IDLERACK runs a real benchmark on the graphics card in your machine, right in this tab, prices it against live rentals on the open market, and racks it next to every card that has been measured.
Every card gets priced the same way a rental market prices it: by what it can actually do, measured on your hardware, not guessed from the name on the box.
A compute shader multiplies big matrices and streams memory on your GPU for a few seconds. Nothing is installed, nothing leaves the tab except the result.
Hundreds of live rental offers, grouped by GPU class. Your card lands between the two classes it measures closest to, at their real median price.
Give it a name and it takes a slot in the public rack, next to the data center cards and every other visitor's machine. Anyone can walk the rack.
Takes about ten seconds. Close other heavy tabs first. Laptops on battery run slower, which is exactly what a renter would see.
ready. the probe runs on your GPU through WebGPU. press run.
Each 1U slot is a real GPU for rent right now, or a visitor's card that went through the probe. Hover a slot. The bright ones are people.
Median per-GPU price of every offer we can see, class by class. The bar runs from the cheapest offer to the dearest, the white tick is the median.
| Class | VRAM | Offers | Median | Range | $ per TFLOP·h |
|---|
$IDLE is the rack's token on Robinhood Chain. Here is plainly what works today and what doesn't yet.
Everything on this page runs on real numbers: your GPU, live offers, a public relay for the rack.
IDLERACK does not rent your GPU out and holds no one's money. The price is what the open market pays for a card that measures like yours.