8 GPU HGX B300 AI server, air cooled
Allocation required- GPUs
- 8 x NVIDIA B300, 288 GB HBM3e each
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 3 TB DDR5 RDIMM across 24 or 32 slots
- Chassis
- 10U, about 14 kW, air cooled
Blackwell servers, tested and delivered. San Diego, since 1978.
NVIDIA B300, B200, RTX PRO 6000 and RTX 5090 servers, GPU workstations and NVIDIA components. From the leading OEMs, plus our own white label configurations where they fit. Specified for your workload, tested before delivery and quoted within one business day.
What we sell
Four server lines, GPU workstations and the NVIDIA parts around them.
- B300 servers8 GPU HGX B300 nodes, air or liquid cooled, single or as 8, 16 and 64 node pods.See HGX B300 servers
- B200 servers8 GPU HGX B200 nodes, the Blackwell platform with the widest availability.See HGX B200 servers
- RTX PRO 6000 serversPCIe servers with 2 to 10 RTX PRO 6000 Server Edition GPUs, 96 GB each.See RTX PRO 6000 servers
- RTX 5090 servers4 and 8 GPU servers on RTX 5090, the lowest cost per GPU for inference.See RTX 5090 servers
- GPU workstationsDeskside and rackmount, with 1 to 4 RTX PRO 6000, RTX PRO 4500 or RTX 5090 GPUs.See workstations
- NVIDIA GPUs, networking and partsGPU cards, ConnectX and BlueField adapters, switches, CPUs, memory and NVMe.See parts and networking
Who buys from us
Computer Modules and our DVEO video division have supplied aerospace, defense, government, technology and broadcast customers from San Diego since 1978.
Aerospace, defense and government
Technology and telecom
Broadcast and media, through our DVEO division
Logos belong to their owners and record a purchasing relationship, not an endorsement.
Shop by GPU
Start from the accelerator and we show every system class that carries it.
Compare all GPUs- HGX B300Blackwell Ultra
- HGX B200Blackwell
- HGX H200Hopper
- H200 NVLHopper
- RTX PRO 6000 Blackwell Server EditionBlackwell
- RTX PRO 6000 BlackwellBlackwell
- RTX PRO 4500 BlackwellBlackwell
- L40SAda Lovelace
- GeForce RTX 5090Blackwell
- GB300 NVL72Blackwell Ultra
- Vera Rubin NVL72RubinComing soon
- HGX Rubin NVL8RubinComing soon
Featured systems
Representative classes we quote most often. Every one is configured to order.
All GPU servers
8 GPU HGX B200 AI server
- GPUs
- 8 x NVIDIA B200, 180 GB HBM3e each
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 3 TB DDR5 RDIMM
- Chassis
- 8U, about 14 kW, air cooled
8 GPU RTX PRO 6000 PCIe AI server
- GPUs
- 8 x RTX PRO 6000 Blackwell Server Edition
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 2 TB DDR5 RDIMM
- Chassis
- 4U, about 6.5 kW, air cooled
4 GPU PCIe AI server, 2U
- GPUs
- 4 x RTX PRO 6000 Blackwell Server Edition
- CPU
- 2 x AMD EPYC 9005 or 2 x Intel Xeon 6
- Memory
- Up to 1.5 TB DDR5 RDIMM
- Chassis
- 2U, about 3.5 kW, air cooled
- Choosing a GPU server for LLM inferenceSize the server from the model, the quantization and the concurrency
- HGX vs PCIe GPU serversWhat separates an HGX node from a PCIe server, and a rule for choosing
- RTX PRO 6000 Blackwell editions explainedWorkstation, Max-Q and Server Edition, and which one to buy
- LLM trainingHGX nodes and pods with NVLink and 800 Gb/s fabric, at three scales
- LLM inferenceServers sized to the model, the quantization and the token rate
- Fine tuningSingle nodes with the GPU memory for LoRA, full fine tuning and pretraining
- Rendering and visualizationRTX PRO systems for a desk, a render farm and a studio
- Simulation and HPCGPU and interconnect choices for a lab, a department and a center
- Research and educationShared GPU compute from one workstation to a cluster
Ready to specify a system?
Tell us the workload, the site and the timeline and we come back with a specified system.
- (858) 613-1818
- sales@computermodules.com
- Monday to Friday, 8am to 5pm Pacific