Two Nvidia data center GPUs, one Blackwell and one Pascal: up to 192 GB HBM3e against 24 GB GDDR5, 23x the memory bandwidth on the B100, and NVLink 1,800 GB/s on the B100 where the P40 is PCIe only.
At a glance
Updated 4 days ago- Memory
- 192 GB
- HBM3e
- Memory bandwidth
- 8.00 TB/s
- From
- --
- 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.
Price history
B100 vs P40 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | B100 | P40 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 192 GB HBM3e | 24 GB GDDR5 | 8x more on B100 |
| Bandwidth | 8,000 GB/s | 346 GB/s | 23x higher on B100 |
| Bus width | 384-bit | ||
| Compute · dense | |||
| FP4 | 7,000 TFLOPS | Not supported | B100 only |
| FP8 | 3,500 TFLOPS | Not supported | B100 only |
| FP16 / BF16 | 1,750 TFLOPS | ||
| INT8 | 3,500 TOPS | 47 TOPS | 74x higher on B100 |
| FP32 | 60 TFLOPS | 12 TFLOPS | 5x higher on B100 |
| FP64 | 30 TFLOPS | Not supported | B100 only |
| Platform | |||
| Interconnect | NVLink 1,800 GB/s | PCIe only | NVLink on B100 only |
| Released | Q4 2024 | Q3 2016 | 8 years newer on B100 |
| Precision | |||
| Supported formats | FP4FP6FP8FP16BF16TF32FP32FP64INT8 | FP16FP32INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: B100 datasheet and P40 datasheet.
Similar GPUs
Other comparisons in the same class.
Our data for Nvidia B100 was last updated on Sept. 7, 2026, and for Nvidia P40 on Sept. 7, 2026.