The Future of App Compatibility on Snapdragon

Welcome everyone to Espeaks from eweek. I'm Corey Nolles. For enterprise IT teams, application compatibility is no longer just a question of whether software opens. Organizations also need to understand how applications perform, how they interact with security and management tools, and whether today's devices can support the AI workloads being built for tomorrow. Joining us to discuss the evolving Windows application ecosystem is Manish Seshmuk, senior director of product management for Snapdragon at Qualcomm.

We'll explore native emulated [snorts] and hybrid applications, the growing role of ondevice AI, and what IT leaders should consider when evaluating Windows on Snapdragon. Manish, welcome to Espeaks. >> Corey, thank you so much for having me over here and looking forward to exciting discussions. >> Same same. I guess to start off, most people know Qualcomm first through smartphones and mobile technology. Can you walk us a bit through Qualcomm's journey into the PC market and what made this the right time to bring Snapdragon more deeply into Windows computing? >> Yeah, I mean I I think most of the people um know about Qualcomm, but Qualcomm I mean it's been around for 40 plus years, right?

It's a technology company. It really stands for quality communications. That's what our um history started out from. So think 2G, 3G, 4G, 5G. We've been pioneers in some of these technologies. U but over time we have built our Snapdragon application processor. So we started out with smartphones, Android phones. I remember working on the first Android smartphone uh while at Qualcomm over here. So it goes long ways back, but since then we've evolved into multiple business units.

The Snapdragon processor now is not just part of the smartphones but also PC as you said XR variables auto IoT devices even robotics and data center so it's evolved quite a lot and uh compute as we call it or PCs has been a recent introduction uh for Snapdragon although it's been almost 3 years now since our first launch of the device uh maybe a little over two years um and it's been great um it's It's been very successful and we'll hear today through our conversations how successful it's been. >> So across PCs, data centers, and edge devices, we're seeing a lot more emphasis on combining performance with power efficiency.

What's driving that shift? And how is it changing what enterprise IT leaders expect from endpoint hardware? >> Yeah, great and very very deep question, right? I mean IT decision makers in my opinion think very traditionally but the mindset is changing with AI. I mean you just we are just at back of the health crisis with co uh you know we're just coming back and we we are on to the next crisis energy crisis compute crisis the IT decision maker today has to think about all those and really built their um device and and infrastructure for the future.

So let's break it down, right? I mean what I mean by compute crisis as an example, right? So when you think about AI compute today, you think about data centers and primarily GPUs. But yeah, >> um from IT decision standpoint, you have to think about inferencing because once you deploy AI into your infrastructure, you have to utilize it for your use cases, which is through inferencing. Um and if you can't really buy GPUs, enough GPUs, you have to make it available across all the devices that are u available.

So if you really think about um the employee base, every employee [clears throat] has a smartphone. Uh maybe you have all these PCs running around. And for Qualcomm, the way we look at it is it's any device that is connected has a computational capability. So >> we we think about solving the compute crisis through uh providing and enabling all these AI inference points uh for IT decision makers to utilize through their employees devices. And secondly, the energy crisis again when we say energy crisis, it really is power that is drawn by these data centers.

And for any AI workload, it's going to translate into power. And um by having a distributed workload that works across multiple devices, it makes it much more efficient. And Snapdragon has a lot of legacy and efficient hardware architecture that will go into that. But >> um just going back to your comment, right? I mean an IT decision maker has to think about AI as a future. They have to think about how do I utilize compute and how do I save energy.

So those those in my opinion are are are good shifts. They are they are I agree. >> Yeah, Snapdragon has that efficiency inherited into into its uh architecture. Right. So unlike x86 which are most of the PCs today um they rely on like complex instruction set computer um or x86 architecture is that whereas Qualcomm is really riskbased architecture. So risk based is really reduce instruction set computer. So by nature of it, it's going to run much more efficiently than any of the x86 systems that they have.

So if you have Snapdragon PCs, um it's going to run much more efficiently, especially for your AI workloads running locally as an example for future. We also have specialized hardware within um uh Snapdragon for AI. Um so traditionally um these PC processors have um CPU which is your central processing unit. You have GPU which is your graphics processing unit and majority of the AI workloads and the way uh people are utilizing AI inference is on GPU today >> but it's not really meant for AI.

