while your model
rests, it earns.
The gaming or dev GPU you already own, working between frames. One client turns idle cycles into income — by doing real AI work, not wasting watts on throwaway hashes.
millions of rigs nobody else will serve.
Datacenter-only networks are too big to bother with your card. Root Layer was built for it.
- Idle ~90% of the day, depreciating for nothing
- Locked out of datacenter-only networks
- Mining means burning power on throwaway hashes
- One-click client turns idle cycles into income
- Tiered marketplace — even a modest GPU has a job
- Earn by doing real AI work, not wasting watts
one client, zero config.
Auto-detect the GPU, pick models by profit, and start earning. Mock-up — figures illustrative.
Detected GPU
RTX 4070 · 12GB
Earning · 24h
$3.84
Jobs · 24h
12,540
Validation
99.2%
| Model | VRAM | Throughput | $ / day | Status |
|---|---|---|---|---|
| bge-large-en · embeddings | 2.1 GB | 1,800 tok/s | $4.10 | serving |
| MiniLM · classification | 0.9 GB | 5,200 tok/s | $3.45 | idle |
| whisper-small · speech-to-text | 2.8 GB | 48× realtime | $2.90 | idle |
| Florence · OCR | 1.5 GB | 320 img/h | $2.20 | idle |
built for the rig you already own.
one-click mining client
Auto-detects your GPU and picks the models that pay best for your card. No CUDA wrangling, no config files.
port existing rigs in one click
HiveOS / minerstat integration and a built-in profit-switch — bring a whole farm over without re-tooling.
paid for verified work
Not empty hashes — staked validators run duplicate & canary checks. Real compute, proven, then paid.
token-subsidized rewards
Rewards top up in $RL, released over time, to keep rigs active through the bootstrap and align you with the network.
earn from real demand, not just emissions.
Two streams: declining $RL emissions for supplying compute, plus a share of every paid job. As real demand grows, paid work outweighs emissions.
The window your card sits unused — and could be earning instead.
Illustrative yield for a mid-tier GPU running batch jobs overnight.
The majority of supply is earned by miners and useful work — not insiders.
Figures illustrative — real earnings depend on card, models, and live network difficulty.
your card's weakness is irrelevant.
latency & reliability don't matter
Consumer-GPU variance is a dealbreaker for real-time serving. We chose batch demand, where it simply doesn't.
cheap, parallel throughput
Batch buyers want exactly what millions of idle consumer GPUs already provide: cheap, parallel throughput at scale.