AMD GPU designed for cloud gaming and streaming, not AI compute.
For smaller projects, the Radeon Pro V520 is available around $0.38/hr per GPU (on-demand). This tier offers a 56% discount compared to the higher end of the market ($0.87/hr).
USA 15 configs 1x-4x
Heads up: We do our best to keep these specs & prices accurate. However, cloud costs may fluctuate based on region, usage, and other factors not listed here. These are estimates based on common setups and are for informational purposes only. Always verify current rates & exact specs with the provider before provisioning.
8GB GDDR6. AMD cloud gaming and visualization GPU. Used primarily for game streaming services.
Designed for game streaming, not AI compute. No Tensor Core equivalent. Very limited AI capability.
The median on-demand price has held steady around $0.87/hr per GPU across providers.
With 8GB of VRAM, the Radeon Pro V520 is mainly limited to small quantized models and lightweight GPU workloads.
The Radeon Pro V520 has 8GB of VRAM. Multi-GPU setups increase total memory, but that memory is not automatically pooled across GPUs.
The Radeon Pro V520 has 512 GB/s of memory bandwidth. Higher bandwidth helps with faster data transfer between GPU memory and compute cores.
The Radeon Pro V520 supports 4 precision formats. Training: FP16, FP32. Inference: INT4, INT8.
No. The Radeon Pro V520 is a PCIe-only GPU with no Infinity Fabric, so it is better suited to single-GPU inference and smaller-scale workloads than large distributed training jobs.
Radeon Pro V520 pricing currently ranges from $0.19/hr to $0.87/hr per GPU, depending on the provider, instance type, and billing model.
At 720 hours per month, one Radeon Pro V520 can cost between $137.45 to $624.24 per month, depending on the provider. Reserved and spot pricing can lower that further.
The Radeon Pro V520 is available from 1 cloud provider: Amazon Web Services.
Yes. We currently track 15 Radeon Pro V520 listings across 1 cloud providers:
| Billing type | Listings | Avg $/GPU/hr |
|---|---|---|
| On-demand | 5 | $0.70/hr |
| Reserved | 10 | $0.44/hr |
| Feature | Value |
|---|---|
| GPU Architecture | RDNA |
| Lithography | TSMC 7nm FinFET |
| Stream Processors | 2304 |
| Compute Units | 36 |
| Peak Engine Clock | 1.6 GHz |
| Peak Half Precision (FP16) Performance | 14.75 TFLOPs |
| Peak Single Precision Matrix (FP32) Performance | 7.4 TFLOPs |
| Peak Single Precision (FP32) Performance | 7.4 TFLOPs |
| Peak Double Precision (FP64) Performance | 460 GFLOPs |
| Peak INT4 Performance | 58.98 TOPs |
| Peak INT8 Performance | 24.94 TOPs |
| OS Support | Windows Server 2022 - 64-Bit Edition, Windows Server 2019 - 64-Bit Edition, Windows 10 - 64-Bit Edition, Ubuntu® 22.04 LTS |
| Dedicated Memory Size | 8 GB |
| Dedicated Memory Type | HBM2 |
| Memory Interface | 2048-bit |
| Memory Clock | 1 GHz |
| Peak Memory Bandwidth | 512 GB/s |
Turing data center card for cloud graphics and low-power inference. From $0.06/hr per GPU across 9 providers.
Ampere data center card used for cloud graphics and inference on AWS. From $0.43/hr per GPU across 2 providers.
Dual-GPU Maxwell card for legacy virtual desktop deployments. From $0.75/hr per GPU across 1 provider.
Last updated
