AMD MI325X
vs
Nvidia A40
Two data center GPUs, one CDNA 3 and one Ampere: 256 GB HBM3e against 48 GB GDDR6, 8.6x the memory bandwidth on the MI325X, and Infinity Fabric 896 GB/s on the MI325X where the A40 is PCIe only.
At a glance
Memory
- MI325X
- 256 GB
- A40
- 48 GB
5.3x more on MI325X
Memory bandwidth
- MI325X
- 6,000 GB/s
- A40
- 696 GB/s
8.6x higher on MI325X
Cheapest on-demand
- MI325X
- $2.25 /GPU/hr
- A40
- $1.05 /GPU/hr
2.2x more expensive on MI325X
Price comparison
Median price per GPU per hour by billing type, with the cheapest listing under it.
| Billing | MI325X | A40 | Difference |
|---|---|---|---|
| On-demand | $3.07 /GPU/hr from $2.25 | -- /GPU/hr from $1.05 | |
| Reserved | $1.95 /GPU/hr from $1.48 (12 mo) | -- /GPU/hr from $0.61 (24 mo) | |
| Spot | Sold out | No spot listings |
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 on-demand and reserved on the A40: 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
MI325X vs A40 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | MI325X | A40 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 256 GB HBM3e | 48 GB GDDR6 | 5.3x more on MI325X |
| Bandwidth | 6,000 GB/s | 696 GB/s | 8.6x higher on MI325X |
| Bus width | 8,192-bit | 384-bit | 21x wider on MI325X |
| Compute · dense | |||
| FP8 | 2,614.9 TFLOPS | Not supported | MI325X only |
| FP16 / BF16 | 1,307.4 TFLOPS | 149.7 TFLOPS | 8.7x higher on MI325X |
| INT8 | 2,614.9 TOPS | 299.3 TOPS | 8.7x higher on MI325X |
| FP32 | 163.4 TFLOPS | 37.4 TFLOPS | 4.4x higher on MI325X |
| FP64 | 81.7 TFLOPS | Not supported | MI325X only |
| Platform | |||
| Interconnect | Infinity Fabric 896 GB/s | PCIe only | Infinity Fabric on MI325X only |
| Board power | 1,000 W | 300 W | 3.3x higher on MI325X |
| Released | Q4 2024 | Q4 2020 | 4 years newer on MI325X |
| Precision | |||
| Supported formats | FP8FP16BF16TF32FP32FP64INT8 | FP16BF16TF32FP32INT4INT8 | |
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: MI325X datasheet and A40 datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | MI325X /GPU/hr | A40 /GPU/hr |
|---|---|---|
Bentaus
|
On-Demand $2.25 | -- |
Cyfuture AI
|
On-Demand $3.02 | -- |
DigitalOcean
|
On-Demand $3.80 | -- |
TensorWave
|
Custom $2.25 | -- |
Vultr
|
Spot $2.00 | -- |
Exoscale
|
-- | On-Demand $1.05 |
Database Mart
|
-- | Reserved (24mo) $0.61 |
Utho
|
-- | Reserved (36mo) $0.65 |
Runpod
|
-- | Reservation on request |
Sesterce
|
-- | On-Demand $1.21 |
EmpirioLabs AI
|
-- | On-Demand $1.49 |
Similar GPUs
Other comparisons in the same class.
Our data for AMD MI325X was last updated on Sept. 7, 2026, and for Nvidia A40 on Sept. 7, 2026.