This “GeForce Beyond” presentation ended up being just a small part of the larger GTC 2022 keynote, which covered topics far beyond just PC graphics. Nvidia CEO Jensen Huang addressed topics across the company’s line of cloud and AI services available to applications such as robotics and self-driving cars, as well as highlighting the company’s work in powering the technology behind the metaverse.
The keynote kicked off at 8 a.m. PT today, and just ended. Check below for a rundown of the updates from Nvidia or rewatch the RTX 4090 event video.
LiveLast updated September 20, 2022 9:37 AM
Jensen is finally wrapping up (whew!), summarizing everything Nvidia has announced, ranging from the RTX announcements up to its acceleration libraries and Omniverse Cloud offerings. With that, GTC has officially kicked off, and the keynote is done. Thanks for hanging with our liveblog! Stick with Digital Trends for all our continuing coverage of GTC.
Chatbots and 3D avatars
Nvidia is playing a demo of how its powering the future of chatbots and 3D customer service avatars. This brings together its large language models and its Omniverse service, to help companies create and improve these types of customer service avatars. In this case it’s looking at Ace, a non-coding application for creating these AI-driven customer service bots. It’s awfully specific, but clearly something that can use multiple aspects of Nvidia’s technology and platforms.
The ARM-based Grace CPU
Nvidia has given some details on its ARM-based Grace CPU, and how it works with the Hopper GPU. These are high-performance products, of course, not for consumers. Still, interesting to see Nvidia work on its first CPU, and it’s first product built on ARM.
Systems will be available in the first half of 2023.
Large language models and Hopper
Jensen is talking about its new large language models, including a model specifically designed for biology and chemistry. Nvidia is collaborating with the Broad Institute, bringing new libraries and acceleration to the scientific research.
These large language models need to be fast though, and that’s where Hopper comes in, the architecture behind these powerful GPUs. Jensen says large language models will be the most important AI of our time, and Hopper is the engine it runs on. Jensen made a silly Star Trek reference, which is bringing back some of the dork factor this ultra-series presentation desperately needed.
Nvidia Qoda for quantum simulation
Jensen has announced Nvidia Qoda, which is a hybrid classical-quantum computer. This is fascinating. Jensen’s talking about bringing the future of quantum computing into the present moment. Basically, Qoda can simulate what quantum computers can do, long before it’s actually arrived.
Nvidia is speaking to developers and investors
Jensen has gone from speaking directly to PC gamers, then to creatives and developers with Omniverse, and now it feels like he’s speaking specifically to investors and industry partners. I gotta say — Nvidia hasn’t done a great job understanding how to market and communicate this stuff. Really makes you wonder why Nvidia doesn’t hold a dedicated event to PC gaming, split from this really dense stuff.
It’s not that we’re not interested in what Nvidia is doing with Omniverse and cloud computing, but the company just doesn’t do a great job communicating that to the average person. Jensen loves talking in numbers and acronyms, but come on, man! Paint the big picture for us. This should be really exciting for those interested in tech, but it gets lost in the weeds quickly.
Nvidia announces the Jetson Orin Nano
Jensen has moved on to robotics computers and announced the Jetson Orin Nano. He says it’s 80x faster than its predecessor.
He’s also announced a partnership with Siemens, which is using Orin IGX and Nvidia’s Metropolis platform. He’s now moving on to medical use cases for technologies, including for robot-assisted surgeries. Nvidia Isaac is the company’s industrial robotics platform, and Jensen is providing lots of examples of use cases. He even said it’s like having a “pre-Quantum computer.” Nvidia has a lot invested in these robotics platforms, and it’s all available in the Omniverse Cloud, of course.
Cars designed in Omniverse, Drive Concierge
Bringing on the tools together, Nvidia shows a demo of cars being designed in Omniverse, including the infotainment and digital dashboards.
Next, a real-looking demo of someone driving in a self-driving car, assisted by a digital assistant called Drive Concierge — and some emotional music, of course.
Drive Orin is the second-gen processor specifically for self-driving cars, and Jensen says it’s already been selected by 40 different cars.
Nvidia announces the Drive Thor chip for self-driving cars
Jensen’s moving on to self-driving cars now. Interestingly, they’ve replaced a Drive Atlan, a car for automotive that never actually came out. But he says with Ada Lovelace, they had access to way more performance, so now Atlan is no more.
