Nvidia CEO: We're open to Intel making our chips

Matching TSMC is not for the faint of heart, though, Huang warns


GTC Nvidia is considering expanding its supplier base by getting at least some of its chips made in Intel factories.

"They're interested in us using their foundries," said Nvidia CEO Jensen Huang regarding Intel, during a virtual press conference on Wednesday. "We're very interested in exploring it."

Demand for Nvidia's GPUs has outstripped supply as gamers, crypto-miners, and others scramble to get hold of the gear. This, and scalping, has forced up retail prices. The GPU shortage is so dire, buyers are paying premiums for older components, such as Nvidia's GeForce GTX 1080Ti, which can handle high-level gaming and graphics and is priced higher than when it was released in 2017.

This situation hasn't been a problem for Nvidia, which last month reported monster profits. That said, the chip designer is trying to increase its output capacity by growing its roster of suppliers and adjusting its product mix so that it can ship a range of components fabricated on older and newer process nodes.

And while Nvidia is trying to book as much factory capacity it can, Intel is spending billions on advanced factories in the US and Europe as it relaunches its foundry business, TSMC-style. Supply, meet demand; demand, meet supply.

Strength in diversity

"We've built diversity in the number of process nodes that we use," Jensen said. "We qualified a lot more process nodes. We're in more fabs than ever. We've expanded our supply base four-fold in the past two years."

That lines up with Nvidia's plans to sell multiple generations of GPUs at the same time. As the COVID-19 coronavirus took hold, the biz saw success selling GPUs based on its recent Ampere microarchitecture alongside the previous generation Turing architecture.

"It's given us the opportunity ... to continue to sell both the current generation as well as the Turing generation," said Nvidia CFO Colette Kress said earlier this month at a Morgan Stanley event. "We've been doing that to provide more supply to our gamers. We may see something like that continue in the future. It was successful with Ampere and we'll see as we move forward."

Nvidia unified its consumer and enterprise GPUs under the Ampere microarchitecture, though it could separate them again. At its GTC shindig this week, it introduced the datacenter-grade TSMC-made 4nm H100 GPU, the first silicon based on its latest Hopper microarchitecture, and is supposedly developing a separate Lovelace architecture for consumer GPUs.

Nvidia hasn't offered a release date for its next-generation consumer GPUs. Kress said the hunger for Ampere chips hasn't yet been sated, and a transition to next-generation gaming processors also depends on advances in such things as real-time ray tracing, which is maturing rapidly and has made its way to hundreds of gaming titles.

"Ray tracing being in its second generation with Ampere has really expanded the capabilities of how you build out games," Kress said. "Three years ago when we started out with ray tracing, it was a chicken and the egg kind of challenge – what comes first, the infrastructure or the games? Now they're both here."

Old pals

Interestingly, Huang has a strong working relationship with Intel CEO Pat Gelsinger, who previously led VMware. VMware is an important cog in Nvidia's datacenter strategy. During the press conference, Huang said Nvidia sees Intel as a partner, adding that Intel's strategy under Gelsinger to focus on manufacturing is the right way to go.

"I'm encouraged by the work that is done at Intel. I think this is a direction they have to go and and we're interested in looking at the process technology," Huang said. We're told details need to be ironed out, though.

"We have to align technology. The business models have to be aligned. The capacity has to be aligned. The operations process and the nature of the two companies have to be aligned. It takes a fair amount of time and a lot of deep discussion," he said.

Being a foundry the caliber of TSMC is not for the faint of heart. This is a change not just in process technology and investment of capital. It's a change in culture

(By the way, we can't help but recall that Intel not only designs its own graphics processors built into its x86 chipsets, it's also working on blurring the line between integrated and discrete graphics. Manufacturing discrete Nvidia GPUs would be quite a thing to see from Intel.)

Nvidia has built strong foundry partnerships with TSMC and Samsung, which provide design and validation expertise to see the chip blueprints through manufacturing and into shipments. Partnerships that Intel would have to elbow itself in between.

Like Apple, Nvidia has worked closely with TSMC. The pair have figured out how to bundle together a collection of complex technologies to form the H100 GPU. The 80-billion transistor chip uses TSMC's 4N process and chip-on-wafer-on-substrate packaging to tightly bind HBM3 RAM and math-crunching cores.

"Being a foundry the caliber of TSMC is not for the faint of heart," said Huang, referring to Intel. "I mean this is a change not just in process technology and investment of capital. It's a change in culture."

