Which workloads suit SolarCompute?
Practical guidance for developers, researchers, and paying customers on where our distributed solar-powered network adds value.
Quick comparison
| Workload | Best fit | Scheduling tolerance | Typical use |
|---|---|---|---|
| GPU compute | Tower / Rack systems | Medium best as scheduled batch jobs | Model training, fine-tuning, batch inference |
| CPU & batch | Case / Tower systems | High tolerant of variable availability | ETL, data pipelines, analytics |
| Rendering & visualisation | Tower / Rack systems | Medium parallelisable batch jobs | Offline rendering, VFX, video encoding |
| Real-time / low-latency | Regionalised / edge placement | Low needs closer placement or hybrid setup | Interactive inference, serving with strict latency targets |
GPU Workloads
Request quote for GPURecommended for model training, fine-tuning, large-batch inference, and accelerated rendering. SolarCompute machine classes provide GPU-optimised capacity when energy and site suitability align.
- Common uses: deep learning training, model fine-tuning, batch inference, 3D rendering and graphics workloads.
- Best fit machine classes: Tower System (mid-tier GPU capacity) and Rack System (high-capacity GPU deployments).
- Developer notes: schedule batch training and large jobs to align with site energy availability for cost and emissions benefits.
- How to start: contact hello@solarcompute.net.au or use Request a Quote to discuss GPU needs and job sizing.
CPU & Batch Processing
Request quote for CPUCPU-bound workloads, data processing, ETL, and background jobs run effectively on general-purpose compute nodes. These workloads are often flexible in timing and can take advantage of intermittent renewable supply.
- Common uses: data pipelines, feature extraction, mapping, batch analytics, and jobs tolerant to scheduling.
- Best fit machine classes: Case System for smaller, localised CPU tasks; Tower System for moderate throughput.
- Developer notes: use job scheduling and retry logic to handle variable availability windows.
- How to start: provide expected job duration, concurrency, and latency tolerance when requesting a quote.
Rendering & Visualisation
Request quote for RenderingGPU-accelerated rendering, video processing, and visualization workloads are a strong fit where parallel GPU throughput is required.
- Common uses: offline rendering, VFX, batch video encoding, and ray-tracing workloads.
- Best fit machine classes: Tower and Rack systems depending on throughput requirements.
- Developer notes: prefer batch job submission and parallel task splitting to maximise throughput and align with energy windows.
Real-Time & Low-Latency
Request quote for Low-latencyWorkloads requiring strict low-latency (live inference, interactive services) may need regionalised deployments and closer network placement. Discuss target latency and SLA expectations with our team.
- Common uses: live inference, interactive model serving, and services with tight latency requirements.
- Considerations: distributed architectures, caching, and edge placement are key to reducing latency.
- How to start: provide expected 95th/99th percentile latency targets when contacting hello@solarcompute.net.au.
Developer & Integration Notes
- Use job schedulers and queuing systems to tolerate variable energy availability.
- Design workloads to be checkpointable for long-running training and batch jobs.
- Parallelise tasks and prefer bulk submission for rendering and batch inference.
- For production services with strict SLAs, discuss regional placement and hybrid architectures with our growth team.
Ready to deploy workloads on SolarCompute?
Request a quote with your workload profile, or review machine classes to match your job requirements.