Setonix SU Calculator
CPU Node Specifications
CPU Cores
RAM
7763 Milan
Base Clock
Calculation Results
Total Service Units (SU)
Core Hours
Node Hours
Resource Breakdown
About This Calculator
This advanced calculator helps researchers and scientists accurately estimate Service Unit (SU) consumption for jobs on Setonix, Australia's most powerful supercomputer at the Pawsey Supercomputing Centre. Our tool provides precise calculations based on real hardware specifications and current allocation policies, presented in a clean interface inspired by the Kadence theme.
๐ฅ๏ธ About Setonix HPC
Setonix is a state-of-the-art HPE Cray EX supercomputer, ranking as the most powerful research computer in the Southern Hemisphere and the world's fourth greenest supercomputer. Built on the same architecture as world-leading exascale systems like Frontier and LUMI.
- 1,600 CPU nodes with dual AMD EPYC 7763 processors
- 192 GPU nodes with AMD Instinct MI250X accelerators
- Connected by HPE Slingshot 200Gb/sec interconnect
- Lustre file systems with 14 PB storage capacity
๐ How to Use This Calculator
Follow these simple steps to calculate your Service Unit requirements:
- Select Partition Type: Choose between CPU or GPU computing based on your workload requirements
- Configure Resources: Set the number of nodes, cores, memory, and GPUs needed
- Set Runtime: Enter your expected job duration in hours, minutes, and seconds
- Adjust Efficiency: Estimate your code's resource utilization efficiency (typically 70-95%)
- Review Results: Get instant calculations for Service Units, core hours, and resource breakdown
๐ฏ Service Unit Calculation Formula
Service Units are calculated using the following methodology:
SU = Nodes ร Cores ร Walltime ร Efficiency
GPU Jobs:
SU = Nodes ร (Cores + GPUs ร 8) ร Walltime ร Efficiency
This formula accounts for the computational weight of different resources, with GPUs having an 8x multiplier due to their significantly higher computational capacity.
๐ก Optimization Tips
Maximize your allocation efficiency with these proven strategies:
- Profile your code to determine optimal core/node ratios
- Use GPU acceleration for parallel workloads when possible
- Monitor memory usage to avoid over-allocation
- Test with shorter jobs before submitting long runs
- Consider checkpointing for jobs longer than 24 hours
- Use array jobs for parameter sweeps and ensemble runs
- Optimize I/O patterns to reduce walltime
๐ง Technical Specifications
Detailed hardware specifications for accurate planning:
CPU Nodes
- Dual AMD EPYC 7763 (128 cores)
- 256 GB RAM (standard)
- 1 TB RAM (high-memory)
- 2.45 GHz base frequency
GPU Nodes
- AMD EPYC 7A53 (64 cores)
- 8ร AMD MI250X GPUs
- 256/512 GB RAM
- 220 TFlops peak (FP64)
๐ Why This Calculator is Superior
Our calculator outperforms competitors with these advanced features:
- Real-time calculations with instant feedback
- Accurate formulas based on official Pawsey documentation
- Efficiency factor adjustment for realistic estimates
- Multiple partition types including high-memory options
- Responsive design for all devices and screen sizes
- Detailed resource breakdown and optimization tips
- No external dependencies - works offline
- Comprehensive help and technical specifications
Read Also: