AMDMI250/300
In comparison to traditional GPUs, cloud-based AMD MI250 and MI300 GPUs offer enhanced flexibility and scalability. Developers can easily spin up or down instances to match their changing HPC processing needs without having to worry about hardware maintenance or upgrades. Cloud-based GPUs also offer lower costs and faster deployment, allowing developers to focus on their core AI/ML development work rather than managing hardware.
The AMD MI250/300 is perfect for a wide range of workloads
Deploying AI based workloads on CUDO Compute is easy and cost-effective. Follow our AI related tutorials.
Available at the most cost-effective pricing
Launch your AI products faster with on-demand GPUs and a global network of data center partners
Virtual machines
The ideal deployment strategy for AI workloads with a MI250/300.
- Up to 8 GPUs / virtual machine
- Flexible
- Network attached storage
- Private networks
- Security groups
- Images
Enterprise
We offer a range of solutions for enterprise customers.
- Powerful GPU clusters
- Scalable data center colocation
- Large quantities of GPUs and hardware
- Optimise to your requirements
- Expert installation
- Scale as your demand grows
Specifications
Browse specifications for the AMD MI250/300 GPU
| Starting from | Contact us for pricing |
| Architecture | CDNA2 | CDNA3 |
| Name | AMD Instinct™ MI250 | AMD Instinct™ MI300X |
| Family | Instinct |
| Series | MI200 Series | MI300 Series |
| Lithography | TSMC 6nm FinFET | TSMC 5nm | 6nm FinFET |
| Stream Processors | 13,312 | 19,456 |
| Compute Units | 208 | 304 |
| Peak Engine Clock | 1700 MHz | 2100 MHz |
| Peak Half Precision (FP16) Performance | 362.1 TFLOPs | 1.3 PFLOPs |
| Peak Single Precision Matrix (FP32) Performance | 90.5 TFLOPs | 163.4 TFLOPs |
| Peak Double Precision Matrix (FP64) Performance | 90.5 TFLOPs | 163.4 TFLOPs |
| Peak Single Precision (FP32) Performance | 45.3 TFLOPs | 163.4 TFLOPs |
| Peak Double Precision (FP64) Performance | 45.3 TFLOPs | 81.7 TFLOPs |
| Peak INT4 Performance | 362.1 TOPs | N/A |
| Peak INT8 Performance | 362.1 TOPs | 2.6 POPs |
| Peak bfloat16 | 362.1 TFLOPs | 1.3 PFLOPs |
| Thermal Design Power (TDP) | 500W | 560W Peak | 750W Peak |
| GPU Memory | 128 GB HBM2e | 192 GB HBM3 |
| Memory Interface | 8192-bit | 8192-bit |
| Memory Clock | 1.6 GHz | 5.2 GHz |
| Peak Memory Bandwidth | 3.2 TB/s | 5.3 TB/s |
| Memory ECC Support | Yes (Full-Chip) |
| GPU Form Factor | OAM Module |
| Bus Type | PCIe® 4.0 x16 | PCIe® 5.0 x16 |
| Infinity Fabric™ Links | 8 |
| Peak Infinity Fabric™ Link Bandwidth | 100 GB/s | 128 GB/s |
| Cooling | Passive OAM |
| Supported Technologies | AMD CDNA™ 2 Architecture, AMD ROCm™ - Ecosystem without Borders, AMD Infinity Architecture | AMD CDNA™ 3 Architecture, AMD ROCm™ - Ecosystem without Borders, AMD Infinity Architecture |
| RAS Support | Yes |
| Page Retirement | Yes |
| Page Avoidance | N/A | Yes |
| SR-IOV | N/A | Yes |
Ideal uses cases for the AMD MI250/300 GPU
Explore uses cases for the AMD MI250/300 including Enhanced natural language processing, Faster deep learning training, Real-time video analytics.
Enhanced natural language processing
Utilise the efficient architecture of AMD MI250 and MI300 GPUs to accelerate natural language processing tasks such as text classification, sentiment analysis, and machine translation. This allows developers to build more sophisticated chatbots, voice assistants, and other NLP-driven applications that can understand and respond to human language faster and more accurately.
Faster deep learning training
Train deep neural networks up to 4x faster with cloud-based AMD MI250 and MI300 GPUs compared to traditional CPUs. This enables developers to experiment with larger datasets and complex models, leading to improved model accuracy and better decision-making insights for customer applications.
Real-time video analytics
Deploy the parallel processing capabilities of AMD MI250 and MI300 GPUs and perform real-time video analytics in the cloud. Developers can analyse live video streams, detect objects, classify actions, and track movements with minimal latency, enabling applications such as smart surveillance and autonomous vehicles.
Browse alternative GPU solutions for your workloads
Access a wide range of performant NVIDIA and AMD GPUs to accelerate your AI, ML & HPC workloads
NVIDIA H100 SXM
from $2.45 /hr
Deploy performant H100s on-demand with CUDO Compute.
NVIDIA H100 PCIe
from $2.45 /hr
Deploy performant H100s on-demand with CUDO Compute.
NVIDIA HGX B200
Pricing on request.
Scale with high performance HGX B200 GPUs on our reserved cloud.
NVIDIA GB200 NVL72
Pricing on request.
Scale with high performance GB200 NVL72 GPUs on our reserved cloud.
NVIDIA A800 PCIe
from $0.80 /hr
Deploy performant A800s on-demand with CUDO Compute.
NVIDIA H200 SXM
Pricing on request.
Deploy performant H200s on-demand with CUDO Compute.
NVIDIA B100
Pricing on request.
Scale with high performance B100 GPUs on our reserved cloud.
NVIDIA A100 PCIe
from $1.50 /hr
Deploy performant A100s on-demand with CUDO Compute.
NVIDIA RTX 4000 SFF Ada
Pricing on request.
Deploy performant RTX 4000 SFF Adas on-demand with CUDO Compute.
NVIDIA RTX A4000
Pricing on request.
Scale with high performance RTX A4000 GPUs on our reserved cloud.
NVIDIA RTX A5000
from $0.35 /hr
Deploy performant RTX A5000s on-demand with CUDO Compute.
NVIDIA RTX A6000
from $0.45 /hr
Deploy performant RTX A6000s on-demand with CUDO Compute.
An NVIDIA preferred partner for compute
We're proud to be an NVIDIA preferred partner for compute, offering the latest GPUs and high-performance computing solutions.
Also trusted by our other key partners:
Frequently asked questions
Talk to sales
Reserve GPUs. Access a MI250/300 GPU Cloud alongside other high performance models for as long as you need it.
Deployment & scaling. Seamless deployment alongside expert installation, ready to scale as your demands grow.
"CUDO Compute is a true pioneer in aggregating the world's cloud in a sustainable way, enabling service providers like us to integrate with ease"
Get started today or speak with an expert...
Available Mon-Fri 9am-5pm UK time