Disaggregate your GPUs. Scale your compute.
Traditional GPU servers lock valuable resources into fixed configurations—often leaving compute underutilized across your environment.
The Liqid EX-5410P, available through Mazda Computing, enables a composable infrastructure model that pools GPUs, storage, and accelerators into a shared resource that can be dynamically allocated to any connected system.
The Liqid EX-5410P, available through Mazda Computing, enables a composable infrastructure model that pools GPUs, storage, and accelerators into a shared resource that can be dynamically allocated to any connected system.
What is the EX-5410P?
|
The EX-5410P is a 4U PCIe Gen5 expansion chassis that allows up to 10 high-performance GPUs or accelerators to be shared across multiple servers using a high-speed PCIe fabric.
Instead of assigning GPUs to a single server, resources can be:
|
|
Key Benefits
Maximize GPU utilization
Eliminate stranded compute by dynamically allocating GPUs where they are needed most.
Increase infrastructure flexibility
Support multiple workloads across AI, HPC, and research environments without fixed hardware limitations.
Reduce overprovisioning
Avoid purchasing excess GPU capacity for peak demand—allocate resources as workloads shift.
Enable composable infrastructure
Create a shared pool of GPUs, NVMe, and accelerators that can be orchestrated across systems.
Eliminate stranded compute by dynamically allocating GPUs where they are needed most.
Increase infrastructure flexibility
Support multiple workloads across AI, HPC, and research environments without fixed hardware limitations.
Reduce overprovisioning
Avoid purchasing excess GPU capacity for peak demand—allocate resources as workloads shift.
Enable composable infrastructure
Create a shared pool of GPUs, NVMe, and accelerators that can be orchestrated across systems.
Technical Overview |
How it works
|
|
Using Liqid Matrix software, the EX-5410P enables:
|