‘Not All NPUs Are Created Equal’: Qualcomm’s Nitin Kumar Talks ‘Disruptive’ Snapdragon X Elite on Copilot+ PCs

Microsoft introduced its new line of artificial intelligence (AI) PCs last month, announcing Copilot+ Windows laptops packing new silicon, that can match up to, and even surpass, the performance of a MacBook running on Apple’s in-house M series chips. The new thin and light Copilot+ laptops, from both Microsoft and its OEM partners Acer, ASUS, Dell, HP, Lenovo, and Samsung, promise all-day battery life, AI features like Recall, Live Captions, Co-creator and more.

A host of these Copilot+ laptops are powered by Qualcomm’s Arm-based Snapdragon X Elite platform, announced at the Snapdragon Summit in October 2023. The American chipmaker, a major player in the smartphone SoC market, now hopes to take on chip giants Intel and AMD, and Apple’s in-house silicon, to make its mark in the PC segment. Microsoft is reportedly confident that the Snapdragon X Elite-powered Windows laptop can finally beat Apple’s MacBook models. In fact, as per internal testing from Qualcomm, the X Elite is said to be 28 percent faster than Apple’s M3 chipset.

At the eve of Computex Taipei last week, Qualcomm president and CEO Cristiano Amon took the stage and highlighted the capabilities of the new line of Copilot+ PCs. “The PC is reborn,” Amon said in his keynote, calling X Elite-powered Copilot+ PCs “the fastest, most intelligent Windows PCs ever built.” At Computex, the new laptops running on Qualcomm silicon were on display, showcasing AI features like Live Captions and Recall. Gadgets 360 got a chance to interact with X Elite-powered Microsoft Surface laptops at a Qualcomm demo room and try out some of the features first-hand.

Microsoft Surface Copilot+ laptops at Qualcomm’s demo room at Computex Taipei

At the sidelines of the tech trade show in Taipei, we also caught up with Nitin Kumar, vice-president of product management for Snapdragon chipsets at Qualcomm. In our chat, we covered a lot of ground — from the company’s partnership with Microsoft, to new Copilot+ features on the X Elite chip. We also touched upon the expectations from the new platform in terms of gaming performance and the path ahead for AI-powered laptops. Here are slightly edited and condensed excerpts from our chat:

Gadgets 360: Can you talk about the journey that has led to the X Elite chip launch and Qualcomm coming back into the PC game to position itself against traditional heavyweights like Intel and AMD, making it more of a three-horse race?

Nitin Kumar: Absolutely. I’ll actually take a step back and walk you through how we have envisioned this, why we are in this business and what we are trying to achieve. Of course, Qualcomm has a long history and legacy in smartphones, and India is a very predominant market for us; Snapdragon is a very well-known brand name for India market for smartphones.

If you look at the last decade, Qualcomm, has invested predominantly in smartphones in multiple technologies. Every year with every generation, we have driven the technology forward, we’ve driven the experience forward. When you draw a parallel to that with the PC industry, to be very honest, it hasn’t seen any significant innovation. When you look at the same, let’s say, last ten years or beyond, be it in terms of improvement or user experience or feature set or even battery life for that matter, things haven’t improved significantly.

So, we have the technological know-how and we wanted to bring that level of innovation onto the PC industry to provide that disruption. So that when you innovate or when you do more tasks or whatever you want to do with your PC, you get that better capability. So, our journey has been from that vision, that we have our core strength in leading technology, absolutely leading battery life — that’s our DNA — and we want to leverage our core strength, our DNA onto the PC industry and along with partnering with Microsoft with full Windows experience natively available on Arm architecture-based Snapdragon processor — we’ve had two or three generations of product in the past — but this one is absolutely foundational, absolutely unique, absolutely disruptive with the level of technology that we have packed in the X series processors. We’re very proud of that. That’s what the journey has been and that’s our vision: to be disruptive in the marketplace where the PC industry hasn’t been.

Intel unveiled its Lunar Lake laptop chip at Computex Taipei

Gadgets 360: From the consumer perspective, it feels like, in the past year, the pace of innovation, especially on the chipset side, has been pretty quick. We’ve already seen multiple iterations of advanced AI capable chips coming out. If a consumer wants a laptop with AI-ready features and they look at options running on AMD chips, or Intel’s (upcoming) Lunar Lake chips, and now there’s one that are running on Snapdragon. How does the consumer decide what to get? PC consumers know that Intel and AMD have been in the PC market for long, and Snapdragon is now coming in (with X Elite). Reported benchmark numbers say that it’s as powerful or more powerful than the M3 chip from Apple. But how do you make the consumer make that switch in their heads?

