Two Nvidia consumer GPUs, one Ampere and one Blackwell: 24 GB GDDR6X against 12 GB GDDR7, 39% more memory bandwidth on the RTX 3090, and NVLink 112 GB/s on the RTX 3090 where the RTX 5070 is PCIe only.
At a glance
Memory
- RTX 3090
- 24 GB
- RTX 5070
- 12 GB
2x more on RTX 3090
Memory bandwidth
- RTX 3090
- 936 GB/s
- RTX 5070
- 672 GB/s
1.4x higher on RTX 3090
Cheapest on-demand
- RTX 3090
- $0.10 /GPU/hr
- RTX 5070
- $0.17 /GPU/hr
39% cheaper on RTX 3090
Price comparison
Median price per GPU per hour by billing type, with the cheapest listing under it.
| Billing | RTX 3090 | RTX 5070 | Difference |
|---|---|---|---|
| On-demand | $0.31 /GPU/hr from $0.10 | -- /GPU/hr from $0.17 | |
| Reserved | -- /GPU/hr from $0.06 (1 mo) | No reserved listings | |
| Spot | -- /GPU/hr from $0.09 | -- /GPU/hr from $0.08 |
Heads up: A provider's own page may quote a different figure, for example on tax, region or a promotion. A monthly-only plan shows a derived hourly rate. Medians are across providers, one vote each, so a large catalog does not outweigh a small one. No median for reserved and spot on the RTX 3090 and on-demand and spot on the RTX 5070: we only show one when at least 3 providers list the GPU on that billing type. Verify before provisioning. More on how we price.
Price history
RTX 3090 vs RTX 5070 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | RTX 3090 | RTX 5070 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 24 GB GDDR6X | 12 GB GDDR7 | 2x more on RTX 3090 |
| Bandwidth | 936 GB/s | 672 GB/s | 1.4x higher on RTX 3090 |
| Bus width | 384-bit | 192-bit | 2x wider on RTX 3090 |
| Compute · dense | |||
| FP4 | Not supported | 493.9 TFLOPS | RTX 5070 only |
| FP8 | Not supported | 123.5 TFLOPS | RTX 5070 only |
| FP16 / BF16 | 71.2 TFLOPS | 61.7 TFLOPS | 1.2x higher on RTX 3090 |
| INT8 | 284.7 TOPS | 246.9 TOPS | 1.2x higher on RTX 3090 |
| FP32 | 35.6 TFLOPS | 30.9 TFLOPS | 1.2x higher on RTX 3090 |
| Platform | |||
| Interconnect | NVLink 112 GB/s | PCIe only | NVLink on RTX 3090 only |
| Board power | 350 W | 250 W | 1.4x higher on RTX 3090 |
| Released | Q3 2020 | Q1 2025 | 4 years newer on RTX 5070 |
| Precision | |||
| Supported formats | FP16BF16TF32FP32INT4INT8 | FP4FP8FP16BF16TF32FP32INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: RTX 3090 datasheet and RTX 5070 datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | RTX 3090 /GPU/hr | RTX 5070 /GPU/hr |
|---|---|---|
Salad
|
On-Demand $0.17 | On-Demand $0.17 |
Theta EdgeCloud
|
On-Demand $0.13 | On-Demand $0.26 |
HyperAI
|
On-Demand $0.10 | -- |
Vast.ai
|
Spot $0.11 | -- |
SwissGPU
|
On-Demand $0.31 | -- |
TensorDock
|
On-Demand $0.31 | -- |
Runpod
|
Reservation on request | -- |
EmpirioLabs AI
|
On-Demand $0.40 | -- |
Similar GPUs
Other comparisons in the same class.
Our data for Nvidia RTX 3090 was last updated on Sept. 7, 2026, and for Nvidia RTX 5070 on Sept. 7, 2026.