TSMC has the experience of working closely with hundreds of companies worldwide on chip design, supply chain logistics, and other manufacturing needs, Huang said. By comparison, Intel in the past made chips for itself, and will now have to replicate TSMC's operations when working with other companies.

"The ability to dance with all of these different operations teams, supply chain teams, is not for the faint of heart," Huang reiterated.

A manufacturing partnership means Nvidia will have to adapt its GPU design to Intel's advanced process technology that includes Foveros, which is for packaging, and EMIB, a high-speed communication link for processing blocks inside a package.

Small is beautiful

Nvidia, which has a preference for large monolithic dies, also shed further light on its production strategy at GTC.

It appears Nvidia supports the use of chiplets. These are dies, or tiles as Intel calls them, that contain CPU cores, AI engines, other forms of hardware acceleration, and other blocks of IP, and a single chip package can contain more than two or more chiplets. AMD takes the chiplet approach in its Zen family of microprocessors. Intel is moving its manufacturing facilities toward chiplets as it tries to welcome more businesses to use its manufacturing facilities.

Specifically, Nvidia's CEO suggested a customer could design their own chiplet with application or workload specific acceleration, and have the die manufactured and placed alongside one or more Nvidia GPU dies in a single package.

"Do I believe in chiplets in the future? There will be little tiny things that you can connect directly into our chip. As a result, a customer could do a semi-custom chip with just a tiny engineering effort and connect it into ours and differentiate it in their own datacenter in their own special way," Huang said.

Nvidia also welcomed, to a degree, the Universal Chiplet Interconnect Express (UCIe), which is a common-language interconnect for a wide array of GPUs, CPUs, AI engines, and other accelerators to link up inside a chip package.

UCIe was revealed earlier this month, and its launch supporters included Intel, TSMC, AMD, and Arm. Nvidia and Apple weren't listed as supporters at the time. Nvidia's CEO isn't ruling anything out.

"As soon as the UCIe spec is stabilized, we'll put it into our chips as fast as we can," Huang said, adding that "it will take, as it did with PCI Express, about half a decade or so." His approval of UCIe opens a pathway for UCIe-supporting Nvidia chiplets to make their way into Intel-manufactured packages on advanced nodes.

"UCIe has the benefit of allowing us to connect many things to our chips and now allow us to connect our chips to many things," Nvidia's CEO said. ®

Broader topics


Other stories you might like

  • Despite 'key' partnership with AWS, Meta taps up Microsoft Azure for AI work
    Someone got Zuck'd over

    Meta’s AI business unit set up shop in Microsoft Azure this week and announced a strategic partnership it says will advance PyTorch development on the public cloud.

    The deal [PDF] will see Mark Zuckerberg’s umbrella company deploy machine-learning workloads on thousands of Nvidia GPUs running in Azure. While a win for Microsoft, the partnership calls in to question just how strong Meta’s commitment to Amazon Web Services (AWS) really is.

    Back in those long-gone days of December, Meta named AWS as its “key long-term strategic cloud provider." As part of that, Meta promised that if it bought any companies that used AWS, it would continue to support their use of Amazon's cloud, rather than force them off into its own private datacenters. The pact also included a vow to expand Meta’s consumption of Amazon’s cloud-based compute, storage, database, and security services.

    Continue reading
  • Atos pushes out HPC cloud services based on Nimbix tech
    Moore's Law got you down? Throw everything at the problem! Quantum, AI, cloud...

    IT services biz Atos has introduced a suite of cloud-based high-performance computing (HPC) services, based around technology gained from its purchase of cloud provider Nimbix last year.

    The Nimbix Supercomputing Suite is described by Atos as a set of flexible and secure HPC solutions available as a service. It includes access to HPC, AI, and quantum computing resources, according to the services company.

    In addition to the existing Nimbix HPC products, the updated portfolio includes a new federated supercomputing-as-a-service platform and a dedicated bare-metal service based on Atos BullSequana supercomputer hardware.

    Continue reading
  • In record year for vulnerabilities, Microsoft actually had fewer
    Occasional gaping hole and overprivileged users still blight the Beast of Redmond

    Despite a record number of publicly disclosed security flaws in 2021, Microsoft managed to improve its stats, according to research from BeyondTrust.

    Figures from the National Vulnerability Database (NVD) of the US National Institute of Standards and Technology (NIST) show last year broke all records for security vulnerabilities. By December, according to pentester Redscan, 18,439 were recorded. That's an average of more than 50 flaws a day.

