The Next Decade of Compute

Geof: Good day and welcome to the Arm Viewpoints podcast, Episode One, The Next Decade of Compute. And I have to say I’m really excited about the topic and the guest we have to discuss it. With us today is none other than Rene Haas, President of Arm’s IP Products Group, IPG and a member of Arm’s Executive Committee. For anyone who has attended any major Arm event in recent years, Rene will be no stranger to you, as one of the company’s strategic leaders, and an experienced technology panel moderator in his own right. Red Eye is poised to offer us something really special today. Welcome, Rene.

Rene: Thank you, Jeff. Glad to be here.

Geof: So with that, I’m going to jump straight in Rene, I understand that Today is a big day for Arm with a major announcement about a new architecture that will drive a vast array of specialized processing solutions, the Armv9 architecture. So what makes this Armv9 architecture so important for the future of computing? No, no pressure.

Rene: Thanks again, Jeff. It is a big day Arm celebrated our 30th birthday at the very end of last year. And we’re here to talk today about version nine of the architecture. So for the engineers out there, you can do the math, and you realize that this is not something that we do every year, or every couple years, it’s actually every few or several years that we actually come out and talk about something new in the architecture. What’s really special and unique about version nine is what it means across everything we’re doing across compute. Now we’re known for CPUs, and Armv9 is really about the CPU architecture. But what’s really exciting today is what we’re talking about relative to everything we’re bringing to the whole landscape of compute, not just the CPU, but GPUs and GPUs, we stand for neural network processors, really everything that brings to bear all the technology to have our partners develop the most amazing solutions for their customers.

Geof: And one of the things you talk about with this new architecture is specialized computing, which I find really intriguing. It appears to underpin some really key concepts behind the design of armv9, can you explain it in further detail?

Rene: I spoke about computing being really not just about the CPU, but really about everything that goes into compute and compute technology. And I think the best way I might think about it is in this pandemic, which has really, really changed our lives, it’s become very, very obvious that technology is indispensable, and the entire world went from working in our offices to being really comfortable with or, or maybe not so comfortable with zoom calls every single day. And people learning and working from home kids taking, you know, online classes. But what it did do, I think, is accelerate a number of key trends that we were starting to see and that technology is ubiquitous to everything that we do. Without it you literally cannot get anywhere relative to your daily lives and then when you start to think about all the different ways that technology is coming to bear, it’s natural to think about, not every compute element can solve everything’s probably everyone’s problems. For example, a CPU which we’re known for is a wonderful processing tool. But there are jobs, tasks, workloads that might not need, or might need something beyond just the CPU to accelerate the application, that could be a GPU, that could be an NPU, as I talked about. So really, version nine is around taking all of the key things that we’ve learned over the last 30 years around CPU technology, extending that towards the specialized compute elements. And then really underpinning that all with one of the key elements, which we’ll talk about later, which really underscores v nine,

Geof: And that’s security, right? So what types of use cases will this specialization deliver beyond the really broad range of use cases currently available? The possibilities seem really exciting.

Rene: Yeah. So again, I mentioned security, and version nine is really all about bringing a number of fairly sophisticated security elements to bear. First off, when you think about all the intelligent devices that are around us, I look around my room sitting in my home here, whether it’s a set top box, whether it’s my toaster, there’s a lot of things that are intelligent, that have microprocessor in them and they are connected. When these devices are connected these intelligent devices, they are subject to security breaches, intrusions, getting access to data in such a way that you didn’t intend to. So as a result, really keeping security at the forefront of what we’re doing is going to be really, really critically important and there’s a number of different ways that that can come to bear. One of the most powerful computers that exists out there today is what everyone carries in their pocket. And that’s your smartphone, everything is in your smartphone, your it’s a digital wallet, you make payments with, it’s really everything is there. That is the amount of security that needs to be tied to what goes on to that device is huge. So when you think about things such as simple as apps that sit in the app store, you want to make sure that those applications that are in the app store that use the operating system that use things like the supervisor and hypervisor of the CPU, that the data, the data that’s being used by the application is completely protected, you’ve got to do some things around that data to physically isolate it and protected off from anything that could get access to it. And that that happens really at the architectural level. So some things that we’re going to talk about with v9 really are those type of areas, ie you’ve now put into the smartphone through version nine. Think about it, no separate data realms, if you will, where the data lives in a protected category, where because of things that we put inside the micro architecture, we can assure that when that end application has been hacked, or some bad actors there, you’re not going to get access to that data.

