiple storage nodes or controllers, everything needed is contained and automated within a single appliance. .
Flexible Fault Tolerance
P-series PNY appliances feature variable RAID protection which provides various levels of resilience. With the recommended configuration being RAID 5, with RAID 1, 10, 5 and 6 all being supported. In a RAID 5 configuration the data and parity are distributed in blocks and not just on one drive. For example, with 12x 15TB drives a single S3-2450P has a raw capacity of 180TB. However, with RAID5 protection enabled and the equivalent of 1 drive worth of capacity is used for parity, the useable capacity drops to 165TB.
| System Specifications | ||
|---|---|---|
| Feature | PNY 3S-1050 | PNY 3S-2450 |
 |
 |
|
| Storage Capacity | Up to 150TB | Up to 360TB |
| Bandwidth | 23GB/s | 23GB/s |
| Connectivity | 2 x QSP56 HDR InfiniBand /200Gb/s Ethernet | 2 x QSP28 EDR InfiniBand /100Gb/s Ethernet |
| Software Platform | PNY | PNY |
| Export compatibility | NVMe-oF (InfiniBand or Ethernet) NFS (InfiniBand or Ethernet – RDMA or TCP) |
NVMe-oF (InfiniBand or Ethernet) NFS (InfiniBand or Ethernet – RDMA or TCP) |
| Software Licence | 3 years / 5 years on request | 3 years / 5 years on request |
| Form Factor | 1U - 660mm deep | 2U - 710mm deep |
| Warranty | 3 years | 3 years |
| Support | 3 years 24/7 Premium Support | 3 years 24/7 Premium Support |