CPU compute
Simulations, data processing and other CPU-bound work.
₹1.49per core-hour
- MPI simulations split by domain
- Data jobs split into shards
- Watchdog requeues chunks on node failure
- Pre-verified Python and OpenFOAM images
Pay for the compute you use.
Jobs run on idle campus hardware, so you pay for execution time only. Contributors earn credits for the time their machines spend on your chunks.
GPU compute
ML training, Blender rendering and parallel GPU workloads.
₹12.50per RTX 3090-hour
- CUDA 12 PyTorch image
- Blender renders split by frame range
- DDP or Local-SGD for multi-node training
- Priority queue
Against the cloud
Per-hour cost for one GPU. Cloud figures are approximate on-demand list prices, converted to rupees.
| Provider | Per hour |
|---|---|
| AWS EC2 p3.2xlarge | ₹224.50 |
| Google Cloud, V100 | ₹286.00 |
| Azure NV6 | ₹198.25 |
| SpareCycles, RTX 3090 | ₹12.50 |