"Never gonna give you up" - Nvidia Says It Is Still Developing a Full Spectrum of Arm CPUs

[Lightwreather]

Summary

Although Nvidia had to drop its plans to buy Arm, Nvidia CEO Jensen Huang revealed during the company's earnings call tonight that it still has big plans to create a full portfolio of Arm-based CPUs for a wide range of applications spanning from the data center to robotics.

The company says that its inaugural Grace CPU is on-track to be released in the first half of 2023, but its release marks just the beginning of Nvidia's long CPU journey. Huang says the company has a 20-year license for Arm's architecture, so it will use the instruction set to develop CPUs for a wide range of applications spanning from tiny SoCs for robotics to high-end processors for supercomputers.

 

Quotes

Quote

"We are on track to launch our Arm-based Grace CPU targeting giant AI and HPC workloads in the first half of next year," said Jensen Huang, chief executive of Nvidia, at the company's earnings call with financial analysts and investors. "Our 20-year architectural license for Arm's IP allows us the full breadth and flexibility of options across technologies and markets."

"You are going to see a lot of exciting CPUs coming from us, and Grace is just the first example. You're going to see a whole bunch of them beyond that," said Huang. "We love to see the expansion of CPU footprints, and we are just thrilled that Arm is now growing into robotics, autonomous vehicles, cloud computing, supercomputing. We intend to bring the full spectrum of Nvidia's accelerated computing platform to Nvidia Arm CPUs."

"We have multiple Arm projects going in the company from connected form devices to robotics processors, such as the new Orin [SoC] that is going into autonomous vehicles, industrial automation, and so on," said Huang. "You could expect us to do a lot of CPU development around the Arm architecture. […] One of the things that developed nicely over the last couple of years is the success that Arm has seen in hyperscalers and data centers. It really motivated them to accelerate the development of higher-end CPUs."

"Our strategy is accelerated computing; that is ultimately what we do for a living," stated the head of Nvidia. "We will deliver on our three-chip strategy across CPUs, GPUs, and DPUs. "Whether x86 or Arm, we will use the best CPU for the job. Together with partners in the computer industry, we will offer the world's best computing platform to tackle the most impactful challenges of our time."

 

My thoughts

So, apparently Nvidia will not give up on ARM just yet, even though they've given up on their aquisition of ARM. Jokes aside,  this is fine by me, good even. I'm all for more competition in the tech space, and with NVidia joining in the fray, we just might see all the Big three (R, G and B) competiting in the CPU space and GPU space. Well, we still know almost next to nothing about nvidia's CPUs (apart from Grace), so we'll have to wait and see (we might be waiting till beyond 2023)

 

Sources

Tom's Hardware

Earning call transcript - The Motley Fool

[Stahlmann]

5 minutes ago, J-from-Nucleon said:

wide range of applications spanning from the data center to robotics.

So basically it doesn't matter for us? (yet)

 

We'll have to see what they have planned for the consumer market - if anything at all.

[Needfuldoer]

Why not! They can still license ARM just like everyone else.

[Athan Immortal]

Not giving up? They own a perpetual license for the ARM processors anyway? I don't see what would have changed.

3 minutes ago, Needfuldoer said:

Why not! They can still license ARM just like everyone else.

Exactly!

[Lightwreather]

6 minutes ago, Athan Immortal said:

Not giving up? They own a perpetual license for the ARM processors anyway? I don't see what would have changed.

So umm, that's the joke?

 

[Needfuldoer]

5 minutes ago, J-from-Nucleon said:

So umm, that's the joke?

 

We let you down.

[Laborant]

On a side note: Nvidia already sells products with their own ARM CPU core designs.

 

They started with a variant of the Tegra K1 (Denver Dual core) in 2014. For the Tegra X1 (As present in the Nintendo Switch) they opted for ARM original cores (A53 and A57) but for the Tegra X2 (or TX2) they use a combination of ARM originals alongside of two Denver2-cores.

For their current Lineup of Automotive and Embedded Systems (Xavier Lineup), they use their own cores again, named Carmel.
 

The upcoming Orin lineup (succeeding the Xavier lineup) will use again ARM original cores (12x Cortex-A78AE), but for the following Atlas lineup, they announced that "Grace-next" CPUs will be used.

 

We use these chips for developing Embedded-AI and High Performance Embedded Tasks (e.g. image processing on the edge) in the form of their Jetson-Modules.

[porina]

Way I see it, this is the status quo so not unexpected. If they were to drop Arm there would have to be a replacement or they get out of certain business. RISC-V isn't even close, and I'm not sure anything else makes sense.

 

I still wonder what alternative universe would be like had the deal gone though.

[jasonvp]

1 hour ago, Laborant said:

On a side note: Nvidia already sells products with their own ARM CPU core designs.

 

The Mellanox SmartNICs also have ARM cores, though I'm uncertain if those are "NVidia's own" or someone else's design.

 

[RejZoR]

Would be nice if NVIDIA entered market of mobile SoC's and compete against Qualcomm and Mediatek. Their Tegra was pretty amazing, but wasn't very efficient because it was a separate module. If they made the whole thing themselves, well, that would be pretty sick. I certainly wouldn't mind a smartphone running NVIDIA SoC...

[Obioban]

It's almost as if they only wanted to buy ARM to limit competition, not because there was an actual need.

[Spindel]

Totally OT:

 

But I need to start a semiconductor company making CPUs for light weight applications. I'm going to call it Light Electronic Gadgets or LEG.