Um at Snapdragon what we have is a new processor called NPU neural processing unit which is really meant to be running machine learning workloads right so think your neural network operations um it's really meant to run those so when you as an IT decision maker are thinking about um these future and thinking about most efficient compute hardware um Snapdragon already sort of like is built towards that >> I think it's really interesting that you bring up NPUs and and we're seeing those more and more now and I was wondering if you could kind of explain the value of the NPU and what it does different.

I talked about compute and energy u u efficiency both alongside what npu is as I said it's a very AI ccentric um processor so really meant for running your neural workloads so if you have a model running on npu it's going to much run much more efficiently and it's going to run quick so there are kpis like tokens uh per second which is essentially how fast uh an AI model could run and Then there is tokens uh time to first tokens. So how quickly AI responds and how um fast and then there's also accuracy.

So like when you ask a question to AI how accurately so we've been measuring across all three KPIs and NPU consistently beats any other processor. So uh we strongly believe the NPU is the right way. The other way of looking at it is when you are an IT decision maker you know AI is future. these AI um agents are going to run in the background all the time. Think security as an example, right? Uh we have more and more security application providers now utilizing NPU because um they want to run it all the time in the background, run it very efficiently and not bother the foreground processes like that are consumed by CPU or GPU, right? and Snapdragon architecture with dedicated NPU for AI workloads makes it much much more future ready for what IT decision makers want.

IT decision makers also run into a couple of um problems today which is like how do I use AI? I mean a lot of companies have thought about what AI use cases are them for today but it's still evolving. I mean, every six months, Corey, I look at it and like, "Oh my gosh, I didn't know AI could do that." So, it's going to continue to evolve, right? And >> the other problem consistently I hear and I see is we keep running out of tokens. Um, so with with Snapdragon and with NPU, you're able to save a lot of your costs for cloud because it's going to run locally on NPU and it's going to run efficiently in the background without hurting any of your employee experiences that are running in the foreground.

So Snapdragon is the way to go in my opinion. >> That's so awesome. Another question, application compatibility is often a big concern raised when an IT organization starts thinking about a new compute platform. [snorts] Where does the Windows on Snapdragon application ecosystem stand today like particularly for enterprises who are moving from traditional x86-based environments? bringing a new architecture in Windows ecosystem wasn't going to be easy but we work really hard for last few years and I I I think we are at a point now where consumer apps is not an issue anymore like >> you buy a Snapdragon PC I know you have core >> I have one and it's not been an issue for me >> yeah yeah and and you you bring up and any apps games you know it just works out of box so um we've changed a lot of that perception in last couple of years right once we've initially launched a lot of people had concerns but not anymore.

When you think about enterprises and when you think about IT decision makers they they think about app compat in sort of like three folds. One is the popular consumer apps that your your employees want to use. Of course that's not an issue anymore like I mentioned. The second is the popular IT apps or IT apps that are specific to my company that I need to run. And the third bucket is legacy apps. Maybe I have home homegrown apps or applications that I've been using for last decade.

So that's the third bucket, right? So consumer apps as I said it's done. Um IT apps, we've done quite a lot of work over last couple years. Um so we really started out with >> interviewing hundreds of IT decision makers and saying okay what are the top IT applications that would be required for us to support and we've been going out talking to all these IT apps makers making sure their application is compatible basically functionally compatible the second is >> they can port it to uh Windows and Snapdragon natively that means if there is a kernel mode driver dependencies that applications have.

We go work with them. We go resolve those uh dependencies as an example. And then further there are some applications like EET as an example is a security IT security application. They went ahead and integrated the AI on Snapdragon NPUs, right? So we continuously engage uh with these IT application vendors. I I feel proud to say that you know we today we've um support majority of all popular IT applications out today. um security as a great example um and uh we continue to stretch that.

One of the good things that we've worked on is identifying all these top IT categories, IT application categories and and making sure all those run. The third is the legacy applications and this is where I think it's important for us to work with these enterprises to ensure their workload run on the Snapdragon PCs. And the harder it sounds like, oh, I've had this app for a number of years and it doesn't run. Uh, it's actually not that hard because Microsoft and Qualcomm work together on this Prism emulation.

