Two Nvidia data center GPUs, one Ampere and one Blackwell Ultra: 48 GB GDDR6 against up to 288 GB HBM3e, 11x the memory bandwidth on the B300, and NVLink 1,800 GB/s on the B300 where the A40 is PCIe only.
At a glance
Updated 5 minutes ago- Memory
- 48 GB
- GDDR6
- Memory bandwidth
- 696 GB/s
- From
- $0.49 /GPU/hr
- 6 providers
- Memory
- 288 GB
- HBM3e
- Memory bandwidth
- 8.00 TB/s
- From
- $6.50 /GPU/hr
- 30 providers
Price comparison
Cheapest listed price per GPU per hour by billing type, and the median across providers.
| Billing | A40 | B300 | Difference |
|---|---|---|---|
| On-demand | $0.49 | $6.50 | 13x more expensive on B300 |
| Spot | No spot listings | $2.99 | |
| Reserved | $0.61 | $4.25 | 7x more expensive on B300 |
| Median | $0.77 | $7.87 | 10x more expensive on B300 |
Median: Median across providers, one vote each, so a large catalog does not outweigh a small one.
Price history
A40 vs B300 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | A40 | B300 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 48 GB GDDR6 | 288 GB HBM3e | 6x more on B300 |
| Bandwidth | 696 GB/s | 8,000 GB/s | 11x higher on B300 |
| Bus width | 384-bit | 8,192-bit | 21x wider on B300 |
| Compute · dense | |||
| FP4 | Not supported | 14,000 TFLOPS | B300 only |
| FP8 | Not supported | 4,500 TFLOPS | B300 only |
| FP16 / BF16 | 149.7 TFLOPS | 2,250 TFLOPS | 15x higher on B300 |
| INT8 | 299.3 TOPS | 153.5 TOPS | 1.9x higher on A40 |
| FP32 | 37.4 TFLOPS | 75 TFLOPS | 2x higher on B300 |
| FP64 | Not supported | 1.2 TFLOPS | B300 only |
| Platform | |||
| Interconnect | PCIe only | NVLink 1,800 GB/s | NVLink on B300 only |
| Board power | 300 W | 1,100 W | 3.7x higher on B300 |
| Released | Q4 2020 | Q1 2025 | 4 years newer on B300 |
| Precision | |||
| Supported formats | FP16BF16TF32FP32INT4INT8 | FP4FP6FP8FP16BF16TF32FP32FP64INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: A40 datasheet and B300 datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | A40 /GPU/hr | B300 /GPU/hr |
|---|---|---|
Runpod
|
On-Demand $0.49 | On-Demand $7.89 |
Exoscale
|
On-Demand $1.05 | -- |
Database Mart
|
Reserved (24mo) $0.61 | -- |
Utho
|
Reserved (36mo) $0.65 | -- |
Sesterce
|
On-Demand $1.21 | -- |
EmpirioLabs AI
|
On-Demand $1.49 | -- |
GPU.ai
|
-- | On-Demand $7.89 |
Lyceum
|
-- | On-Demand $7.99 |
GPUaaS
|
-- | Custom on request |
Packet·ai
|
-- | Custom on request |
Impossible Cloud
|
-- | Reserved (60mo) $4.32 |
Canopy Wave
|
-- | Custom on request |
Bentaus
|
-- | On-Demand $6.50 |
6 providers list the A40 and 30 the B300. Every listing is on the A40 page and the B300 page.
Similar GPUs
Other comparisons in the same class.
Our data for Nvidia A40 was last updated on Sept. 7, 2026, and for Nvidia B300 on Sept. 7, 2026.