Jensen is imagining a single computer that runs a car, which currently is managed by multiple computers — all run on software on Drive Thor. Jensen is connecting what Nvidia has done with Drive Sim for testing self-driving cars. Using AI and neural networks, entire 3D scenes are recreated from real sensor data, and then placed into Omniverse. It’s not hard to see why this kind of thing is essential for testing autonomous vehicles.
Nvidia GDN, the portal to the omniverse
Jensen talks about the three types of hardware that power Omniverse — RTX computers for creatives, OVX servers, and the Nvidia GDN (graphics delivery network). Jensen is contrasting the GDN to a CDN (content delivery network), speaking about how it empowers graphics to be sent all around the world without the need for local computing. Omniverse Cloud is the suite of applications, or an “infrastructure as a service,” as Jensen calls. Omniverse Cloud was announced earlier this year, but Jensen says they’re available on AWS today.
Using digital twins for manufacturing and robotics
Jensen has introduced a few different partners, including Lowe’s and Amazon Robotics, to show how the use of digital twins built on the Omniverse will continue to be important in the future. He’s made the point that no physical products that runs on software can be released to the public without having that product be fully tested on its digital twin in 3D simulation.
Where are the specs?
Jensen has already moved on to Omniverse, but I’m sure most PC gamers are skeptical of how quickly Jensen moved over the three new GPUs that were announced. The above slide is the only technical specs offered for the RTX 3090 Ti during the presentation, but the full specs have been posted on Nvidia’s website: https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4090/
The metaverse, the “3D internet”
Jensen has moved on to Omniverse and the metaverse, which he calls “the 3D internet.” Everyone seems to have a different definition of what the metaverse really is, but I can see why Nvidia wants to spin it that way.
Jensen’s talking about updates to Omniverse and the VFX pipeline. “Omniverse connects 3D worlds,” he says.
The RTX 4090 and 4080 is official!
Jensen has shown some very minor specs for these first new Ada Lovelace chips. For rasterized games, Jensen says it’s up to 2x faster than the previous generation. For ray traced games, it’s over 4x faster. Jensen is talking about overclocking Ada Lovelace GPUs, as well as the power efficiency, though he hasn’t gotten into the details.
He says the RTX 4090 is 2x faster in Microsoft Flight Simulator and 3x faster in Portal RTX than the RTX 3090 Ti.
The RTX 4090 is available on October 12 at $1,599. The RTX 4080 will be $1,199 for the 16GB model and $900 for the 12GB model.
Cyberpunk 2077 and Microsoft Flight Simulator demoed with DLSS 3
Nvidia shows a demo of Microsoft Flight Simulator, a traditionally CPU-limited game, boosting frame rates in the scene from 66 fps to 135 fps.
Nvidia has also announced Portal RTX, reimagined now with DLSS 3 and ray tracing.
DLSS 3 officially announced
Jensen is talking about 3rd-gen RTX, including DLSS 3. The difference in this latest generation? “A revolution in neural rendering,” as it says. DLSS 3 predicts entirely new frames, not just pixels. Jensen says it’s four times faster than traditional performance. He even says it benefits CPU-limited games this time around.
Nvidia announces Ada Lovelace
“Today, we’re announcing Ada Lovelace,” Jensen proudly says. This is the “one GPU” that was powering the toy racing car scene we just saw. Nvidia says it’s worked close with TSMC for Ada Lovelace, designing the 4N process made for GPUs.
This prerecorded presentation just doesn’t have the quirky weirdness that use to come with live Nvidia events in the past. But this is the world we live in now, unfortunately.
Jensen takes the virtual stage
Jensen has donned his iconic leather jacket and beamed into a virtual space to start off GTC. He’s giving an overview of the keynote, talking about how it will touch on both new RTX products. He’s already moved on to Omniverse Cloud, showing how much of the focus of today’s show may be away from PC hardware.
“Future games will be simulations,” he says, showing off a impressive, lifelike 3D scene involving some toy cars racing around a bedroom.
The presentation kicks off in 10 minutes
Nvidia’s livestream is started up. Here’s the description from its YouTube stream:
“PC enthusiasts, don’t miss the GeForce Beyond special broadcast. Learn about the latest breakthroughs in gaming, creating, and graphics technology — from NVIDIA founder and CEO Jensen Huang in the opening of his GTC keynote on Sept 20 at 8 a.m. Pacific time (UTC-7).”
The big question remains: How many GPUs from the RTX 40-series will Nvidia actually announce? We’ll have to wait and see.