Nitin Kumar: Great question. And we think about that a lot, and I completely understand the crowded aspect of the messaging and the crowded aspect of the marketplace. Let me first give you a few data points on how we are looking at it and what we are doing as well. You’re correct, our competition in the X86 world has had a significantly longer legacy in terms of PC space. So, we completely understand that, and they have a presence associated with that. But at the same time, if you look at what Snapdragon is known for, especially in India market — for premium Android experience, you want that mobile device experience, you want the best experience on an Android phone that you can get, you got to go Snapdragon.

So, there is a brand equity that we have with Snapdragon. And we are bringing that same brand promise, that we have worked hard over generations to create with our mobile phone, onto a PC, as well. So, when as a consumer you have the absolute leading-edge Android smartphone powered by Snapdragon 800 series, 8 Gen 2, 8 Gen 3 platform, what you get is premium experience, premium battery life. So, for the consumer, it should be a seamless transition to look at. ‘OK, I go with Snapdragon X Elite, I’m bound to get the premium experience and premium battery life.’ That gives us a big, big edge over there.

And even with everybody talking about AI capability, I think there needs to be a second-level filter on what that really means. And how do you decipher the two out. Not all NPUs are created equal, if you will. What truly matters? Of course, one is just peak performance: Okay, I have 40 TOPS (trillions of operations per second) NPU or 45 tops NPU. But what truly matters is to think of it from a metric, which is NPU performance per power consumption, or NPU performance per Watt. What all I can load up on my NPU and get that task done as quickly or burning the least amount of power.

That’s where the efficiency, the mobility aspect will come in, and that’s where the Snapdragon promise, which comes from the mobile space, will tie into PC and help us differentiate in the market. We are very confident in our approach, in the promise that we are delivering, and we have made our comparisons. We’re very confident that these devices will be the absolute best Windows 11 Copilot+ devices in the marketplace. And with the strength of our partners, ecosystem and partnership with Microsoft, I think we will disrupt the market.

The Snapdragon X Elite was announced at Snapdragon Summit in October last year

Gadgets 360: We’ve also heard from Qualcomm that 1,200 games have been tested on the X Elite chip, but performance numbers have not been made public yet. And I’m not going to get into that. But what can gamers expect from Qualcomm?

Nitin Kumar: Very good point. It’s definitely in our interest to set the right expectation in the marketplace on what these systems are capable of. In the initial phase, we have been focused predominantly on consumer and then commercial space. To be very honest, gaming as a category is not our initial focus and the initial set of devices (on Snapdragon silicon) are not going to compete with the gaming category of devices.

Having said that, there is a class of gamers that are casual gamers, not necessarily hard-core gamers. But when you look at casual gaming, the performance the system is driving, it’s actually meeting or beating any other integrated GPU class of product for a casual gaming experience that is available today. And more importantly, you are able to achieve that same level of graphics and gaming capability at half the power. So, think from a casual gamer perspective, this will drive a better experience, not from a performance perspective, but battery-life perspective, and overall, you get a better system.

Gadgets 360: We tried out some of the AI features on Snapdragon-powered Microsoft Surface Copilot+ laptops. Some of them weren’t as accurate when it came to localised prompts. But Live Captions worked well with Hindi. How do you plan to localise these AI features?

Nitin Kumar: Qualcomm has been integrating an AI engine on our mobile platforms for almost a decade now. So, the class of leading-edge processors that we’ve had for our smartphone lineup, they’ve all had an AI engine of some sorts, half a TOPS, one TOPS — doesn’t matter. But we have been innovating on AI for long. The PC industry has just started to talk about it. So, think of all the AI models, AI work, AI optimisation that we have worked on the mobile front, we’re bringing all that out onto our PC platform as well — all that has been optimised, localised, and regionalised for the different regions.

So live translation has a very popular use case on smartphone model. All those AI use cases have been optimised on Snapdragon architecture. The fundamental architecture on smartphone and PC is exactly the same, so you can take that model and run it on a PC, and it will have a phenomenal result because it’s already so finely, optimally tuned for Snapdragon architecture.

