GL-G1U-4G
Up to 4 GPUs
1U scale-out GPU server for up to 4 PCIe GPUs. Host platform: AMD EPYC 7002. Validated with GenesysBench™ and deployed under GLIDS; AI workloads run on the GLabs AI Stack.
- GPUs
- up to 4
- Memory
- 8 DIMM slots
- Height
- 1U
- Sockets
- 1

Specifications
Pick the host platform. The chassis, power and cooling follow it.
Previous generation. Still supported by GLabs; ask a solutions architect about availability.
| Form factor | 1U Rackmount |
|---|---|
| Processor | 1 Socket SP3 (LGA4094), AMD EPYC™ 7003 (with AMD |
| Memory | 8 DDR4 DIMM slots (1DPC) |
| Expansion | 3 PCIe4.0 x16 for dual slot cards1 PCIe4.0 x16 for a dual slot card 2 PCIe4.0 x16 for low-profile cards |
| Storage | 8 E1.S or M.2 (PCIe4.0x4) |
| Networking | 2 RJ45 (1GbE)1 dedicated IPMI |
| Power | 1+1, 80-PLUS Platinum, 2000W CRPS |
| Cooling | 8 dual-rotor 40x56 mm fans |
| Dimensions (D × W × H) | 880 x 438 x 43.5 mm (34.6" x 17.2" x 1.7") |
Previous generation. Still supported by GLabs; ask a solutions architect about availability.
| Form factor | 1U Rackmount |
|---|---|
| Processor | 3D V-Cache™ Technology )/7002 series processors |
| Memory | 8 DDR4 DIMM slots (1DPC) |
| Expansion | 3 PCIe4.0 x16 for dual slot cards1 PCIe4.0 x16 for a dual slot card 2 PCIe4.0 x16 for low-profile cards |
| Storage | 2 Hot-swap 2.5" NVMe (PCIe4.0 x4) drive bays |
| Networking | 2 RJ45 (1GbE)1 dedicated IPMI |
| Power | 1+1, 80-PLUS Platinum, 2000W CRPS |
| Cooling | 8 dual-rotor 40x56 mm fans |
| Dimensions (D × W × H) | 880 x 438 x 43.5 mm (34.6" x 17.2" x 1.7") |
Exact CPU, memory, drive and network parts are fixed on your quote. GPU and platform names are trademarks of their respective owners.
Views


Configure GL-G1U-4G
Choose the host, memory, storage and networking. Your request goes straight to a Solutions Architect, who replies with a validated build and quote.
- 01System and host
- 02Processor
- 03Memory
- 04GPU
- 05Storage
- 06Networking
- 07Software stack
- 08Send request
Tell us what you're building.
A solutions architect responds directly — no ticket queue.


