Nvidia's GPU Technology Conference, or GTC is taking place this week, after the company was forced to cancel the in-person event back in March due to the pandemic. As it started today, the company has dropped a massive amount of announcements across its many areas of business, including a new professional GPU, the RTX A6000.
This is Nvidia's first Ampere-based professional GPU, which means it's using the same architecture as the recently-launched RTX 3080 and 3090 consumer cards. Usually, Nvidia's professional GPUs use the Quadro branding, but for whatever reason, Nvidia seems to have dropped it from the name here, though you'll still find it under the Quadro category on Nvidia's website.
Being based on the Ampere architecture, the RTX A6000 has new CUDA cores that promise to double the speed of single-precision floating point (FP32) operations; new RT cores that deliver up to twice the throughput in ray tracing; and third-generation Tensor Cores, with Tensor Float 32 (TF32) precision, which promises five times the AI training throughput compared to the previous generation. Additionally, it has 48GB of GPU memory in a single card, and if that's not enough, the new third-generation NVLink bridge can link up two cards for a total of 96GB, and it touts more interconnect bandwidth between GPUs.
Nvidia also introduced the Nvidia A40, a GPU for data centers that features many of the same enhancements, including the 48GB of memory and NVLink support. Both cards also support PCIe Gen 4 x16.
As mentioned above, there are a ton of announcements coming from GTC. Nvidia introduced the Maxine platform for video conferencing software, which leverages AI to deliver capabilities such as face alignment, gaze correction, denoising, and notably, super resolution. Nvidia says this software can greatly reduce the bandwidth (by up to a factor of 10 compared to H.264 compression) used in videoconferencing, by capturing key points of a user's face and reanimating the facial expressions on the other end.
Another major announcement is that Nvidia is bringing many of its tools to the Arm platform, including Nvidia AI, Nvidia Rapids, RTX, and the HPC SDK. These tools target many areas of technology, including data centers, high-performance computing, robotics, and of course, PCs. Among the use cases that these tools can enhance, Nvidia includes cloud gaming, an increasingly crowded market.
Additionally, Nvidia introduced the CloudXR platform, which will leverage Nvidia GPUs on Amazon EC2 P3 and G4 instances to stream AR and VR experiences to users. Companies like Lucid Motors are already using the technology to create a design showroom for its Lucid Air electric car, though of course there are other, potentially more interesting applications.
Recently, Microsoft introduced Microsoft Editor to help users with grammar corrections and suggestions, and today, Nvidia also announced that this tool can now tap into its Triton Inference Server, at least in Word for the web. Many of the announcements from GTC are highly technical, but you can check the full list of news here if you're interested.