Now the question is, what do you do from a Qualcomm perspective to enable the developers to optimise these new models for the PC industry? The PCs may require a different set of models, different AI use cases, and how do you enable that? For the last 10 years, we’ve been working on this AI journey on smartphone that we’re bringing to PC, so we have a whole set of AI tool sets, AI SDKs, AI stack models that we’ve made available to all the developers. So, we give them all the documentation and tool sets, and you as a developer can take that tool set, start writing your code. The tool set will give you all the changes required to make the model fully optimised on Snapdragon architecture. We have built that over a long time.

On top of that, at Microsoft Build, about two weeks back, we launched something called AI Hub for compute. It takes the ability of a developer to write their AI application on Snapdragon to the next level and makes it super convenient for them. You log in to AI Hub for compute, and it has about 70 or 100 models that are already optimised — stable diffusion is one model, Llama 2 is another, Hugging Face models are already there, that are already optimised for Snapdragon architecture.

What you can also do is, if let’s say you are a developer and you want to optimise for Snapdragon PC, you can create your model, upload it onto that portal, and it will run that model on a physical Snapdragon X Elite PC, and give you back the results to tell you, ‘hey, this is how your runtime was, this is where the optimisations can be, this is how you can debug it’ — all remotely. You don’t even need to have a Snapdragon X Elite as a developer device available with you. You can do it all remotely. You can build your own app with it. It will help you deploy it as well. All of it is super seamless and the traction on that has been tremendous since we launched it.

So, some of the models may not be fully optimised, as you said, they may not be regionalised. But we expect a lot of the local community, local developers — we have a partnership going on with local Indian developers — to partner and enable those AI experiences as locally as they want for their respective markets.

AI was the dominant theme at Computex Taipei this year

Gadgets 360: AI features are still in the stage of infancy. We’ve seen features that are focused more on helping creatives with their work, providing them new tools. But when do you see some of these features evolving and helping people and communities who are more in need, like aiding people with disabilities or people with less access?

Nitin Kumar: The variety of applications that we’re seeing, to be honest, is actually quite wide, ranging on a broad spectrum. There are creative applications like Co-creator, that help you create images with the help of prompts. Then, there is Llama type models, or large language models, which can write essays. So that’s on the productive side. I’ll give you a few more examples. Djay Pro application can help DJs mix music. You play any song, and it can quickly take the vocals out or just take the drums out or take the bass out and isolated noises. That is an apt algorithm to run on the NPU in a very low latency form fashion, locally on the device.

Then think of it from a developer perspective: You’re a software developer, you can generate code on the fly locally. There are a variety of these applications. I truly believe this is the start of a new sort of revolution and a new class of applications, because the power of the platform is just immense.

Gadgets 360: We’ve seen how quick the pace of innovation for AI has been in the past one-and-a-half years. It feels like when OpenAI kicked open the door, everything else fell out, as well. Do you see this intense pace of innovation being sustained or will it slow down and reach a plateau soon? The PC market in the past decade, before AI, seemed to have flatlined.

Nitin Kumar: Because of the disruption that this technology brings, as with any new technology, I think the pace initially will be quite disruptive and at an inflection point. You can draw a parallel to when smartphones were just coming around, be it the Apple ecosystem or the Android ecosystem. And the App Store was coming out with just 10 apps, 1,000 apps, and then 10,000 apps and then millions. And then it just grew very rapidly from there. And a whole series of use cases emerged out of the device, that were just unimaginable. You can go back another decade and think of the Internet as that moment. It was very hard to say (at the time) how many websites will show up or what kind of use cases will show up, or how does this thing even evolve? It’s unimaginable at that point.

But you know that the essence of the technology has enough of a spark to drive that kind of a revolution. So, my personal belief is that AI is definitely there. And of course it’s very widespread — from a cloud AI perspective, server AI perspective; we’re focusing a lot more on the on-device perspective. When you look at it in totality, this will completely drive a new set of devices, experiences, innovation, apps. The usability aspect of it is just hard to imagine, but that spiral effect, at least for a few years — it might be longer, as well — will continue at a very, very rapid pace. And from a Snapdragon perspective, we’re very excited about where we’re headed.

---