AMD has recently made an official announcement through a driver update, confirming the release of the Radeon AI PRO R9700 GPU. This cutting-edge graphics card is specifically designed to cater to AI acceleration and high-performance computing needs, targeting both tech enthusiasts and professional users.

The Radeon AI PRO R9700 GPU is based on the RDNA 4 architecture and boasts a Navi 48 chip equipped with 64 compute units (totaling 4096 stream processors) along with 128 AI accelerators. It operates with a total board power (TBP) of 300 W. Complementing its robust processing capabilities, the card is fitted with 32 GB of GDDR6 memory, which not only provides high bandwidth through a 256-bit memory bus but also doubles the memory capacity of its predecessor, the Radeon AI PRO 9070 XT. In terms of computational prowess, the R9700 delivers exceptional performance figures, offering 47.8 TFLOP of FP32 processing power, 191.4 TFLOP for FP16/BF16, and 382.7 TFLOP for both FP8 performance and 382.7 TOPS for INT8. Enhanced with structured sparsity, the INT4 processing can achieve up to 1531 TOPS, showcasing a twofold increase in FP16 performance and fourfold enhancement in INT8/INT4 performance compared to the Radeon PRO W7800. Furthermore, the R9700 supports Wave Matrix Multiply Accumulate (WMMA) instructions to streamline and optimize AI model computation efficiency.

Engineered to handle substantial AI workloads efficiently, the R9700 is equipped with 32GB of video memory, adept at meeting the demands of sophisticated AI tasks like DeepSeek R1 Distill Qwen 32B Q6 and others. Performance evaluations indicate that the R9700 eclipses its predecessor, the Radeon PRO W7800, in the DeepSeek R1 task, boasting up to a 5x speedup in some instances when compared against the competing RTX 5080 with 16GB of video memory. AMD highlights that the 32B 6-bit model, supported by the 32GB video memory, surpasses the 8B 6-bit model in inference accuracy. Additionally, for text-to-image tasks, the FP16 on the 8B model outperforms the 1B version, underscoring the vital role of substantial graphics memory in enhancing AI models' quality.

The R9700 is compatible with the PCIe 5.0 platform and can be scaled to support 4-way multi-GPU configurations. It provides a 128GB shared memory pool, sufficient to handle large-scale models like Mistral 123B and DeepSeek R1 70B, which require up to 112-116GB of memory. The card's dual-slot design incorporates a blower cooler and will be released in July 2025 through partners such as ASUS, ASRock, Gigabyte, PowerColor, Sapphire, XFX, and Interactive.

AMD expands its AI acceleration strategy with the ROCm 7 software suite, spanning three main product lines: Ryzen AI MAX APUs for compact to medium-sized Large Language Models (LLMs); Radeon AI PRO GPUs for multi-GPU edge inference with similar-sized LLMs; and Instinct AI accelerators for large-scale LLM inference and training on a rack scale. The R9700, as the flagship of the Radeon AI PRO series, reinforces AMD's position in the edge AI computing market.

On the processor side, AMD's Zen 5 architecture introduces the Ryzen 5 9600X3D, featuring 3D V-Cache technology within a six-core, twelve-thread design. Following the success of the Ryzen 7 5800X3D, this processor is fine-tuned for gaming and high-performance computing, equipped with stacked caches for quicker data retrieval. Compared with the standard Ryzen 5 9600, the 9600X3D offers a markedly increased cache capacity, estimated at 64MB of L3 cache, cutting latency in gaming and complex applications. Tailored for commercial purposes, the Ryzen 9000 PRO series delivers enhanced security and management tools for enterprise-grade workstations and servers.

Additionally, AMD continues to make strides in gaming tech, crafting a comprehensive gaming chip roadmap for Xbox. While detailed specifications for the Ryzen 5 9600X3D and Ryzen 9000 PRO series remain under wraps, they are anticipated to be available by the third quarter of 2025. These would complement the R9700 GPUs, solidifying AMD's stronghold in high-performance computing and AI domains.

These latest product releases demonstrate AMD's unwavering dedication to advancing in AI and computational performance. The Radeon AI PRO R9700 GPU meets the escalated resource demands of modern AI models through its substantial memory capacity and formidable processing power. Meanwhile, the Ryzen 5 9600X3D and Ryzen 9000 PRO series furnish consumers and professionals with more efficient processing choices, amplifying AMD's foothold in both consumer and enterprise markets. This development not only bolsters AMD's competitiveness but also provides tech enthusiasts with enhanced high-performance hardware options.