Nvidia GeForce Beyond: Live updates for RTX4090 launch

Jensen will be speaking about Nvidia’s new large language models. This includes a model that is specifically tailored for biology and chemical. Nvidia will collaborate with the Broad Institute to provide new libraries as well as acceleration for scientific research.

Hopper is the architecture behind these powerful GPUs and Hopper is necessary to speed up large language models. Jensen believes large language models will become the most important AI in our time. Hopper is its engine. Jensen made a hilarious Star Trek reference. This is helping to bring back the dork factor needed for this ultra-series presentation.

Jensen announced Nvidia Qoda which is a hybrid quantum-classical computer. This is amazing. Jensen speaks of bringing quantum computing’s future into the present. Qoda is able to simulate the capabilities of quantum computers long before they actually arrive.

Jensen is now speaking to industry partners and investors, rather than PC gamers. Nvidia has not done a good job of understanding how to market this stuff. You have to wonder why Nvidia doesn’t host a dedicated event on PC gaming.

We aren’t interested in Nvidia’s Omniverse or cloud computing projects, but the company doesn’t do a good job communicating this to the average person. Jensen enjoys using acronyms and numbers. But, come on! Let’s see the bigger picture. This should be exciting for anyone interested in tech. However, it is easy to get lost in the weeds.

Jensen is now focusing on robotics computers, and has announced the Jetson Orin Nano. He claims it is 80x faster than the predecessor.

He has also announced a partnership to Siemens, which uses Orin IGX technology and Nvidia’s Metropolis platform. Now, he’s focusing his efforts on medical use cases of technologies, such as for robotic-assisted surgical procedures. Jensen has many examples of Nvidia Isaac, the company’s industrial robots platform. Jensen even described it as a “pre-Quantum” computer. Nvidia has invested a lot in these robotics platforms. All of it is available in the Omniverse cloud, of course.

Nvidia displays a demonstration of Omniverse-designed cars, complete with digital dashboards and infotainment, by bringing together the tools.

Next, you will see a real-looking demonstration of someone driving in self-driving cars, assisted by Drive Concierge, a digital assistant that plays emotional music.
Drive Orin, the second-gen processor designed for self-driving cars is Jensen’s favorite. Jensen claims that 40 cars have already selected it.

Jensen will soon be driving self-driving cars. They’ve replaced Drive Atlan, an automotive car that never came out, which is quite interesting. He says that Ada Lovelace gave them access to a lot more performance so Atlan is gone.

Jensen imagines a single computer running a car. Currently, multiple computers are managing the car. All of these computers run software on Drive Thor. Jensen connects what Nvidia did with Drive Sim to test self-driving vehicles. Omniverse is able to create 3D scenes from real sensor data using AI and neural network. This is why it’s so important to test autonomous vehicles.

Jensen speaks about the three types o hardware that power Omniverse — RTX computer for creatives, OVX server, and Nvidia GDN (graphics distribution network) Jensen compares the GDN to a CDN (content distribution network), and explains how it allows graphics to be sent around the globe without any local computing. The Omniverse Cloud suite of applications is also known as an “infrastructure-as-a service” or the “infrastructure for a cloud”. Jensen said that Omniverse Cloud was first announced in the beginning of this year. However, they are now available on AWS.

Jensen introduced several partners including Lowe’s as well as Amazon Robotics to demonstrate how digital twins built on Omniverse will be useful in the future. He made it clear that physical products running on software cannot be released to the general public without being fully tested in 3D simulation on the digital twin.

Jensen has moved on to Omniverse. But most PC gamers will be skeptical about how fast Jensen transferred the three new GPUs. 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/

Jensen now moves on to Omniverse. He also calls the metaverse “the 3D internet.” Although everyone seems to have a different idea of what the metaverse is, I can see why Nvidia would like to make it that way.

Jensen is talking about Omniverse updates and the VFX pipeline. He says, “Omniverse links 3D worlds.”

Jensen has provided some minor specifications for the Ada Lovelace Chips. Jensen claims it is up to 2x faster for rasterized games than the previous generation. It is over 4x faster when ray traced. Jensen hasn’t got into the details, but he is talking about overclocking Ada Lovelace GPUs.

He said that the RTX 4090 is faster in Microsoft Flight Simulator than the RTX 3090 Ti and Portal RTX is 3x faster.

The RTX 4090 will be available October 12th at $1,599. The RTX 4080 is $1,199 for the 16GB model, and $900 for 12GB.

Nvidia displays Microsoft Flight Simulator demo, which is a CPU-limited game. It boosts frame rates from 66 to 135 FPS.

Portal RTX was also announced by Nvidia. It is a reimagined portal with DLSS 3 (ray tracing) and DLSS 3.

Jensen is referring to 3rd-gen RTX including DLSS3. What is the difference between this and previous generations? According to the article, “A revolution is neural rendering.” DLSS 3 can predict entirely new frames and not just pixels. Jensen claims that it is four times faster then traditional performance. It even benefits CPU-limited gaming, he claims.

Jensen proudly declares that “Today, We’re announcing Ada Lovelace”. This was the “one GPU” behind the toy racing cars scene that we just witnessed. Nvidia claims it has worked closely with TSMC on Ada Lovelace’s design of the GPU-friendly 4N process.

Prerecorded presentations don’t have the same quirky weirdness as live Nvidia events. Unfortunately, this is the current world.

Jensen has donned the iconic leather jacket and beaming into a virtual space in order to launch GTC. He will give a brief overview of the keynote and talk about how it will affect both new RTX products. He has already moved to Omniverse Cloud, demonstrating how far the show’s focus may be from PC hardware.

He says that future games will be simulations. He shows off an impressive 3D scene featuring some toy cars racing through a bedroom.

Nvidia’s stream is now live. Here’s the description taken from their YouTube stream.
“PC enthusiasts, don’t miss the GeForce Beyond special broadcast. Jensen Huang, NVIDIA’s founder and CEO, will be opening his GTC keynote at 8 a.m. Pacific (UTC-7), September 20th.
The big question is: How many GPUs will Nvidia actually reveal from the RTX 40 series? We will have to wait and watch.