ASRock today introduced the Radeon AI PRO R9700 Creator graphics card. The card is based on the PRO R9700 GPU AMD launched earlier this week, which is targets both the edge AI acceleration and professional visualization markets. The R9700 is based on the 4 nm "Navi 48" silicon, which it maxes out, with all 64 CU being enabled for 4,096 stream processors, 128 AI accelerators, and 64 RT accelerators. A key difference between the RX 9070 XT gaming GPU and the PRO R9700 is memory, with the latter getting 32 GB of it—double the size. This should help with fairly large AI models. The card also uses PCI-Express Gen 5 memory pooling features in workstations with up to four of these.

To make the card capable of exactly this, ASRock designed its PRO R9700 Creator graphics card to have a strictly 2-slot form-factor, with a lateral blower-based cooling solution that has cutouts behind the blower, and clearance of the backplate toward the tail-end, to help with airflow to adjacent cards. The card also implements the 16-pin 12V-2x6 power connector. This is ASRock's second card from this generation to do so, the other being the Radeon RX 9070 XT Taichi. The connector is located at the tail-end of the card, and not the top. This is important, as the card becomes capable of 3U rackmount chassis installations, something not possible with top-oriented power inputs, particularly with the strict cable bending restrictions of 12V-2x6. The card is 27.1 cm long and 11.1 cm tall. Display outputs include four DisplayPort 2.1a.

More...