Back to guide

Article

What is a GPU-hour?

A GPU-hour is one hour of access to a specific GPU. If a buyer rents four hours on an RTX 4090, that is four RTX 4090 GPU-hours. IdleChip keeps this unit concrete because different cards have different VRAM, performance, drivers, power limits, and software support.

Why not use a generic compute unit?

Generic compute units only work after a platform defines a fair normalised benchmark across hardware. IdleChip is currently scoped to verified GPUs, so the clearest unit is GPU-hours plus specs. Buyers should compare price together with model, VRAM, online status, FP32 estimates, and future benchmark scores.

How pricing works

Seller asks are prices per GPU-hour for remaining listed capacity. Buyer bids are requested GPU-hours and optional budgets. Market pages show the best bid, best ask, spread, price history, and cheapest seller offers first.

Unit

GPU-hours

Price

$/GPU-hour

Context

Specs and benchmarks

Use this while browsing markets

The order book and each GPU market page use this same unit. Open a market, compare the best bid and ask, then come back here if the unit needs a refresher.

GPU-Hours Explained - IdleChip