Cristiano Amon Takes over as Qualcomm CEO: Reiterates Focus on Custom CPUsby Andrei Frumusanu on July 2, 2021 10:00 AM EST
- Posted in
As of June 30th, Qualcomm’s Cristiano Amon has taken over as the company’s CEO, replacing his predecessor Steve Mollenkopf, who has now retired. Prior to the appointment, Amon had a long history and tenure at Qualcomm filling engineering roles, and previously filling the role of president of QCT (The company’s semiconductor business).
In statements to Reuters, Amon had made comments regarding the company’s future CPU roadmap, which come to further contextualise the company’s completed acquisition of NUVIA last March.
"We needed to have the leading performance for a battery-powered device," Amon said. "If Arm, which we've had a relationship with for years, eventually develops a CPU that's better than what we can build ourselves, then we always have the option to license from Arm."
The wording here is again very bullish on Qualcomm’s part, reinforcing the idea that the company is extremely confident in NUVIA’s CPU microarchitecture and that it will have no issue in differentiating itself in terms of performance compared to what Arm has available in terms of CPU IP. Last March, the company had noted that work on integrating NUVIA’s custom CPU core into a laptop-oriented Snapdragon SoC would be an immediate focus, with Amon now stating that they are planning on bringing such a design to market in 2022.
In terms of timeline and against which Arm core the NUVIA design might compete against depends on when exactly in 2022 the new chip might make it to market – if it’s in the first half, then we’ll see it compete against the already announced Cortex-X2 cores from Arm. If it’s in the latter half, it’s possible it will be positioned against Arm’s next-gen Sophia cores. In either case, Qualcomm seems confident in terms of beating the Arm Cortex designs, which bodes well for next-gen Snapdragons.
Amon’s comment that if Arm is able to build a better CPU than Qualcomm’s own designs is also reminiscent of the company’s previous generation custom CPU endeavours: the last time the company had employed a custom microarchitecture was in the 2016 Snapdragon 820 with its Kryo cores. Competing Cortex cores had been faster and more power efficient in a smaller area footprint, which lead the company to use those designs instead, and eventually leading to Qualcomm dissolving its CPU design teams – a decision which later ended up with no in-house design capabilities up until the recent NUVIA purchase.
Post Your CommentPlease log in or sign up to comment.
View All Comments
mode_13h - Friday, July 2, 2021 - linkYou should distinguish between what he says and what he really thinks/does. Publicly, they're always going to put a positive spin on whatever happened. They have to, unless it's so bad (like Intel's 10 nm delays) that there no way to hide what a debacle it is.
ikjadoon - Friday, July 2, 2021 - linkHonest leadership can be positive, too. Look at Dr. Lisa Su when they launched Zen. Look at Arm when they launch new cores. Do any of them sulk, “Yeah, this is good enough. Don’t expect we need to much faster for Windows / Android” ?
Amon comes from a long line of arrogance at Qualcomm. See their Court battles, mass royalties, monopolization, etc.
Stop lying to your customers and investors: turns out, this option has no negative consequences.
Raqia - Friday, July 2, 2021 - linkThey devoted CPU design resources to Centriq rather than mobile which was canned during the Broadcom hostile takeover bid for Qualcomm to appease shareholder demand for lower R&D costs. The part was very good in terms of efficiency:
but those gains at the SoC level would get diluted by other components and by accelerator card power budgets (which is why a pivot to AI100 PCI-e cards which go dozens into a rack and better multiplies their expertise in power efficiency makes sense.)
So it was probably the least promising of their initiatives in terms of long term value at that time since even though margins were incredibly high, competition from AMD in the x86 space with sticky software stacks would quickly erode margins and hyperscalers were starting to do their own in-house SoC designs based on ARM IP.
Nuvia started their business specifically in server CPU parts to attract venture capital because Intel had such artificially high margins and with typical SoCs sporting GPUs, ISP/DSP, modems etc. they could not realistically do anything else with their specific team expertise, so I honestly don't think they much care that it's servers in particular, just something that could leverage their bread and butter CPU core IP design prowess. However, with the trouble companies are having booking leading edge foundry capacity, hyperscalers doing in-house SoC designs and newer servers integrating ever more disciplines outside of the CPU core IP itself, they would find it hard to distinguish themselves if they had to do a part at 10nm, without an integrated FPGA, on package cache tiles, or specialty high bandwidth interconnects. A sale to someone like Qualcomm seems win-win given this.
mode_13h - Friday, July 2, 2021 - linkSo, you think he's all hype? Or was his point that it was pre-release and its UI issues are probably fixable via software? Or maybe he was more interested in just being in the game and on a *trajectory* for competing with iPhone?
