Ethernet-based memory appliance for faster training and low-latency, enterprise-scale production inference.
Penguin Solutions MemoryAI KV Cache Server is an Altus-based 4U rackmount server designed specifically for use as a KV cache server during AI inference. It features dual AMD EPYC CPUs and enables up to 11 TB of CXL-based memory (SMART CXA-8F2W).
The server provides 3TB main DDR5 memory and up to 8 1TB CXL Add-In-Cards (AICs) and uses dual 400Gb/s Ethernet NICs for network connectivity.
Up to 11 TB of DDR5 to create pooled memory for KV cache storage.
CXL Add-In-Cards to provide significantly expanded memory capacity critical for KV cache tiering.
Up to 24 DDR5-6400 MT/s ECC memory modules to deliver speed and efficiency for data-intensive workloads.
Dual AMD EPYC™ 9005 processors.
Full specifications for MemoryAI KV Cache Server.
| Specification | Details |
|---|---|
| Form Factor | 4U Rackmount |
| Processors |
|
| Motherboard | Chipset: SoC |
| Board Management |
|
| Memory |
|
| Storage System |
|
| Networking |
|
| PCI Expansion Slots | 8x PCIe Gen5 x16 FHFL, 2x PCIe Gen5 x16 LP |
| GPU |
|
| External I/O Interfaces |
|
| Supported Operating System | RHEL, Rocky |
| Power System | Power Supply Size: 4x 3000W 80Plus Titanium |
| Regulatory Compliance | CE, FCC |
| Mounting Hardware | Standard Rails included |
| Operating Environment |
|
| System Dimensions |
|
| Warranty | 3 Years Standard |
Download product documentation and resources
Our team of experts is ready to help you find the perfect solution for your business needs. Get personalized advice and competitive quotes.
We're here to help with any questions