Skip to content
GIRIVON
Infrastructure+
Systems+
Components+
Computing Devices+
Solutions+
Compute Tools+
Knowledge+

HPC Cluster Planner

Plan an HPC cluster from its node design: theoretical peak (Rpeak) from CPUs and GPUs, an estimated sustained figure from an efficiency you set, plus power, heat and the racks needed by space and by power.

Nodes

Your measured or assumed ratio of sustained to peak.

Facility

42U minus switches, PDUs and cable management.

Results

CPU peak (FP64)
144.2 TFLOPS
GPU peak (FP64, vendor)
0 TFLOPS
Cluster Rpeak
144.2 TFLOPS
Estimated sustained
100.9 TFLOPS
IT power
35.2 kW
Heat load
1,20,107 BTU/h
Racks needed1 by space, 2 by power
2
Peak per watt
4.1 GFLOPS/W

Power, not space, limits rack density. Consider higher-power racks or liquid cooling.

What it calculates

  • CPU and GPU peak FP64
  • Cluster Rpeak
  • Estimated sustained performance
  • Power and heat
  • Racks by space and by power
  • GFLOPS per watt

Worked example

Using the default inputs below; change any input above to recalculate.

Inputs
Compute nodes32
CPUs per node2
Cores per CPU64
Sustained vector clock2.2 GHz
FP64 FLOPs per cycle per core16 (AVX2 2 FMA or AVX-512 1 FMA)
GPUs per node0
GPU FP64 peakNVIDIA H100 SXM5 80GB — 34 TFLOPS
Sustained efficiency (HPL-like)70 %
Power per node (max)1100 W
Rack units per node1 U
Usable U per rack38 U
Power limit per rack20 kW
Results
CPU peak (FP64)144.2 TFLOPStheoretical
GPU peak (FP64, vendor)0 TFLOPSVendor-reported
Cluster Rpeak144.2 TFLOPStheoretical
Estimated sustained100.9 TFLOPSestimated
IT power35.2 kWcalculated
Heat load1,20,107 BTU/hcalculated
Racks needed2calculated
Peak per watt4.1 GFLOPS/Wtheoretical

Methodology

CPU peak = nodes × CPUs × cores × sustained clock × FP64 FLOPs per cycle. GPU peak uses vendor-reported FP64 figures. Sustained performance applies your efficiency factor. Racks are the larger of the counts required by rack units and by the per-rack power limit.

Formulas

Rpeak
R = Σ_CPU (cores × f × FLOPs/cycle) + Σ_GPU peak_FP64
Racks
max(⌈nodes × U ÷ usable_U⌉, ⌈nodes ÷ ⌊rack_kW ÷ node_kW⌋⌉)

Assumptions

  • FP64 vector peak is used for CPUs; GPU FP64 figures are non-tensor vendor values.

Limitations

  • Efficiency is an input; real codes range widely.
  • Switches, storage and management nodes are not included.