AWS Graviton3
Efficient Arm compute for compatible parallel workloads.
- Scale
- 126–504 cores
- Memory
- 196–783 GB
- Nodes
- 2–8
Eferka Technologies HPC capability
High-performance computing
Run the study your workstation cannot.
Scale demanding CFD models across Arm and x86 compute architectures—with the cores and memory needed for larger meshes, longer transients and broader design-space exploration.
CFD / PARALLEL RUNResources available
3D modelMeshDomain decomposed
Arm / x86 Parallel execution Batch ready
1,146cores available
4,235 GBRAM available
12scalable presets
Arm + x86architecture options
Architecture choice
Match the platform to the solver.
Choose for compatibility, memory demand and scale. We confirm solver, licensing and workload requirements before execution.
AMD EPYC · 3rd gen
Flexible x86 capacity from compact jobs to large distributed runs.
- Scale
- 95–1,140 cores
- Memory
- 341–4,093 GB
- Nodes
- 1–12
AMD EPYC · 4th gen
High core density and memory capacity for the largest studies.
- Scale
- 191–1,146 cores
- Memory
- 706–4,235 GB
- Nodes
- 1–6
Compute matrix
Start at the right scale. Expand when the model demands it.
The configurations below represent the currently supplied capability envelope. Final allocation depends on platform availability, solver compatibility and licensing.
| Preset | Compute architecture | Credits / hr | Nodes | Cores | RAM |
|---|---|---|---|---|---|
| Extra Small | ArmAWS Graviton3 | 34 | 2 | 126 | 196 GB |
| Small | ArmAWS Graviton3 | 63 | 5 | 315 | 489 GB |
| Medium | ArmAWS Graviton3 | 84 | 8 | 504 | 783 GB |
| Extra Small | x86AMD EPYC · 3rd gen | 32 | 1 | 95 | 341 GB |
| Small | x86AMD EPYC · 3rd gen | 47 | 2 | 190 | 682 GB |
| Medium | x86AMD EPYC · 3rd gen | 62 | 4 | 380 | 1,364 GB |
| Large | x86AMD EPYC · 3rd gen | 78 | 6 | 570 | 2,046 GB |
| Extra Large | x86AMD EPYC · 3rd gen | 146 | 12 | 1,140 | 4,093 GB |
| Small | x86AMD EPYC · 4th gen | 63 | 1 | 191 | 706 GB |
| Medium | x86AMD EPYC · 4th gen | 85 | 2 | 382 | 1,412 GB |
| Large | x86AMD EPYC · 4th gen | 140 | 4 | 764 | 2,823 GB |
| Extra Large | x86AMD EPYC · 4th gen | 180 | 6 | 1,146 | 4,235 GB |
About credits: credits per hour are a relative platform-consumption measure, not a currency price. Commercial terms are confirmed for each engagement.
Designed for simulation
Use more compute where it changes the engineering answer.
Larger meshes
Increase spatial resolution where local workstation memory and core counts become restrictive.
Transient studies
Run longer physical times or finer time steps without turning the solver queue into the project bottleneck.
Parameter sweeps
Compare operating points and configurations in parallel instead of waiting for serial runs.
Automotive CFD
Support intake, exhaust, thermal-fluid, multiphase and flow-distribution investigations.
Engagement workflow
A controlled path from model to result.
Compute is paired with engineering judgment. We scope compatibility and resources before committing the workload.
- 01
Review
Confirm solver, license, model size, memory demand and expected runtime.
- 02
Right-size
Select architecture, node count and memory headroom for the study.
- 03
Execute
Run the agreed case set with monitored parallel execution.
- 04
Transfer
Return solver outputs and engineering-ready result packages through the agreed channel.
Before a run
What we need to size your workload
- Solver and version
- License arrangement
- Cell or element count
- Steady or transient setup
- Expected case count
- Data-transfer requirements
Ready to scale?
Tell us what limits the current run.
Share the solver, approximate model size and the study you need to complete. We will recommend a suitable configuration.