- AMD Radeon AI PRO R9700S delivers 47.8 TFLOPS of maximum computing power FP32
- AMD equips R9700S with 32GB GDDR6 memory for AI workloads
- PCIe 5.0 x16 support enables high-bandwidth communication for AI applications
AMD has unveiled the Radeon AI PRO R9700S, a passively cooled workstation GPU that is part of the R9000 series and promises enterprise-class AI performance.
Following the launch of this series earlier this year, several major brands have confirmed that their products are equipped with R9700 cards, indicating strong industry adoption.
The R9700S has the same Navi 48 RDNA4 configuration as the standard R9700, offering 64 compute units and 4096 stream processors.
Passive cooling for high-density AI workloads
The boost clock reaches up to 2920 MHz and provides maximum FP32 performance of 47.8 TFLOPS, making it a powerful choice for enterprise AI workloads.
Designed for compact racks and multi-GPU configurations, the R9700S uses a 32GB GDDR6 buffer on a 256-bit bus with 64MB of Infinity Cache.
PCIe 5.0 x16 support ensures high-bandwidth communication with compatible workstation platforms.
The “S” designation on the R9700S indicates a quiet design that replaces the cooling fan of previous models.
This approach uses system airflow instead of integrated fans, making the GPU suitable for compact racks where multiple cards work close together.
Despite the passive cooling, the card maintains a TDP of 300W and runs on a single 2×6 12V port.
This approach reflects AMD’s focus on maintaining computing power without adding active components.
The company also launched the R9600D, which houses 48 compute units and 3,072 stream processors with identical memory.
However, the R9700S offers higher performance for large AI models.
Both cards support Linux ECC memory options and are fully compatible with AMD PRO Edition software and ROCm integration.
The R9700S is optimized for tasks such as generative AI inference and training large language models.
This allows businesses to move demanding workloads from CPU cores to GPU accelerators.
The first introduction of the R9700 was in rack systems such as Elsa’s Veluga-D A70S G6 workstation.
Pair the GPU with an AMD Ryzen 7 9700X processor, 64GB of DDR5 RAM, and PCIe expansion slots for additional accelerators.
With these chips, AMD offers enterprise-class solutions for AI computing with an emphasis on memory bandwidth, power and quiet operation.
However, the high performance of this chip raises questions about sustained thermal temperatures during high-density operation.
Therefore, organizations deploying multiple devices must consider system-level airflow and thermal management to maintain stability under AI peak loads.