    However just 1,212 vulnerabilities were reported in Microsoft products last year, said BeyondTrust, a 5 percent drop on the previous year. In addition, critical vulnerabilities in the software (those with a CVSS score of 9 or more) plunged 47 percent, with the drop in Windows Server specifically down 50 percent. There was bad news for Internet Explorer and Edge vulnerabilities, though: they were up 280 percent on the prior year, with 349 flaws spotted in 2021.

    Continue reading
  • ServiceNow takes aim at procurement pain points
    Purchasing teams are a bit like help desks – always being asked to answer dumb or inappropriate questions

    ServiceNow's efforts to expand into more industries will soon include a Procurement Service Management product.

    This is not a dedicated application – ServiceNow has occasionally flirted with templates for its platform that come very close to being apps. Instead it stays close to the company's core of providing workflows that put the right jobs in the right hands, and make sure they get done. In this case, it will do so by tickling ERP and dedicated procurement applications, using tech ServiceNow acquired along with a company called Gekkobrain in 2021.

    The company believes it can play to its strengths with procurements via a single, centralized buying team.

    Continue reading
  • HPE, Cerebras build AI supercomputer for scientific research
    Wafer madness hits the LRZ in HPE Superdome supercomputer wrapper

    HPE and Cerebras Systems have built a new AI supercomputer in Munich, Germany, pairing a HPE Superdome Flex with the AI accelerator technology from Cerebras for use by the scientific and engineering community.

    The new system, created for the Leibniz Supercomputing Center (LRZ) in Munich, is being deployed to meet the current and expected future compute needs of researchers, including larger deep learning neural network models and the emergence of multi-modal problems that involve multiple data types such as images and speech, according to Laura Schulz, LRZ's head of Strategic Developments and Partnerships.

    "We're seeing an increase in large data volumes coming at us that need more and more processing, and models that are taking months to train, we want to be able to speed that up," Schulz said.

    Continue reading
  • We have bigger targets than beating Oracle, say open source DB pioneers
    Advocates for MySQL and PostgreSQL see broader future for movement they helped create

    MySQL pioneer Peter Zaitsev, an early employee of MySQL AB under the original open source database author Michael "Monty" Widenius, once found it easy to identify the enemy.

    "In the early days of MySQL AB, we were there to get Oracle's ass. Our CEO Mårten Mickos was always telling us how we were going to get out there and replace all those Oracle database installations," Zaitsev told The Register.

    Speaking at Percona Live, the open source database event hosted by the services company Zaitsev founded in 2006 and runs as chief exec, he said that situation had changed since Oracle ended up owning MySQL in 2010. This was as a consequence of its acquisition that year of Sun Microsystems, which had bought MySQL AB just two years earlier.

    Continue reading
  • Beijing needs the ability to 'destroy' Starlink, say Chinese researchers
    Paper authors warn Elon Musk's 2,400 machines could be used offensively

    An egghead at the Beijing Institute of Tracking and Telecommunications, writing in a peer-reviewed domestic journal, has advocated for Chinese military capability to take out Starlink satellites on the grounds of national security.

    According to the South China Morning Post, lead author Ren Yuanzhen and colleagues advocated in Modern Defence Technology not only for China to develop anti-satellite capabilities, but also to have a surveillance system that could monitor and track all satellites in Starlink's constellation.

    "A combination of soft and hard kill methods should be adopted to make some Starlink satellites lose their functions and destroy the constellation's operating system," the Chinese boffins reportedly said, estimating that data transmission speeds of stealth fighter jets and US military drones could increase by a factor of 100 through a Musk machine connection.

    Continue reading
  • How to explain what an API is – and why they matter
    Some of us have used them for decades, some are seeing them for the first time on marketing slides

    Systems Approach Explaining what an API is can be surprisingly difficult.

    It's striking to remember that they have been around for about as long as we've had programming languages, and that while the "API economy" might be a relatively recent term, APIs have been enabling innovation for decades. But how to best describe them to someone for whom application programming interfaces mean little or nothing?

    I like this short video from Martin Casado, embedded below, which starts with the analogy of building cars. In the very early days, car manufacturers were vertically integrated businesses, essentially starting from iron ore and coal to make steel all the way through to producing the parts and then the assembled vehicle. As the business matured and grew in size, car manufacturers were able to buy components built by others, and entire companies could be created around supplying just a single component, such as a spring.

    Continue reading

Biting the hand that feeds IT © 1998–2022