NVIDIA launches RTX A4500 20GB and RTX A2000 12GB
NVIDIA RTX A4500 is equipped with GA102 GPU featuring 7168 CUDA cores. It is, therefore, the lowest core count from all existing GA102 GPU graphics cards.
The RTX A4500 has a TDP of 200W, which is 30W less than RTX A5000. Both cards actually have a lot in common, such as the external design, but this is mostly due to the fact that they are both based on the same GPU. With a reduced CUDA core count to 7168, the number of Tensor Cores has also been reduced to 224. In terms of memory, the card offers 20GB GDDR6 memory across a 320-bit bus. With a memory speed of 16 Gbps, the card has 640 GB/s of maximum bandwidth.
NVIDIA claims 23.7 TFLOPS in FP32 compute power and 46.2 TFLOPs in RT performance. The RTX A5000 is therefore 17% faster at 15% higher typical board power.
NVIDIA today also announced it will now offer RTX A2000, the slowest Ampere workstation card, with 12GB GDDR6 memory. There are no changes to the specifications otherwise, in fact, even the 70W TDP has remained intact. This is a GA106-based model with 3328 CUDA cores and up to 8 TFLOPS of single-precision compute power.
Both cards are for desktop workstation systems and they offer quad DisplayPort 1.4 support (either full-sized DP or miniDP in the case of the RTX A2000, which is a low-profile design).
|NVIDIA RTX Ampere Workstation Series|
|VideoCardz||RTX A6000||RTX A5000||RTX A4500||RTX A4000||RTX A2000|
|Boost Clock||~1800 MHz||~1695 MHz||~1653 MHz||~1563 MHz||~1200 MHz|
|FP32 Performance||38.7 TFLOPS||27.8 TFLOPS||23.7 TFLOPS||19.2 TFLOPS||8 TFLOPS|
|Memory Config.||48GB G6 384b||24GB G6 384b||20GB G6 320b||16GB G6 256b||6GB/12GB G6 192b|
|Launch Date||October 2020||April 2021||November 2021||April 2021||August/November 2021|