NVIDIA today formally announced the new professional graphics accelerator RTX A4500. The manufacturer did not release any official press releases on this matter. The novelty just appeared on the website of one of its partners, PNY company.
The professional RTX A4500 GPU is powered by the GA102 processor with 7168 CUDA cores. At the moment, this is the smallest number of CUDA cores among all GA102-based video cards. The chip has 224 tensor cores, 56 RT cores. The video card is equipped with 20 GB GDDR6 video memory with support for 320-bit bus.
Taking into account the data transfer rate of 16 Gb / s per contact, the bandwidth of the memory subsystem is 640 GB / s. According to NVIDIA, the theoretical performance of the RTX A4500 accelerator is 23.7 teraflops (FP32) or 46.2 teraflops (FP16).
The claimed power consumption of the new product is 200 W, which is 30 W less than that of the more powerful RTX A5000. Both cards are based on the same GPU and have a similar design.
NVIDIA also announced the start of sales of the most compact graphics card based on the Ampere architecture, the RTX A2000 model. This is also a professional graphics card designed for workstations. Recall that her company was announced back in August . At the heart of the RTX A2000 is a GA106 GPU with 3328 CUDA cores and 12GB of GDDR6 memory. The power consumption of the video card is only 70 watts. The declared performance level is 8 teraflops (FP32).