Geof: That’s fantastic. It also looks to me like the architecture will have a big impact on AI processing by several magnitudes. So what types of specialized processing Do you think that will enable in the future?

Rene: Yeah, it’s a great question. I mean, when you think about AI, and how sophisticated it’s starting to get, and that’s just, you know, from thinking about applications that predicted, which are going to which you’re going to go say, and I’m like, I’m using email applications, I’m sure you do, too, Jeff, that do a pretty good job now of predicting, you know, the sentences I’m going to write and where things are going to go. And that’s just a small, small, small micro example of where AI is finding itself into use cases. And you know, what, it’s starting to work, you know, voice assistance, for example, I think the prediction is going to be 8 billion voices that by 2024, which is probably the population of the planet, you know, plus or minus by 2024. Those voice assistants are going to be incredibly sophisticated in terms of, you know, recognizing your voice, the patterns of your voice, voice prints, that these will be very natural key ways of how people identify themselves. Just even thinking about my passport, my passport expires in 2023 and I’m old school, I like to look at the stamps in the passport, and then I’m kind of wondering, gosh, that next passport is probably going to be an AI enabled device, it’ll be a combination of voice print, eye print, all those type of things that come together and be far more secure than what a passport looked like, you know, we’ve done a lot of things around the architecture in terms of specific instruction extensions, things called Scalable Vector extensions, we’ve done a lot of work with partners, like Fujitsu, to get a much higher degree of performance than they’ve gotten in the past. Again, v9 is going to be about security, but it’s also going to be about performance.

Geof: Well, picking up on another aspect of security, I’m seeing that the key piece, a key piece of this armv9 architecture is the importance of confidential computing, which if I understand it correctly, involves protecting data within areas that cannot be accessed by platform administrators, service providers or platform software. And I know you talked about that a little bit earlier. But can you explain what this architecture does differently to enable this, and what it will provide in terms of security?

Rene: Yeah, so confidential computing is not a computer program running you know, with a confidential email or something of that nature, it’s there, it’s a much more cool than that. I’ll talk about it in a second. But again, some of the numbers around cybercrime $6 trillion dollars worldwide by 2021, which is an insane number and is set to increase. A lot of that is not the classic person who calls someone up on the phone and tricks them into moving their bank account, it’s done much more silently, it’s done in such a way that the data is going to be it’s going to be taken. So I talked about a little bit a previous example about realms, which allows the data to be in a protected area, such that if the application is hacked, the application is not going to be able to get access to your data, confidential computing is the next level, and that’s actually doing the computation of that data in its own hardware based secure environment. So literally, and I know it sounds a little magical, but think about an application. Now where based upon what’s being run, not only the data, the data areas protected, but the actual compute element that’s running and executing on that data that is also cordoned off away from the normal compute space. So it gives administrators just a much higher degree of control over what’s going on with that platform. So now you have not only the data is in a protected space, but the compute elements is actually computing on that data is also in a confidential protected area, which is huge, right? Because when you start to think about the kinds of things that you can start to cordon off, not only the data section, but actually the computing area in its own space, if you will, it’s going to be huge in terms of the impact.

Geof: So you talked a lot about specialized computing and I’m wondering, from a developer perspective, what’s that going to mean for people who are developing applications?

Rene: For Arm, v9 means a lot. When we talk about specialized computing, as we talked about, it’s not just the CPU, it’s the GPU, it’s the NPU. It’s all the all the programming elements, it’s the tool chains, it’s the compilers, it’s the linkers, the debuggers. That also means that we have to think about making sure that we are with our partners, providing developers the best experience we can so that C, C++, PyTorch, TensorFlow and whatever the next 10 years brings. Rest assured that Arm and our partners in the ecosystem are working really, really hard to make sure that it will just work for the developers, developers have whatever they need, again, whether it’s around their debug environment, their compilers, their assemblers, and really making sure that the best solutions can be developed on our own by our huge amount of community, because the amount of people who develop products and software on Arm is vast. And as we get towards the era of specialized computing, we know that the programming environments are going to be a little bit different for each one of these specialized compute nodes and it’s our job to make sure that it just all works.

