Skip to content
NVNEXUS
info@nvnexus.com
Talk to an Expert →
GPU Comparison

AMD Instinct MI300X vs NVIDIA H200 NVL

How the AMD Instinct MI300X and NVIDIA H200 NVL compare across memory, compute, power and form factor, with the stronger figure highlighted on each row. Vendor-specific figures (CUDA vs stream processors, Tensor vs matrix cores) are listed on their own rows.

Specification CDNA 3 AMD Instinct MI300X NVIDIA alternative: NVIDIA H100 SXM → NVIDIA H200 NVL HOPPER NVIDIA H200 NVL View Product →
Pricing
Price Not sold by NVNexus $41,400.00
Overview
Architecture AMD CDNA 3 Hopper
Memory
Memory 192 GB HBM3 141 GB HBM3e
Memory Bandwidth 5.3 TB/s 4.8 TB/s
Memory Interface 8192-bit –
Compute
Stream Processors (AMD) 19,456 (304 CUs) –
Matrix Cores / AI Accelerators (AMD) 1,216 –
FP32 163.4 TFLOPS (matrix) 67 TFLOPS
FP64 Tensor 163.4 TFLOPS (matrix) 34 TFLOPS / 67 TFLOPS
TF32 Tensor 653.7 TFLOPS / 1,307 (sparse) –
FP16/BF16 Tensor 1,307 TFLOPS / 2,615 (sparse) –
FP8 Tensor 2,615 TFLOPS / 5,230 (sparse) 3,958 TFLOPS (with sparsity)
INT8 Tensor 2,615 TOPS / 5,230 (sparse) –
Platform
Interface PCIe Gen5 x16 –
NVLink / Infinity Fabric Infinity Fabric, 896 GB/s aggregate NVIDIA NVLink (900 GB/s), PCIe Gen5 (128 GB/s)
TDP 750W Up to 600W (configurable)
Form Factor OAM module (8-GPU platform) PCIe dual-slot air cooled
Multi-Instance GPU (MIG) – Up to 7 MIGs @ 18 GB each
Decoders – 7 NVDEC, 7 JPEG
Cooling Passive (chassis airflow) –
Transistors 153 billion –
Not sure which fits?

Talk it through with a specialist

Tell us your workload and we will recommend the right GPU, configuration, and sourcing options.

Talk to an Expert →

More Comparisons