Sapphire Radeon AI Pro R9700 32GB GDDR6 Graphics Card
MPN 32358-01-20G · Brand SAPPHIRE · 2 Part Slot Occupied266.7 (L) × 111 (W) × 39 (H) mm
The SAPPHIRE Radeon AI Pro R9700 delivers 32 GB GDDR6, 4,096 stream processors, 128 AI accelerators, and 1,531 TOPS INT4 performance on AMD's RDNA 4 architecture — purpose-built for demanding local AI and machine learning workloads
Inclusive of GST · GST invoice available
Overview
The SAPPHIRE Radeon AI Pro R9700 is a professional-grade AI accelerator graphics card built on AMD's RDNA 4 architecture and engineered specifically for local AI inference, machine learning training, and large-scale data-set processing. Unlike typical gaming GPUs, the Radeon AI Pro series is purpose-designed to handle multiple machine learning frameworks simultaneously, bringing data-centre-class AI performance to workstation and desktop environments. At the heart of the card sit 4,096 stream processors spread across 64 compute units, backed by 128 dedicated AI accelerators and 64 ray accelerators. This combination enables a peak FP16 throughput of 95.7 TFLOPS and a peak INT4 inference throughput of 1,531 TOPS — figures that make it a compelling option for running large language model inference, diffusion pipelines, and other AI workloads without cloud dependency. Peak INT8 performance reaches 766 TOPS, offering flexibility across quantisation strategies. Memory capacity is a standout attribute: 32 GB of GDDR6 on a 256-bit bus delivers up to 640 GB/s of peak memory bandwidth, complemented by 64 MB of AMD Infinity Cache. This generous VRAM headroom is critical when loading large model weights locally, processing high-resolution datasets, or running multi-model pipelines concurrently. The card connects via PCIe 5.0 x16, ensuring the interface is not a bottleneck on modern platforms. Display output is provided by four DisplayPort 2.1a connectors supporting resolutions up to 7680 × 4320 (8K). The blower-style active cooler with dual ball bearings is a deliberate design choice for workstation and rack environments where exhaust heat must be directed out of the chassis rather than recirculated. Power delivery uses a single 12V-2×6 connector with a rated TDP of 300 W; a minimum 750 W system PSU is required. On the software side, the card supports AMD Software Adrenalin Edition along with technologies including AMD FidelityFX Super Resolution 4, AMD Fluid Motion Frames, AMD Smart Access Memory, AMD FreeSync, and AV1/HEVC/H.264 hardware encode and decode — extending its utility to content creation and media workflows beyond pure AI computation. Verified Specifications: - GPU: AMD Radeon AI Pro R9700 (RDNA 4 Architecture) - SKU / Model No.: 32358-01-20G - Compute Units: 64 | Stream Processors: 4,096 - AI Accelerators: 128 | Ray Accelerators: 64 - Transistor Count: 53.9 Billion - Memory: 32 GB GDDR6, 256-bit bus, 640 GB/s bandwidth - Infinity Cache: 64 MB - Boost Clock: up to 2,920 MHz | Game Clock: up to 2,350 MHz - Peak FP16: 95.7 TFLOPS | Peak FP32: 47.8 TFLOPS - Peak INT4: 1,531 TOPS | Peak INT8: 766 TOPS - Texture Units: 256 - Display Outputs: 4× DisplayPort 2.1a (max 7680×4320) - Interface: PCIe 5.0 x16 - Power Connector: 12V-2×6 | TDP: 300 W - Recommended PSU: 750 W or greater - Cooler: Active Blower Fan, Two Ball Bearing - Dimensions: 266.7 mm (L) × 111 mm (W) × 39 mm (H), 2-slot - OS Support: Windows 11 64-bit, Linux x86 64-bit - Warranty: 3 Years
Datasheet
| Memory Type | GDDR6 |
|---|---|
| Warranty | 3 Years |
| Chipset | AMD RADEON |
| Memory Size | 32 GB |
| Interface | PCI-Express 5.0 x16 |
| GPU | AMD RADEONTM AI PRO R9700 AMD RDNATM 4 Architecture |
| Memory Clock | 640 Gbps |
| Recommended PSU | 750W |
| Resolution | 7680 x 4320 |
| Power Consumption | 300 W |
| Model | 32358-01-20G |
| Ports | DisplayPort |
| Cooler | Blower Fan |
| Note** | Features, Price, Specifications are subject to change without notice. Images used for representation purposes only |
| Pci Express | 5.0 |
| Engine Clock | Boost Clock: Up to 2920MHzGame Clock: Up to 2350MHzBoost Clock is the maximum frequency achievable on the GPU running bursty workload. Boost clock achievability, frequency, and sustainability will vary based on several factors, including but not limited to: thermal conditions and variation in application and workloads.Game Clock is the expected GPU clock when running typical gaming applications, set to typical TGP (Total Graphics Power). Actual individual game clock results may vary |
| Memory Interface | 256-BIT |
| Power Connectors | 12V-2×6 |
| Max Display Support | 4 |
| GPU Core (Cuda Core) | 4096 |
| Tops | Peak INT4 Performance: 1531 TOPSPeak INT8 Performance: 766 TOPS |
| Tflops | Peak Half Precision: 95.7 TFLOPSPeak Single Precision: 47.8 TFLOPS |
| Displays | 4x DisplayPort™ 2.1a |
| Model No. | 32358-01-20G |
| Cooler Type | Active Blower Fan, Two Ball Bearing |
| Compute Units | 64 |
| Miscellaneous | TPP: 3062TPP Density: 8.91 |
| Texture Units | 256 |
| Video Formats | H264 Encode and DecodeH265 / HEVC Encode and DecodeAV1 Encode & Decode |
| Ai Accelerator | 128 |
| Infinity Cache | 64MB |
| Amd Technologies | AMD Software Adrenalin EditionAMD FidelityFX™ Super Resolution 4AMD Fluid Motion FramesAMD Radeon™ Super ResolutionAMD Smart Access MemoryAMD Smart Access VideoAMD Radeon™ BoostAMD Radeon™ Anti-LagAMD Radeon™ Image Sharpening 2AMD Enhanced Sync TechnologyAMD FreeSync™ TechnologyAMD Radeon™ ChillAMD Virtual Super Resolution |
| Operating System | Windows 11 – 64-Bit Edition, Linux x86 64-Bit |
| Ray Accelerators | 64 |
| Transistor Count | 53.9 Billion |
| Memory Size / Bus | 32GB GDDR6256-bit Memory Bus64MB AMD Infinity Cache Technology640 GB/s Peak Memory Bandwidth |
| Stream Processors | 4096 |
| System Requirement | PCI® Express® compatible motherboard with one x16 PCIe slotMinimum 750W or greater power supplyMinimum 8GB of system memory. Recommended 16GB.Supported operating systems include Linux®, Windows 10/11. 64-bit operating system required |
| External Power Connector | 12V-2×6 |
| Form Factor | 2 Part Slot Occupied266.7 (L) × 111 (W) × 39 (H) mm |
Complete your build