So, >> essentially, if you have your application running on x86 architecture, it can run on Windows on Snapdragon in emulation mode. And funny enough, I think I've seen these applications running in the emulation mode on Snapdragon running faster than actual x86 because our CPU is amazingly fast. I mean, it's uh it's been it's been it's been great. So, um yeah, in those three buckets, I would say we feel pretty comfortable as an IT decision makers.

Um the consumer app should be no problem. You shouldn't be worried about it. For IT application, we have great coverage. We maintain our app compat on website called workssonwa.com works on wa.com. >> Okay. >> And then for legacy applications I think we have a pretty solid path with prism emulation. It can run out of box. Um but even if you have any issues we have fantastic u support and field application engineering that can stay on top and help you guys resolve >> well.

We don't native emulated hybrid applications can all be part of a modern Windows environment. How should it think about the differences between those approaches when evaluating performance, efficiency, reliability, and the overall user experience? >> A lot of these IT decision makers have a lot of dashboards as well. Yeah, >> we at Qualcomm as an example, right? um and even at Microsoft and some of these other uh companies that have adopted Snapdragon I I've reviewed a lot of these dashboards. um Snapdragon versus equivalent x86 platform.

First of all, I mean there are KPIs that um if the user experience is acceptable uh for any of the application, it's pretty obvious through testing. Um uh people have observed but IT decision makers also track like number of crashes, number of hangs in the applications over number of users, number of uh months of usage. Um, and Snapdragon has consistently been better platform than any other platform, right? Um, so emulation or not, um, I wouldn't worry too much about it, Corey.

I mean, I think at the end what it matters is is this app usable and does it crash or does it hang, right? It it's usable, it doesn't crash, it doesn't hang, then I think you you got it right. And I think we've been winning across all those. And it's not to say uh that native apps are not important. It's super critical and we've been working with them quite a lot and in general if I look at the PC industry or there was this fruity company that started with this emulation a few years back and they're going to retire the emulation pretty soon.

So a lot of these application vendors are already starting to migrate um and creating native Windows and Snapdragon applications rather than emulation. So the industry is shifting and as I said you know these AP KPIs of application shouldn't really worry about is it emulated or is it native anymore >> that seems like like a lot of a big undertaking as well like I know the success of any platform depends on its software ecosystem and uh and you talked about how you've been working with Microsoft and these other developers but I'm also curious about independent software developers hardware partners making it easier to build optimize applications for Windows on Snapdragon.

Is that a thing you all are working on as well? >> Yeah. Yeah. I mean I I think one of the one of the interesting thing for us on Snapdragon was up until now I think every platform that we worked on like mobile smartphone as an example or XR or auto it was all consumption platform but now PC is also a development platform. So we have to not just make sure that any application that exists out there for Windows PC runs on our platform but also all the tools that are needed to build any software out there runs on Snapdragon.

So I mean Microsoft has been amazing partner again I want to call them out but we went across um identified what tool chains software tool chains uh we need to run across our platforms as well as support and uh sort of like built applications on our uh platform as well and we've been working with them to to enable those and uh majority of the tool chains are Microsoft but there are a lot of open-source projects um that we continue to contribute to and make uh there is Snapdragon adoption in there.

Um so yeah, >> a lot of work like you said you know it wasn't easy a lot of work has gone through and we continue to I think how how we are ready for AI maybe maybe that's where you're headed uh next but that one in my opinion is an interesting challenge. >> It is it is it's going to and it's going to be for a for a hot minute I think. >> Yeah. So for an IT leader evaluating Snapdragon today, what does real world adoption look like? Do you have an example or anything of an enterprise that has deployed a bunch of Snapdragon based PCs >> for confidentiality?

I can't disclose the customer names, but >> Yeah, that's fair. >> Yeah, but I I did talk about Qualcomm being the zero customer for us. Um I mean at Qualcomm, we have um I think more than 35,000 uh devices um already. uh Microsoft has more than 45,000 uh devices and as I said we consistently see across dashboards less hangs less freezes with Snapdragon over the other and of course more much more efficiency like longer battery life um with employee base.

