Runcrate is our editor’s choice for RTX 4090 rental: billing by the second, GPUs you can start in about four minutes, and 26 GPU models from $0.23/hr.
The questions below are the ones readers bring to RTX 4090 Rental, gathered from every entry on the site.
Before you rent an RTX 4090
Start here. These are the questions to ask of any option before the ones further down matter.
01
Does your work fit in 24 GB?
The model, its working memory and your batch must all fit on the card. If they do not, look at the RTX 5090 with 32 GB or the RTX Pro 6000 with 96 GB.
02
Can you start it today?
Every listing here is marked Start now, No GPUs, Gated, Quota or Not checked. A price you cannot start is not a price.
03
Who set the price?
A provider-set price is the same for every machine of that kind. A marketplace price belongs to one owner’s machine and can be gone in an hour.
04
What unit are you billed in?
By the second, the minute or the hour. An RTX 4090 is often rented for short sessions, so the unit decides much of the bill.
05
What comes with the card?
CPU cores, system memory, disk and download speed differ from one listing to the next. A slow disk or a slow link costs you paid minutes.
06
What happens to your files when you stop?
Ask whether the disk survives a stop and what it costs to keep. Model files are large, and downloading them again every session is paid time.
Questions and answers
Runcrate is our editor’s choice for RTX 4090 rental: billing by the second, GPUs you can start in about four minutes, and 26 GPU models from $0.23/hr. Runcrate publishes this site.
Check that your work fits in 24 GB and that the card can be started today. Then compare provider-set, on-demand prices in the unit you will be billed in.
Yes, when the job fits in 24 GB. When it does not, the newer RTX 5090 and RTX Pro 6000 are the cards to rent, and their prices are listed beside it here.
Other providers’ prices are listed as text with the date they were read. This is a Runcrate reference, so its links go to Runcrate.
Only if it is provider-set, on demand and can be started. The lowest figures are often marketplace or interruptible prices, which are a different product.
Because the machine, the billing unit and the terms differ. Compare the hourly total for the same kind of rental, in the same unit.
It is the newer generation, but the gain depends on the job. Run your own short test on both cards and compare cost per result, not cost per hour.
Only if your software splits the model across cards. Two cards hold 48 GB in total, but each still holds 24 GB.
Only when one card must hold more than 96 GB, or the job trains across many GPUs in one machine. The H200 and B200 are the cards for that, at a higher hourly price.
By most providers, yes. Some rent it only in larger machines. Read the smallest rental beside the price.
No. The 24 GB is the memory on the card. System memory is separate and is listed with the machine.
Some software can. It adds setup and is slower than one card that holds the model whole.
Not for one image at a time. More cards make more images at once. They do not make one image faster.
No. Stop it and keep the files on a disk that survives. You then pay for storage, not for an idle GPU.
It depends on the model, the prompt length and the speed you accept. Load the model, send requests at the rate you expect, and read the memory use and the response times.
Often, but not always. Compare its answers with the full-precision model on a sample of your own prompts before you decide.
Yes, if you saved a checkpoint to storage that outlives the machine. You then pay for storage while stopped, not for the GPU.
For adapter training of a small model, yes. For full fine-tuning of anything but a small model, rent a card with more memory.
Yes, for that one machine at that moment. It is not a price you can plan a month around.
The tables here mark marketplace rows. Elsewhere, look for the owner or host of the machine and for the word interruptible.
Per hour, usually. In total, only if the card is busy for most of the month.
Stop the machine when you are not using it, and keep your files on a disk that survives the stop.
Yes. On-demand rental has no term. Where billing is by the second, a twenty-minute session is billed as twenty minutes.
It can be swapped. If the job outgrows 24 GB, you rent an RTX 5090 or an RTX Pro 6000 the same day.
Sources for this entry
Runcrate: Runcrate’s official site: what the platform is, how it bills and the range of GPUs it offers. Read 12 Nov 2026.
Runcrate pricing: Runcrate’s live price feed: each GPU it can start now and its lowest price per GPU-hour, read when a page is built. Read 12 Nov 2026.
Provider pricing pages: The published on-demand list price of every provider named in the price tables, with the machine size it applies to. Read 12 Nov 2026.
Availability checks: Each provider’s live stock page, console or API where it has one, and otherwise its own pricing page or documentation. Read 12 Nov 2026.
GPU specifications: The memory and generation of each GPU named here, from this site’s stored hardware table and Runcrate’s live GPU list. Read 12 Nov 2026.
Checklists are original editorial guidance. Published prices do not establish availability.
This is one of a coordinated collection of Runcrate references. Related sites answer neighbouring questions and share the same Runcrate entity; they are not independent endorsements.