Verification for IDLE Protocol

Benchmark IDLE compute
before it earns.

Run one command on your machine. IdleBench checks your GPU, CUDA, Docker, Network and Score — then publishes a signed trust score buyers can verify before routing workloads.

$npx idlebench run
View leaderboard
idlebench — zsh — 80×24
~ $

How it works

Run the benchmark where the hardware lives.

Read the docs →
01Requires Node.js 18+

Install and run

Install Node.js 18+ from nodejs.org, then run npx idlebench run. The CLI prompts for your wallet address, resource name, and type — then starts checking your hardware.

02Takes 60–120 seconds.

Local verification

IdleBench inspects your GPU, VRAM, CUDA, Docker runtime, CPU cores, RAM, and network latency. All checks run on your machine. Nothing is simulated.

03Public URL + embeddable badge.

Score and publish

Your machine receives a final score out of 100 and a grade. The signed report uploads and creates a public verification page tied to your IDLE endpoint.

What gets checked

Your GPU score
should not be
a screenshot.

Every check runs locally on your hardware. If something cannot be detected, it is marked unknown — not assumed.

30%
GPUGPU Capability

Detects GPU model and VRAM via nvidia-smi. Scores based on known performance profiles. Marks unknown if undetectable — no fabrication.

20%
RuntimeDocker & CUDA

Verifies Docker is installed and running. Tests GPU passthrough via docker run --gpus all. CUDA version checked independently.

20%
PerfBenchmark Performance

Runs lightweight CPU hash workloads and matrix calculations. GPU container tests when Docker GPU runtime is available.

15%
NetworkNetwork Quality

Round-trip latency to IdleBench and IDLE gateway servers. Lower latency = faster response times for buyers routing workloads.

15%
SystemStability & Health

CPU core count, total RAM, and system uptime. A machine under high load scores lower on stability.

Live Leaderboard

Top verified resources.

View all
#ResourceScoreGrade
1
Node 1

Fdv3EQ...vQzZbM

60
Risky
2
node-n1

7ns4fT...bgdQmP

54
Poor

Output

Local checks. Public proof.

terminal
$ npx idlebench run
? Solana wallet 7xKXtg2CW87...gAsU
? Resource name RTX 4090 Training Node
? Resource type GPU
✓ System info Ubuntu 22.04 · x64
✓ GPU detected RTX 4090 · 24 GB VRAM
✓ CUDA 12.3 Available
✓ Docker v24.0.7 · GPU runtime
✓ Network 42 ms latency
Final Score 94 / 100 [Verified A]
idlebench.dev/resource/7xKX...

RTX 4090 Training Node

7xKXtg2CW87d...gAsU · GPU

94
/100
Verified A
GPUNVIDIA GeForce RTX 4090
VRAM24 GB
CUDA12.3 · Available
Dockerv24.0.7 · GPU runtime ✓
OSUbuntu 22.04.3 LTS
CPUi9-13900K · 24 cores
RAM64 GB
Network42 ms · 940 Mbps
97
GPU
100
Runtime
92
Perf
95
Net
90
Stability
sha256: a4f8c2e1b9d3...7f6a1e2c4d8b

IDLE Integration

Attach a verified
score to your IDLE
endpoint.

Buyers need proof before they route workloads. A verified score is that proof — backed by a local hardware check, not a self-reported claim.

Attach your IdleBench grade to any IDLE listing. Buyers see your score before they pay.

1

List resource on IDLE Protocol

earnidle.com

2

Run idlebench on your machine

npx idlebench run

3

Score published to IdleBench

Public verification page created

4

Badge attached to your listing

Embeddable SVG badge

5

Buyers verify before routing

Proof before payment

Get started

Before idle compute
earns, it needs proof.

Run one command. Prove your machine. Publish a trust score buyers can verify before they route workloads.

$npx idlebench run
Read the docs

Requires Node.js 18+ · No account · Free and open

FAQ

Common questions.

Something else? Visit IDLE or read the docs.

Does IdleBench run anything without my consent?

No. Every check is interactive. IdleBench runs lightweight local commands only — GPU detection, Docker checks, CPU benchmarks. Nothing leaves your machine except the structured report you review before uploading.

What if my machine has no GPU?

CPU-only machines are handled gracefully. GPU checks report as unavailable, and the score reflects your actual configuration. CPU resources are a valid type on IDLE.

Is IdleBench affiliated with IDLE Protocol?

No. IdleBench is an independent verification layer built for the IDLE ecosystem. It produces reports that buyers and providers can trust — but it is not a product of IDLE Protocol.

How is the final score calculated?

GPU capability (30%) + Docker/CUDA readiness (20%) + benchmark performance (20%) + network latency (15%) + system stability (15%). The same formula runs in the CLI and on the server.

Can I benchmark a cloud VM?

Yes. Cloud VMs score on what they actually have. The score reflects reality — buyers know exactly what they are getting.

How long does a benchmark take?

Most benchmarks finish in 60–120 seconds. Docker GPU tests take longer on first run if the nvidia/cuda container needs to be pulled.