Geof: I can’t help but notice that this year marks 10 years since the world first learned about Armv8 and I’m sure in the last decade there have been innovations powered by v8 that Arm was pleasantly surprised by with, and that this new armv9 architecture announced today will unlock its own set of opportunities in the next decade. So what are the big differences you see in the world 10 years from now? I know prediction, particularly after last year, it’s a difficult thing to do. But let’s have a crack at it.

Rene: It’s really obviously it’s super challenging to kind of predict what, what 10 years look like and but I would look back and say a couple things. We think about Armv8 when Armv8 came out: 4G was also starting to become a ubiquitous platform 4G LTE, and when you think about 4G, LTE and VA, which was the first 64-bit architecture that we did, I can assure you that that no one stood up and said, the killer app is going to be ride sharing and being able to have location based rentals where I can pick out a number of different homes and areas that I want to rent all enabled by the high-performance compute and low-power that went into smartphones, and then at the same time, a very, very fast bandwidth data network. So if you fast forward to the next 10 years, you’ve got 5G, and I think the Arm partners over the last 30 years, and the total number we talk about is 180 billion chips shipped by all our partners, which is a crazy big number, you know, going back to the population of the earth is you know, seven or 8 billion. I think 180 billion is not only every person who lives on earth today, but I think it was ever lived on earth ever. So these are big, big numbers. That means the data is going to be a tsunami in terms of its uses and where else it will be. So back to the predictions, I’m not a great predictor, but if I think about ubiquitous computing and computers everywhere, and then very, very fast and secure networks that can transmit the data. I think health is one of the big areas that is going to be the tipping point for us going forward. I think the digital wallet that one has in terms of your health information and everything around that, both in terms of not only your health information that allows you to get access to doctors and health clinics much faster. Because today just think about it right. It’s an industry that’s fairly antiquated in terms of how it handles data, just look at how tricky it still is to sign up for COVID vaccines of that nature, I think 10 years from now, because security will get much better, people be much more confident in the information. I think that’ll be a huge breakthrough. And then you just look at how quickly some of these vaccines were developed. Some of these vaccines that were developed, particularly they were around our mRNA and that’s a highly computational difficult problem to go off and solve. So I’m very intrigued about what can be done around the health industry in the next 10 years and hope that Arm technology will be around it. I’m pretty confident that it will and I think v9 will provide a great underpinning to all that.

Geof: Yeah, that’s a fantastic set of predictions. I the one other one that kind of occurred to me, and I wonder what you thought about it was in relation to climate change in a lot of the modelling work, I know that HPC and other high performance implementations have been used to try and do climate modelling. But I’m wondering, as you look out to the nine, what role you see it possibly playing?

Rene: I think there’s two areas, I think there’s a lot that we can do around to your point, armv9 and the compute capabilities, doing a lot of real modelling around what’s going on with climate change, whether it’s weather patterns, you know, certainly here in California we’ve had some horrendous last number of years relative to fires and such. I think one of the other big things that we’re that we’re doing with v9 and really, with everything we’re thinking about in terms of total compute, is reducing the carbon footprint, because it’s very, very clear that sustainability and everything that’s gone around all the good that technology can bring, technology also needs to be responsible for keeping the carbon footprint low and to carbon neutral because if you think you’re the kind of things that that hurt our planet, in terms of sustainability case it’s the amount of energy that we are consuming so we’re very focused on that. I think one of the things armv9 will hopefully allow is even though we’ve got more and more computing, the ARM architecture as we know is well known for its efficiency and low power and that’s a big focus for us. So I’m very hopeful that that will be a big area for us as well.

Geof: I’m sure we’ll be hearing a lot more about where Armv9 takes us. We look forward to bring you more news in the next episode of Arm Viewpoints and look forward to connecting with you all again soon. Thanks for listening today. Thank you