Fall GTC Shows Who Really Cares About Nvidia • The Register

CGC This week’s GPU Technology Conference saw Nvidia do something we haven’t seen much from the chip designer lately: update a consumer product.

For the increasingly enterprise-obsessed tech giant, GTC has become less and less focused on GPUs for gamers and everything to do with capitalizing on new and emerging markets, such as AI, robotics, autonomous vehicles and the always verbose metaverse. By metaverse, in this context, we mean 3D virtual reality worlds in which it is possible to interact and collaborate with simulations, applications and with each other.

Nvidia CEO Jensen Huang, dressed in his signature leather jacket, took the stage – or is it the holodeck? we’re not sure, to unveil a trio of RTX 40 series graphics cards based on the Ada Lovelace architecture of its engineers.

For many who tune into Huang’s nearly hour and 45-minute keynote, that revelation may have been the only solid and relevant announcement at this fall’s event.

Using a number of select benchmarks, Huang bragged about the performance improvements of the RTX 4090 and 4080 graphics cards over their predecessors. The chip designer said the RTX 4090 will deliver 2x-4x higher performance than the company’s previous flagship 3090 TI launched this spring.

Then there is the price of these new RTX units. The cards are some of Nvidia’s most expensive to date. At $ 899 for the 12GB 4080 and $ 1,199 for the 16GB version, the cards cost $ 200- $ 500 more than the 3080 when it launched two years earlier. The weakening of the price on the 4090 is not that serious. At $ 1,599, that’s about $ 100 more than when the 3090 debuted in 2020.

Huang, speaking at a press conference on Wednesday, defended the increase, arguing that the gains in performance and feature set more than offset the higher price. He said the price increase was further justified by higher manufacturing and material costs.

“A 12-inch wafer is a lot more expensive today than yesterday, and it’s not a little more expensive, it’s a lot more expensive,” he said, adding that “our performance with Ada Lovelace is enormously better.”

But beyond the new cards, which Huang spent less than two minutes detailing, he’s back as usual. Here is a summary of Nvidia’s most important announcements at the GTC.

Go back to a dual architecture model

The roughly 15 minutes leading up to the RTX announcement was spent on Nvidia’s new Ada Lovelace architecture, which sees the chip designer once again revert to a dual-architecture model.

Nvidia’s previously announced Hopper architecture will power the company’s HPC and AI-focused processors, such as the H100, while the Ada Lovelace architecture will power Nvidia’s graphics-centric chips.

Named after the 19th century mathematician, the Ada Lovelace architecture is based on TSMC’s 4N process and features Nv’s third-generation real-time ray tracing cores and fourth-generation Tensor cores.

So there’s the split: Hooper primarily aimed at high-performance computing and large AI workloads, and Lovelace primarily aimed at everything else, from cloud server GPUs to game cards.

It is certainly not the first time that Nvidia has used a dual architecture model. Going back two generations, Nvidia’s data center chips, such as the V100, used its Volta architecture. Meanwhile, its consumer and graphics-centric chips, such as the RTX 2000 series and Quadro RTX family, used Turing microarchitecture.

In addition to Nvidia’s RTX 40 series parts, Ada Lovelace also wants to power Nvidia’s RTX 6000 series workstation cards and its L40 datacenter GPUs. However, unlike Hopper, Huang says the new architecture is designed to address a new generation of graphics-centric challenges, including the rise of cloud gaming and the metaverse. Those will need graphics chips somewhere to render those environments real-time: cloud gaming is where the game is mainly rendered in a backend and streamed live over the internet to a screen in front of the user, such as a laptop. or a telephone. This saves players from buying and upgrading gaming platforms and / or carrying them around everywhere.

“In China, cloud gaming is going to be very big and the reason is because there are a billion phones that game developers no longer know how to serve,” he said. “The best way to solve it is cloud gaming. You can reach integrated graphics and you can reach mobile devices.”

