Clusters Explained

What a cluster is. What it isn't. And the honest difference between real datacenter clustering and a pile of desktop cards.

Plain definition
A cluster is two or more computers working together as one.

What a cluster does

A model too large to fit in one GPU's memory can be split across multiple GPUs — or multiple machines — so that each piece holds a slice. The GPUs communicate constantly as they process each token. From the outside, it looks like one system. From the inside, it's a coordinated handoff happening thousands of times per second.

This is called model parallelism. It's how you run a 671-billion-parameter model when no single piece of hardware holds 805 GB of GPU memory.

Why the fabric matters

GPU communication speed determines how fast the handoffs happen. Slow communication creates a bottleneck — the model waits for data instead of computing. This is why NVLink and InfiniBand exist.

  • PCIe bus: ~64 GB/s. Standard computer interconnect. Fine for single-GPU use. Bottlenecks hard in tight multi-GPU coordination.
  • NVLink (within a DGX): 900 GB/s. 14× faster than PCIe. Enables all 8 GPUs in a DGX to act as one unified pool.
  • InfiniBand (between DGX nodes): 400 Gb/s. Connects DGX boxes to each other at near-NVLink speeds, enabling true multi-node clusters.

Real cluster vs. pile of desktop cards

❌ Pile of Desktop Cards
e.g., 4× RTX 4090
ConnectionPCIe lanes / CPU bridge
GPU-to-GPU bandwidth~64 GB/s (shared)
Memory poolingNo — each GPU is isolated
Run a 156 GB model?No — 96 GB total, fragmented
FailoverNone — one card dies, job dies
SchedulingManual / ad hoc
Power per GPU450 W × 4 = 1,800 W
Total VRAM96 GB (4 × 24 GB, not pooled)
Price~$10,000

Desktop cards share a PCIe bus through the CPU. They can cooperate on some tasks (like training a small model), but they can't pool their memory. A model that needs 100 GB cannot run across 4 RTX 4090s — each card still only sees its own 24 GB.

✓ Real Datacenter Cluster
e.g., 2× DGX H200
ConnectionNVSwitch + InfiniBand
GPU-to-GPU bandwidth900 GB/s (NVLink) + 400 Gb/s (IB)
Memory poolingYes — all GPUs are one logical pool
Run a 156 GB model?Yes — 2,256 GB available
FailoverHardware HA, Kubernetes scheduling
SchedulingSlurm / K8s / NGC containers
Power per node10,200 W × 2 = 20,400 W
Total VRAM2,256 GB (16 × 141 GB, fully pooled)
Price~$930,000

DGX nodes use NVSwitch to give all 8 internal GPUs full memory visibility. InfiniBand links the nodes. The result: a 671B-parameter model that needs 805 GB of VRAM fits comfortably, with room to spare, and the system can failover automatically if a node goes down.

The upgrade path

Every build in our catalog fits on this line. Desktop cards are fine for development. DGX systems are where real production begins. Clusters are where scale lives.

1× RTX 4090
$2,399
24 GB
2× H100 PCIe
$60,000
160 GB
1× DGX H200
$450,000
1,128 GB
10-node SuperPOD
$3.5M
11,280 GB
Full Rack
$6.5M
22,560 GB