Two Nvidia data center GPUs on the Pascal architecture: 8 GB GDDR5 against 24 GB GDDR5 and 80% more memory bandwidth on the P40.
At a glance
Updated 13 hours ago- Memory
- 8 GB
- GDDR5
- Memory bandwidth
- 192 GB/s
- From
- $0.65 /GPU/hr
- 1 provider
- Memory
- 24 GB
- GDDR5
- Memory bandwidth
- 346 GB/s
- From
- --
Price comparison
Cheapest listed price per GPU per hour by billing type, and the median across providers.
| Billing | P4 | P40 | Difference |
|---|---|---|---|
| On-demand | $0.65 | -- | |
| Spot | $0.12 | No spot listings | |
| Reserved | $0.34 | No reserved listings | |
| Median | $0.81 | -- |
Median: Median across providers, one vote each, so a large catalog does not outweigh a small one.
Price history
P4 vs P40 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | P4 | P40 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 8 GB GDDR5 | 24 GB GDDR5 | 3x more on P40 |
| Bandwidth | 192 GB/s | 346 GB/s | 1.8x higher on P40 |
| Bus width | 256-bit | 384-bit | 1.5x wider on P40 |
| Compute · dense | |||
| INT8 | 22 TOPS | 47 TOPS | 2.1x higher on P40 |
| FP32 | 5.5 TFLOPS | 12 TFLOPS | 2.2x higher on P40 |
| Platform | |||
| Interconnect | PCIe only | PCIe only | |
| Board power | 75 W | 250 W | 3.3x higher on P40 |
| Released | Q3 2016 | Q3 2016 | |
| Precision | |||
| Supported formats | FP16FP32INT8 | FP16FP32INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: P4 datasheet and P40 datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | P4 /GPU/hr | P40 /GPU/hr |
|---|---|---|
Google Cloud
|
On-Demand $0.65 | -- |
Similar GPUs
Other comparisons in the same class.
Our data for Nvidia P4 was last updated on Sept. 7, 2026, and for Nvidia P40 on Sept. 7, 2026.