AMD released new details about its Radeon AI PRO R9700 graphics card and compared it to the older Radeon PRO W7800 model. The company wants to show customers how much better the new card performs with artificial intelligence tasks. AMD built three different product lines for AI work that target various market segments. The Radeon AI PRO series focuses on edge computing and medium-sized language models. These cards help businesses run AI programs without needing massive data centers.
The R9700 uses AMD's Navi 48 chip with 64 compute units and 4096 stream processors built inside. Engineers packed 128 AI accelerators into the card and gave it 32 gigabytes of GDDR6 memory. The card needs up to 300 watts of power to run at full speed. AMD designed the R9700 to handle advanced AI models like DeepSeek R1 and Mistral Small that need lots of memory to work properly.
Performance tests show the R9700 runs twice as fast as the W7800 when processing DeepSeek R1 models. The new card also beats NVIDIA's RTX 5080 by five times in some AI tasks because it has more memory available. AMD claims the R9700 delivers four times more AI processing power than previous cards. Multiple R9700 cards can work together to create systems with 128 gigabytes of total memory for handling the biggest AI models available today.
The R9700 uses AMD's Navi 48 chip with 64 compute units and 4096 stream processors built inside. Engineers packed 128 AI accelerators into the card and gave it 32 gigabytes of GDDR6 memory. The card needs up to 300 watts of power to run at full speed. AMD designed the R9700 to handle advanced AI models like DeepSeek R1 and Mistral Small that need lots of memory to work properly.
Performance tests show the R9700 runs twice as fast as the W7800 when processing DeepSeek R1 models. The new card also beats NVIDIA's RTX 5080 by five times in some AI tasks because it has more memory available. AMD claims the R9700 delivers four times more AI processing power than previous cards. Multiple R9700 cards can work together to create systems with 128 gigabytes of total memory for handling the biggest AI models available today.
