AMD Instinct MI430X (432GB)

Buying advice

Plan MI430X-based systems for scientific simulation and converged HPC/AI projects that prioritize FP64 arithmetic and 432 GB HBM4 capacity. Match the accelerator to the complete scientific software stack and cluster infrastructure. Its published FP64 peak is a theoretical hardware figure, so compare application-level results and total deployment cost before committing to a system design.

Specifications

LAUNCH

Released
-

GAMING PERFORMANCE

Gaming Performance Index (GPI)
-No applicable GPI estimate.

GRAPHICS BENCHMARKS

3DMark Score
-

MEMORY & STORAGE

Graphics Memory
432 GB HBM4
Memory Bandwidth
23,300 GB/speak theoretical vendorengineeringprojection
Bus Width
-
Memory Transfer Options
-

ARCHITECTURE & COMPUTE

* FP32 describes floating-point throughput, not measured gaming performance. Architecture, clock and method differences affect comparison.

Architecture
-
Process
-
Platform
-
Base Clock
-
Boost Clock
-
Base Clock Range
-
Boost Clock Range
-
Graphics Frequency
-
Max Dynamic Frequency
-
CUDA Cores
-
Max CUDA Cores
-
CUDA Cores per GPU
-
Compute Units
-
Xe Cores
-
FP32 TFLOPS*
-
FP16 (AI) TFLOPS
-
CUDA Compute Capability
-
L2 Cache
-
NVIDIA SMs
-
ROPs
-
RT Core Generation
-
Resolution & frame-rate reference
-
TMUs
-
Tensor Core Generation
-
Transistors
-

POWER & BATTERY

Max GPU Temperature
-
Power
-
Recommended System PSU
-

CONNECTIVITY & EXPANSION

DirectX 12 Ultimate
-
DisplayPort
-
HDCP
-
HDMI
-
PCIe Interface
-
Reference DisplayPort Outputs
-
Reference HDMI Outputs
-
Resizable BAR Support
-
UHBR Mode
-

OTHER SPECIFICATIONS

AV1 Decode
-
AV1 Encode
-
Manufacturer
AMD
NVDEC Engines
-
NVDEC Generation
-
NVENC Engines
-
NVENC Generation
-