AMD MI325X
vs
Nvidia H100
Two data center GPUs, one CDNA 3 and one Hopper: 256 GB HBM3e against 80 GB HBM3 and 79% more memory bandwidth on the MI325X.
At a glance
Memory
- MI325X
- 256 GB
- H100
- 80 GB
3.2x more on MI325X
Memory bandwidth
- MI325X
- 6.00 TB/s
- H100
- 3.35 TB/s
1.8x higher on MI325X
Cheapest on-demand
- MI325X
- $2.25 /GPU/hr
- H100
- $1.50 /GPU/hr
33% cheaper on H100
Price comparison
Median price per GPU per hour by billing type, with the cheapest listing under it.
| Billing | MI325X | H100 | Difference |
|---|---|---|---|
| On-demand | $3.07 /GPU/hr from $2.25 | $3.39 /GPU/hr from $1.50 | 9% cheaper on MI325X |
| Reserved | $1.95 /GPU/hr from $1.48 (12 mo) | $3.16 /GPU/hr from $1.53 (1 mo) | 38% cheaper on MI325X |
| Spot | Sold out | $2.15 /GPU/hr from $0.14 |
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. Verify before provisioning. More on how we price.
Price history
MI325X vs H100 price history
Weekly median on-demand price per GPU per hour
Specs comparison
| Spec | MI325X | H100 | Difference |
|---|---|---|---|
| Memory | |||
| Capacity | 256 GB HBM3e | 80 GB HBM3 94 GB on NVL | 3.2x more on MI325X |
| Bandwidth | 6,000 GB/s | 3,350 GB/s 2,000 GB/s on PCIe 3,938 GB/s on NVL | 1.8x higher on MI325X |
| Bus width | 8,192-bit | 5,120-bit | 1.6x wider on MI325X |
| Compute · dense | |||
| FP8 | 2,614.9 TFLOPS | 1,979 TFLOPS | 1.3x higher on MI325X |
| FP16 / BF16 | 1,307.4 TFLOPS | 989.5 TFLOPS | 1.3x higher on MI325X |
| INT8 | 2,614.9 TOPS | 1,979 TOPS | 1.3x higher on MI325X |
| FP32 | 163.4 TFLOPS | 67 TFLOPS | 2.4x higher on MI325X |
| FP64 | 81.7 TFLOPS | 34 TFLOPS | 2.4x higher on MI325X |
| Platform | |||
| Interconnect | Infinity Fabric 896 GB/s | NVLink 900 GB/s NVLink 600 GB/s on PCIe, NVL | Within 5% |
| Board power | 1,000 W | 700 W | 1.4x higher on MI325X |
| Released | Q4 2024 | Q3 2022 | 2 years newer on MI325X |
| Precision | |||
| Supported formats | FP8FP16BF16TF32FP32FP64INT8 | FP8FP16BF16TF32FP32FP64INT8 | |
Figures are for each model's headline configuration. Where a form factor carries a different number, it is named under the figure.
Compute: Vendor peak figures, dense and per GPU. Not measured throughput.
Source: MI325X datasheet and H100 datasheet.
Where to rent them
Providers listing these GPUs, those with both first, so you can switch without moving clouds.
| Provider | MI325X /GPU/hr | H100 /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 | -- |
Lium
|
-- | On-Demand $1.50 |
HyperAI
|
-- | On-Demand $1.80 |
Vast.ai
|
-- | On-Demand $2.00 |
GPU.ai
|
-- | On-Demand $2.00 |
Runcrate
|
-- | On-Demand $2.19 |
Koyeb
|
-- | On-Demand $2.50 |
Massed Compute
|
-- | On-Demand $2.73 |
Lyceum
|
-- | On-Demand $2.79 |
5 providers list the MI325X and 53 the H100. Every listing is on the MI325X page and the H100 page.
Similar GPUs
Other comparisons in the same class.
Our data for AMD MI325X was last updated on Sept. 7, 2026, and for Nvidia H100 on Sept. 7, 2026.