AMD has detailed its Ryzen Threadripper PRO 9000 WX-Series and Threadripper 9000 Series processors, along with the Radeon AI PRO R9700 graphics card. These products are designed for professional workstations and AI applications. Both product lines are expected to be available starting July 2025.
Threadripper 9000 Series
The new Threadripper 9000 series processors, based on the Zen 5 microarchitecture, offer performance improvements over previous Zen 4 models. AMD reports an average 16% increase in instructions per cycle (IPC) for workstation applications and up to 25% faster performance in AI/ML tasks.
For example, the 96-core Ryzen Threadripper Pro 9995WX is reported to be 13% to 26% faster in workstation benchmarks and 22% to 23% faster in DeepSeek R1 compared to its predecessor. In comparisons with Intel’s 60-core Xeon W9-3595X, AMD states its 96-core Threadripper can be up to 145% faster in some content creation and rendering tasks.
Key specifications include:
- Core Counts: Up to 96 cores and 192 threads for the PRO WX series; up to 64 cores and 128 threads for the standard 9000X series.
- Clock Speeds: All chips can reach up to 5.4 GHz in turbo mode.
- TDP: All models have a 350-watt Thermal Design Power.
- Memory: 8-channel DDR5 memory support at 6400 MT/s with ECC, up to 2 TB RAM for PRO WX models. The standard 9000X series supports 4-channel DDR5.
- PCIe Lanes: Up to 128 PCIe Gen 5 lanes for PRO WX series; 48 lanes for the standard 9000X series.
- Compatibility: Compatible with existing sTR5 platforms after a BIOS update.
These processors are intended for multi-threaded workloads in professional workstations, including rendering, generative AI, scientific computing, software compilation, and CAD/AEC applications. PRO models will be available through OEMs, while standard Threadripper 9000 series and select PRO models will be available at retail.
Radeon AI PRO 9700
The Radeon AI PRO R9700 graphics card, also released at Computex 2025, is designed for AI computing, machine learning, local inference, and model fine-tuning in professional settings. It is expected to be available in July 2025 from various board partners such as GIGABYTE.
The Radeon AI PRO R9700 is based on the RDNA 4 architecture and uses the 4 nm “Navi 48” GPU. It features 64 RDNA 4 compute units (4,096 stream processors) and 128 dedicated AI accelerators. It delivers up to 96 TFLOPS of FP16 dense performance and up to 1531 TOPS (INT4 sparse) for AI inference. AMD claims up to 2x improved performance over the Radeon Pro W7800 in LLM tasks and up to 5x higher performance than the NVIDIA GeForce RTX 5080 in some AI model comparisons.
Key specifications include:
- Memory: 32 GB of 20 Gbps GDDR6 memory with a 256-bit interface, providing up to 640 GB/s memory bandwidth.
- Power: 300 W TBP (Total Board Power), typically using a 12V-2×6 power connector.
- Connectivity: PCIe 5.0 x16 interface and four DisplayPort 2.1a outputs.
- Multi-GPU Support: Supports up to four GPUs in a single system at the driver level, allowing for a combined memory pool of 128 GB.
- Software Support: Full support for the ROCm software stack on Linux, with Windows support planned.
Both the Threadripper 9000 series and the Radeon AI PRO 9700 are aimed at providing solutions for professional workstations and on-device AI acceleration.
