Nvidia introduced its Turing architecture back in late August together with the new Quadro lineup of professional GPUs. Meanwhile the high-end RTX 2000 gaming GPUs based on Turing hit the market and were not received that well; however, the professional cards are just now starting to see availability, as Nvidia enabled the pre-order option for the Quadro RTX 6000 flagship on its website.
Just like the RTX 2080 Ti, the Quadro RTX 6000 profession GPU features the TU102 core with 18.6 billion transistors, being optimized for film production, automotive and architectural designs, as well as scientific visualization. The professional GPU is slightly more potent, integrating 4,608 CUDA cores (versus 4,352), 576 Tensor cores (versus 544) and 72 Ray Tracing cores (versus 68), plus it comes with 24 GB of GDDR6 VRAM instead of 11 GB. Presumably, the default and boost clocks of the professional core are higher than 1,350 MHz / 1,635 MHz, although the single-cooler design may suggest that the frequencies could actually be lower.
Nvidia claims that the Quadro RTX 6000 has a throughput of 16.3 TFLOPS in single precision FP32 mode or double the speed in half precision FP16 mode. With the addition of the Tensor and RT cores, the GPU can push 10 Gigarays/s, 84 Tera RTX-OPS and 130. Tensor TFLOPS. It has a TDP of 295 W andbenefits from the latest implementation of the NVLink connector that allows up to two cards to combine 48 GB of VRAM and up to 100 GB/s bandwidth. Port selection includes 4x DisplayPort 1.4 and a USB-C connector for Nvidia VirtualLink.
The Quadro RTX 6000 cards are now available for pre-orders for US$6,300 on the official site, but Nvidia imposed a 5-card limit per customer.
Are you a techie who knows how to write? Then join our Team! Wanted:
- News Writer (Romania based)
Details here