The metaverse but as a service

However, Ada Lovelace is not limited to cloud gaming applications. Nvidia is positioning architecture as the workhorse of its first software-as-a-service offering, which he believes will allow customers to access its Omniverse hardware and software stack from the cloud.

Omniverse Cloud offers the remote computing and software resources needed to run metaverse applications on demand, from the cloud. The idea is that not all companies want or even have the budget to spend millions of dollars on one of Nvidia’s OVX SuperPods to provide that level of simulation and rendering on the off chance that the metaverse actually goes somewhere. Instead, they can build their own metaverses in the Omniverse Cloud.

For now, Nvidia appears to be wooing large logistics, manufacturing and industrial partners, promising to help them build and display digital twins. These twins are large-scale simulations – each simulation is twinned with the real world, using real data and models – and presented as a way to test and validate designs, processes and systems in a virtual world before they are implemented in the real world.

Yes, it’s all more imaginative modeling and simulation, but with new silicon, interactivity, virtual reality, and invoices.

While Omniverse Cloud is Nvidia’s first foray into managed cloud services, it won’t be the last, according to Huang, who reported that his company is considering a similar model for its other software platforms.

Smarter cars, robots

Nvidia doesn’t just want to power the digital twins of customers’ warehouses and manufacturing facilities. During the keynote, Huang also detailed a series of hardware designed to power anything from autonomous robots to cars.

Huang talked about Drive Thor, Nvidia’s all-in-one computing platform designed to replace the multiplicity of computer systems used in vehicles today.

The technology will make its debut in China, where Nvidia says it will power the Zeekr and Xpeng 2025 vehicle range and QCraft’s autonomous taxi service. That’s, of course, if U.S. export restrictions don’t tighten to the point where Nvidia can no longer provide, a perspective Huang downplayed during Wednesday’s press conference.

Meanwhile, to power the robotic minions lurking alongside the human workers, Nvidia showed off its IGX and Orin Nano platforms.

IGX is based on Nvidia’s previously announced Orin AGX Industrial system-on-modules, but adds high-speed networking. According to Nvidia, one of the first uses of the card will be in surgical robots. Meanwhile, Nvidia’s Jetson Orin Nano modules are designed to tackle less demanding applications.

Great language templates for the masses

As with previous GTCs, the software dominated a sizable chunk of the keynote. Two of the biggest releases for this fall’s event were Nvidia’s large-language model (LLM) services called NeMo and BioNeMo.

The services aim to make it easier for AI researchers and biologists to use LLMs looking to gather information from complex datasets. The services allow customers to insert existing data into customizable base templates with minimal effort. For example, it has been suggested that BioNeMo could be used to accelerate research into protein folding.

Every single company, in every single country, speaking every single language probably has dozens of different skills that their company could adapt our large language model to go and perform

However, looking beyond the medical field, Huang predicts LLMs will have broad applicability for the vast majority of businesses. “My feeling is that every single company, in every single country, that speaks every single language probably has dozens of different skills that their company could adapt to our big language model to go and perform,” she said.

“I’m not exactly sure how big this opportunity is, but it’s potentially one of the biggest software opportunities ever.”

Hopper in production

Finally, Nvidia has provided an update on the availability of the long-awaited Hopper H100 GPUs, which it claims have gone into series production and will begin shipping to OEM system builders next month.

Announced at Nvidia’s GTC spring event, the 700W GPUs promised AI performance that’s 6x faster than the outgoing A100 thanks to support for 8-bit floating point computing. Meanwhile, for HPC applications, Nvidia claims the chip will deliver 3x the performance in double-precision FP64 calculations.

However, those hoping to get their hands on Nvidia’s internal DGX H100 servers, complete with their own custom interconnect technology, will have to wait until the first quarter of 2023, a full quarter after forecast.

While Nvidia blamed the DGX system’s increased complexity, a likely culprit is the Intel Sapphire Rapids processors used in the systems, which were reportedly delayed until the end of the first quarter. ®