H200
141GB HBM3e per GPU
View GPU All GPUsInstinct MI300X
CDNA 3 architecture192GB HBM3 per GPU
GPU Catalog / AMD
Instinct MI300X
GPU infrastructure.
Available in 4 weeks through reservation: an AMD CDNA 3 accelerator with 192GB of HBM3 memory for generative AI and HPC workloads.
- Architecture
- CDNA 3
- GPU memory
- 192GB HBM3 per GPU
- Availability
- Available in 4 weeks - reserve only
- Reservation rate
- $2.50 / GPU / hr
- Lead time
- 4 weeks
Reservation
Reserve capacity.
Available in 4 weeks through reservation.
- Access
- Available in 4 weeks - reserve only
- Rate
- $2.50 / GPU / hr
- Estimated lead time
- 4 weeks
- Qualification
- Available in 4 weeks through reservation.
Hardware profile
Where Instinct MI300X fits.
AMD Instinct MI300X combines CDNA 3 architecture with 192GB of HBM3 memory per GPU. AMD positions it for generative AI and high-performance computing applications, supported by the ROCm software ecosystem.
Choose MI300X when the workload benefits from a 192GB memory envelope and an AMD ROCm software path, with capacity planned through reservation.
AMD product information- 01
Generative AI
Reserve high-memory AMD accelerators for large-model training and inference workloads.
- 02
High-performance computing
Run scientific and technical workloads that use CDNA 3 compute and HBM3 memory bandwidth.
- 03
ROCm-based development
Build and validate GPU-accelerated applications against the AMD ROCm software ecosystem.
Deployment sizing
Size before you provision.
Hardware fit depends on the model, precision, batch size, software stack, and total working set. Published use cases are not performance guarantees.
- 01 / Memory
Confirm the working set
Include weights, caches, activations, batches, and framework overhead.
- 02 / Software
Validate the stack
Check CUDA, driver, framework, precision, and container requirements.
- 03 / Capacity
Plan the lead time
MI300X capacity is available in 4 weeks through reservation.