NovaPodNOID GPU
NOID GPU products
Rent an NVIDIA card for Parano1d (NOID). These five products are the cards with a published accelerated hashrate. The marketplace ask is what you pay, per second.
Before you launch
- - The card is RTX 5090, RTX 4090, L40S, RTX 3090, or A100.
- - The host driver is 580 or newer. Older drivers do not start the accelerated miner.
- - The payout address is a public NOID address. A seed phrase never goes on the pod.
- - The pod is stopped when the job is done. Idle time is still billed.
Execution steps
STEP 01
Pick a NOID GPU product
Open the price page for RTX 5090, RTX 4090, L40S, RTX 3090, or A100. Compare the live ask with the median before you rent.
STEP 02
Launch the NOID GPU template
Start the NOID GPU template from Templates. It is an SSH pod with a CUDA base. The miner is yours to run; NovaPod does not hold the payout address.
STEP 03
Point it at a public address
Use only the public payout address. Published pool notes require NVIDIA driver 580 or newer for the accelerated path on compute capability 8.0 and above.
STEP 04
Stop the pod when you are done
The clock is the marketplace ask, counted per second. Stopping the pod stops the GPU charge.
Five cards, one job
NOID GPU mining on the accelerated path is published for compute capability 8.0 and newer: A100, RTX 3090, RTX 4090, L40S, and RTX 5090. A NOID Pool note from September 2026 lists these pool-side rates with the accelerated engine: RTX 5090 about 155 MH/s, RTX 4090 about 121 MH/s, L40S about 112 MH/s, RTX 3090 about 49 MH/s, and A100 about 40 MH/s. Those figures are the pool's, not a NovaPod guarantee. Power limit, clocks, and driver change the result.
Without the accelerated engine the same note puts throughput at roughly one third. GTX 10-series cards are outside this product. Rent one of the five cards above, then read that card's price page for the current ask.
What you pay
The product is the GPU rental. NovaPod bills the live marketplace ask per second while the pod runs. There is no separate NOID fee on the rental. A miner or pool fee, if the software you run charges one, is outside this invoice.
Fund the balance with USDT, USDC, SOL, or ETH, launch the card, and stop it when the run is over. A cheap ask that drops offline is still a spot machine. Keep the work restartable.
Pre-warmed Container Cache & 1s Crypto Settlement
Base CUDA runtimes and foundational weights are distributed across verified high-bandwidth nodes. Boot instances in under 15 seconds with zero KYC and pay with USDT, USDC, SOL, or ETH.
Budget and reliability tips
- - Published hashrates are pool measurements. Your pod's rate depends on the card, the driver, and the power limit.
- - Driver 580 or newer is required for the accelerated miner. Confirm it before a long rental.
- - Stop the pod when shares are no longer being sent. Per-second billing continues on an idle GPU.