So that has been amazing story for us to rally around. Um in just last 12 months Corey we've sold to more than 25,000 unique customers in commercial and enterprise segment. So that just tells you like the the vast uh volume and the pace of it, right? It's just been in last 12 months. So just tells you the pace of um enterprise and and commercial adoption >> and that there's an interest. >> Yes, 100%. And these are customers that are multipleiz customers like small, medium scale, med mid-market or large enterprises. um and stretching across segments like financial like banks and industrial education retail and goes across global as well right so and we continue to um highlight some of these stories on our portal there's an example of Australian education system um using uh Snapdragon PCs as part of their employee database and uh employee base and and what what they observed is you know 75% of their issues device related issues are just gone with Snapdragon So when people think of Snapdragon, they were like, "Oh yeah, Windows on Snapdragon is a new thing.

It must not be stable enough." In fact, it's the other way. It's much more stable than what has been around. It's much more efficient than what has been around. And of course, it's future ready because of AI processing and NPU that I talked about. >> For sure. For sure. And and you know, I'm glad you mentioned uh PCs in the workplace because as organizations start to plan that next PC refresh or maybe an AIPC deployment of sorts, what should they test during the evaluation process and which signals can maybe help them determine whether a platform is ready for both their current application environment and the workloads that are coming around the corner?

Again very deep question but I I think today IT decision makers that we interact with um simply start with hey these are my applications that I need for my employee base do they run right and that's a great way to start but I would I would challenge that and say you need to look beyond that you need to think about all the factors that I mentioned which is AI is here and it's evolving very very rapidly how are you thinking about your AI use cases evolve over time and not to say that every IT decision maker is going to be able to predict the future but what I can tell you is the future of AI is hybrid.

Um so cloud is not going to be enough. I mean everybody keeps running out of tokens. So >> inferencing on the edge is going to be super super critical. So how are IT decision makers anticipating that and how are they making sure they purchase the right hardware which has local AI capability that does not disrupt the other processes that might be happening like I was talking NPU runs in the background without affecting anything that is happening in the foreground on CPU and GPU and as example runs very efficiently that's going to be another factor that they have to think also what are the tools and uh infrastructure that my employees need beyond just even the PC and do those run on this platform.

As an example, we just um >> we just announced our partnership with Hugging Face, right? Any AI developer today starts with a um a model hub like hugging face and they start and then they start experimenting. >> Uh that's that's where I started. >> Yeah. So, so, so partnerships like that that uh Snapdragon um offers and readiness for future is super super critical for um these IT decision makers to think about as well. So them not just thinking about today but also tomorrow and thinking much more holistically from am I actually solving the company issues that might be coming in future which is do I have enough compute am I utilizing it everywhere whatever my employee base has um am I running it much more efficiently am I able to unlock the use cases of future >> excellent well Manish thank you so much for joining us today it's been really interesting >> yeah Corey thank you thank you Yeah, >> before we let you go, where can people go to learn more about Snapdragon, Windows on Snapdragon and Qualcomm's work with the application developer ecosystem? >> Yeah, so a couple of websites.

One is works on wa.com. Um, this is where I think it's a communitydriven, but the list of applications. So, if you feel uncomfortable or you still have a little bit of hesitance, does my application run on Snapdragon? You can always go on that website and search for a particular app. But >> awesome. as you as I mean as people will realize it's just tremendously growing and even if the application is not there doesn't mean that it doesn't run. Um you can always reach out to us on our on our forums and then of course uh for anything related to latest on where can I buy Snapdragon PCs, what is the latest Snapdragon SOC, they can always go to qualcom.com/napdragon and we have a portal dedicated to commercial wins and success and stories like I was talking about previously. >> Excellent.

Thanks again Manish. I really appreciate it. For more conversations with technology leaders, visit eweek.com. Be sure to like this episode, subscribe to the channel, and follow EW week so you don't miss what's next. I'm Corey Nolles. Again, thanks for watching. We'll see you next time on E Speaks.

This transcript was generated automatically from the video's captions and may contain errors.