To some extent, a CEO does need to deliver the hype. However, what counts is that they have a good management team behind them, that can deliver the goods. I've heard a CEO is more an external-facing role. They have a lot of influence on the company, but the most important thing they do is present a good face and message to customers and investors. And history has shown that investor relations is not something to be taken lightly, at Qualcomm.
sonicmerlin - Monday, July 5, 2021 - linkAnd it still lags to this day. They just do a better job of hiding it behind visual flair.
mode_13h - Friday, July 2, 2021 - linkI just wonder what innovations Nuvia can possibly have that make them so much better than everyone else. I know it's staffed by some ex-Apple engineers, but they supposedly left that company with some idea(s) that even it isn't doing.
I see two patents from Nuvia:
US20210151549A1: Integrated System with Power Management Integrated Circuit Having On-Chip Thin Film Inductors
US10911048B1: Dynamically adjustable CMOS circuit
Of these, the second sounds much more intriguing. I'm not too sure what to make of it, though. I wonder if their whole reason for founding the company could be based on such a low-level tweak, or if it's just one of many tricks they had stored up.
ikjadoon - Friday, July 2, 2021 - linkIIRC, they left Apple to build server cores.
I think the latest Jim Keller interview noted if you start with an old / bad design and try to iterate (aka “scale” in marketing), you’ll hit constraints sooner and can’t make large enough leaps per generation.
He definitely seems to be a clean sheet proponent and I think that’s what NUVIA is trying as well. Not new as in “unimaginably new architectures”, but new as in “let’s imagine we can start from scratch and build it for high-perf.”
So, the assumption is that Arm Ltd. is very much simply iterating its cores and not starting fresh enough often. A76, A77, A78, etc.
And, if you know you’re designing for high-performance (via PPA), then maybe your fresh architecture begins differently. IIRC, X1 is just a larger + wider A78. It’s not a new design.
> At a high level, the design could be summed up as being a ultra-charged A78 – maintaining the same functional principles, but increasing the structures of the core significantly in order to maximize performance.
ikjadoon - Friday, July 2, 2021 - linkThis brings up a great point: where are NUVIA’s small cores? Are they pushing a big-core-only laptop?
Most, but not all, in the Arm space advocate for big.LITTLE approaches to save battery life / improve efficiency. Like, I already have the experience of a 50W PL2 on my lap. I don’t like it.
Or back to Arm Ltd. for efficiency? IIRC, Phoenix (NUVIA’s core) is Armv8, so the Arm v9 a510 isn’t a drop-in little core. Back to A55? I don’t even know how Windows uses A55 cores. Are they useful enough?
Would be hilarious if Intel laptops are big.LITTLE in 2023 and Qualcomm releases a big-only Arm CPU.
And the final question: which OEMs will use the NUVIA/QC SoC? Microsoft, Samsung, and HP seem to be the biggest WoA proponents. And yes, current Samsung WOA devices use QC SoCs!
mode_13h - Friday, July 2, 2021 - link> I think the latest Jim Keller interview noted if you start with an old / bad design and try to
> iterate, you’ll hit constraints sooner and can’t make large enough leaps per generation.
I thought that was more in reference to ISA, where you simply can't shake the legacy stuff. Sure, it can apply to other cases, but Apple or ARM can do a new design at least as easily as Nuvia.
> Arm Ltd. is very much simply iterating its cores and not starting fresh enough often.
But isn't the core after A710 supposed to be a new design?
>> At a high level, the design could be summed up as being a ultra-charged A78
You're talking about the X1? Bad example, IMO.
lmcd - Friday, July 2, 2021 - linkThe A76 is considered extremely close to clean-sheet, which invalidates your entire post, as it is still an extremely recent launch.
And no, there are plenty of cases where feeding the execution units better might as well be a fresh design. If we're just gonna cite Keller like it's gospel, he discusses how the execution units are just a functional block, and not where the innovation happens anymore.