Use four accelerators with higher CPU and memory capacity for research groups, multi-user inference and light model fine-tuning.
Nexxys AI & HPC infrastructure
Eight GPUs. Four rack units. One closed loop.
The LM TEK 4U8GPU10KW is an eight-GPU barebone platform built around integrated direct-liquid cooling and platform-level power. Nexxys completes the system with the motherboard, processors, GPUs, memory, storage and networking selected for your workload.
Configure my system
Explore configurations
4U
Standard 19-inch rack
8 x GPU
Full-height PCIe Gen 5
Direct liquid
in-chassis radiator
Open platform
EATX and EEB boards
Engineered as one thermal system
AI density without facility water infrastructure.
The pump, manifold, CPU and GPU water blocks, radiator and reservoir are contained inside the 4U chassis. The pre-filled closed loop requires no facility plumbing or external coolant distribution unit, while the server still demands rack power and room-level heat-removal capacity for the selected components.
Configuration range
One chassis from departmental AI
to production scale.
01 Departmental AI
Start with two professional GPUs, a single enterprise CPU, ECC memory and mirrored or high-performance NVMe storage for inference, development and single-team workloads.
02 Research & fine-tuning
03 Production AI
Populate all eight GPU positions for enterprise inference, production fine-tuning and multi-tenant acceleration, with CPU, RAM, NVMe and networking sized around the workload.
04 Custom engineering
Validate alternative boards, accelerators, networking or storage layouts with Nexxys and LM TEK engineering when the required configuration sits outside the published reference envelope.
The Nexxys approach
A complete system, not an unfinished barebone.
01
Scope
We map the AI workload, software stack, GPU memory, data pipeline and deployment environment.
02
Configure
We select compatible CPUs, GPUs, memory, storage, networking and rack-power requirements.
03
Validate
The complete server is assembled, updated, pressure-tested and system-tested before delivery.
Build the right AI server
Let’s configure the 4U eight-GPU
platform around your workload.
Tell us which models or applications you intend to run, your GPU preference, storage needs and
datacentre constraints. We will translate them into a complete, validated configuration.
datacentre constraints. We will translate them into a complete, validated configuration.