GPUhub Our sponsor
Singapore 4 configs 1x-8x
Top Ada Lovelace consumer GPU for local AI research and rendering.
For smaller projects, the RTX 4090 is available around $0.16/hr per GPU (on-demand). This tier offers a 81% discount compared to the higher end of the market ($0.87/hr). Spot instances start lower, at $0.11/hr per GPU.
Singapore 4 configs 1x-8x
USA 1 config 1x
USA 149 configs 1x-16x
USA 20 configs 1x-8x
USA 7 configs 1x-8x
USA 1 config 1x
USA 28 configs 1x-8x
France 3 configs 1x-8x
UAE 7 configs 1x-16x
USA 1 config 1x
Singapore 2 configs 1x-2x
USA 1 config 1x
USA 1 config 1x
USA 1 config 1x
Switzerland 1 config 1x
USA 8 configs 1x-2x
UAE 6 configs 1x-4x
Heads up: We do our best to keep these specs & prices accurate. However, cloud costs may fluctuate based on region, usage, and other factors not listed here. These are estimates based on common setups and are for informational purposes only. Always verify current rates & exact specs with the provider before provisioning.
24GB GDDR6X with Ada Lovelace FP8 Tensor Cores. Highest-performing consumer GPU of the Ada generation. Can run 13B parameter models at FP16 or 30B+ quantized. Popular for AI research and Stable Diffusion.
Consumer GeForce GPU. No ECC memory. 24GB VRAM is limiting for larger LLMs. Subject to Nvidia's consumer EULA, which may restrict data center use. For production inference, consider the L40S.
The median on-demand price across providers has fallen about 25% since August 2025, from $0.52 to $0.39/hr per GPU. This reflects the market-wide median, which moves both when providers change prices and when lower or higher priced offerings enter the market.
With 24GB of VRAM, the RTX 4090 can typically run 7B to 13B models in FP16, or larger models in 4-bit quantized form.
The RTX 4090 has 24GB of VRAM. Multi-GPU setups increase total memory, but that memory is not automatically pooled across GPUs.
The RTX 4090 has 1,010 GB/s of memory bandwidth. Higher bandwidth helps with faster data transfer between GPU memory and compute cores.
The RTX 4090 supports 7 precision formats. Training: BF16, FP16, TF32, FP32. Inference: FP8, INT4, INT8.
No. The RTX 4090 is a PCIe-only GPU with no NVLink, so it is better suited to single-GPU inference and smaller-scale workloads than large distributed training jobs.
RTX 4090 pricing currently ranges from $0.11/hr to $2.75/hr per GPU, depending on the provider, instance type, and billing model.
At 720 hours per month, one RTX 4090 can cost between $77.83 to $1,978.00 per month, depending on the provider. Reserved and spot pricing can lower that further.
The RTX 4090 is available from 17 cloud providers, including Vast.ai, Oblivus, Novita. Pricing and availability vary by region and billing model.
Yes. We currently track 241 RTX 4090 listings across 17 cloud providers:
| Billing type | Listings | Avg $/GPU/hr |
|---|---|---|
| On-demand | 91 | $0.48/hr |
| Reserved | 104 | $0.43/hr |
| Spot | 46 | $0.29/hr |
| Model Name | GeForce RTX™ 4090 GAMING X TRIO 24G |
| Graphics Processing Unit | NVIDIA® GeForce RTX™ 4090 |
| Interface | PCI Express® Gen 4 |
| Core Clocks | Extreme Performance: 2610 MHz (MSI Center), Boost: 2595 MHz (GAMING & SILENT Mode) |
| CUDA® CORES | 16,384 Units |
| Memory Speed | 21 Gbps |
| Memory | 24GB GDDR6X |
| Memory Bus | 384-bit |
| Output | DisplayPort x 3 (v1.4a), HDMI™ x 1 (Supports 4K@120Hz HDR, 8K@60Hz HDR, Variable Refresh Rate) |
| HDCP Support | Yes |
| Power Consumption | 450 W |
| Power Connectors | 16-pin x 1 |
| Recommended PSU | 850 W |
| Card Dimensions (mm) | 337 x 140 x 77 mm |
| Weight (Card / Package) | 2170 g / 3093 g |
| DirectX Version Support | 12 Ultimate |
| OpenGL Version Support | 4.6 |
| Maximum Displays | 4 |
| G-SYNC® Technology | Yes |
| Digital Maximum Resolution | 7680x4320 |
Source: official Nvidia RTX 4090 datasheet.
Last updated