Business AI and HPC knowledge source code.
Browse by topic
Read by the question you are asking.
Notify me of new work
New articles and whitepapers, when they are published. No marketing.
Up to eight GPUs, outside the rack
Rack-class compute without a server room
Liquid-cooled workstations with the same eight-GPU compute as our rack server, built around the workloads you plan to run and upgradeable as they grow.
On the workstations page
- Why the cooling scales Radiator surface instead of fan speed
- Tested side by side Eight NVIDIA H200 NVL against our rack server
- Sizing advice Size for the density you will grow into
How many GPUs, and which chassis? An engineer sizes it with you.
Featured platform · 4U · 8 GPU New
LM TEK 4U8GPU10KW
8-GPU liquid-cooled AI server in a 4U envelope. Built around off-the-shelf internals and an EK-engineered cooling loop – designed for your server room, not a hyperscale facility.
- Form factor
- 4U rackmount
- GPU slots
- 8 full-height PCIe Gen 5
- Cooling
- Direct-liquid
- Engineered
- In Komenda, EU
Platform details
- Specifications Full platform specifications and reference configurations.
- Inside How the cooling loop, power and chassis are built.
- Validation Thermal and power results, rated and measured.
- Resources Drawings, datasheets and documentation.
- Compatible components Liquid-cooling parts that fit this platform.
Plan and buy
- Servers overview The platform, who it suits and how to start.
- Deployment Fit it into the server room you already have.
- Use cases Inference, fine-tuning, simulation and rendering.
- Ownership Start modestly and upgrade on your own schedule.
- How to buy Buy through a system integrator or distribution partner.
- Modification projects When your application needs more than the standard platform.
Roadmap
Additional server platforms are added as released. Looking for something specific? Talk to engineering.