Linux supercomputer comparison record
Article: https://yenra.com/linux-supercomputer/
Prepared: September 15, 2026
Purpose: compare results for a defined workload and measurement boundary.

QUESTION
Application, input and correctness/convergence target:
Fixed problem size or growing problem size:
Deadline or resource budget:

FOR EACH CONFIGURATION
System / list edition / source URL / measurement date:
Node, CPU and accelerator counts:
Memory / interconnect / storage:
OS, compiler, libraries, application and launch settings:
Benchmark or application:
Precision and operation-count convention:
Rpeak / Rmax if relevant, with units:
Queue / setup / compute / output time, with units:
Repeated runs and variation:
Result correctness:
Average power (kW) and components included:
Elapsed measurement time (hours):
Energy (kWh) = average kW x hours:
Excluded facility loads or other limitations:

COMPARISON
Speedup = baseline elapsed time / candidate elapsed time.
Time reduction (%) = (baseline - candidate) / baseline x 100.
Energy change (%) = (candidate energy - baseline energy) / baseline energy x 100.
Decision, reason and missing evidence:

FICTIONAL EXAMPLE
A: 2 h at 100 kW = 200 kWh.
B: 1.5 h at 150 kW = 225 kWh.
Speedup 1.333...; time reduction 25%; energy increase 12.5%.
These values are invented and share the same measurement boundary.

Primary references:
https://www.top500.org/project/top500_description/
https://www.top500.org/project/linpack/
https://hpcg-benchmark.org/