 So we're here right after the ARM press conference here at Computex and you announced new ARM processor Yes, so this morning we talked about the next generation of ARM Cortex technology We announced the Cortex-A75 and the A55 processors These are the first two processors that match up with ARM's dynamic technology And as you know dynamic technology allows you to build More integrated more heterogeneous systems and are designed ground up to support Better AI functionality Better robustness and reliability features both for enterprise as well as functional safety critical applications such as automotive and How to connect in better acceleration again for better Improved machine learning computer vision capability as well as more integrated infrastructure solutions This also complements the announcement of the Mali G72 GPU that goes along with all of this So dynamic ARM dynamic is does it have to do with the? cash-correct interconnect kind of stuff so ARM dynamic is a platform which will be the basis for the next few generations of processors. Yes, it involves a Level three caching system, but alongside that it is about design It's a cluster that can support up to eight cores and each of those can be different in configuration and capability But architecturally compliant, but also this this system is designed to us to support more enterprise ready robustness features with it and they also support cash stashing and architecturally different Capabilities like data poisoning etc. They're built into CPUs. It's also dynamic built from the ground up so that it can be Certified for acel D or functional safety critical capabilities So are you announcing now that there's more cash closer to each core? something is it something that's similar to what Intel is doing on a Xeon or I Mean without asking specifically, but does that provide a whole new new whole bunch of new performance? so as we've gone about naturally we have level one caching gone up to 64 kilobytes for Configurably for both a 75 and a 55 What we've done this time around As a part of this dynamic capability is each core now has a private L2 again very configurable to reduce the Latency into the core itself But also giving it the level of isolation needed so that each core can operate independently in a larger level 3 system As things get more complex and as performance we go up the performance curve Naturally a large part of the benefit is going to come from well-designed memory systems that also need layered caching for them and Now you announced a whole bunch of new cluster configurations It doesn't have to be a It could be five plus one or something like that. Oh, what is it? So? certainly dynamic is a Platform supports up to eight cores in a cluster It's naturally a good fit for big little because not only can you do? Transactions between cores much faster and transaction of state between cores much faster It uses the same software as before What now this offers you is a flexibility to put in a different set of configurations of big and little in the traditional context before you had Maximum of four cores per cluster. So if you wanted to do one big and seven little it wasn't easy to do without adding a third cluster Whereas now in an eight core cluster, you could have just one big processor Especially in a mainstream kind of setup that gives you the single thread performance you need But then taken up with six seven Little cores that actually give you their throughput and bandwidth while maintaining the power efficiency of that system So the big little idea and all these clusters definitely has turned out to be the right way to do things, right? You want to have different clusters? You want to have Even some a 55s might be smaller than the other a 55s and like some high bigger ones Yes So the one thing that we can certainly say is across the smartphone world and certainly into some of the larger Clamshell form factors as well big little is pretty much everywhere. It's ubiquitous And that comes from the fact that you can do the right amount of compute for the right workload And I actually have the flexibility to do that any time What dynamic does now is takes it to the next level because the efficiency has grown Plus it has given you a lot more scope for mixing and matching the right level of performance for the cluster that you choose and And in your keynote you were talking about notebook laptop Push like there's a new there's a new push there, right? You're doing even more performance, but still keeping it a mobile power consumption So as you know, I mean the a 73 generation Or a last Generation class you are seeing much more push into the clamshell form factor What a 75 does is now pushes up that to another level as we said Compared to 2.4 gigahertz products today in the smartphone world with a 73 an a 75 and 10 nanometer Probably be up 3 gigahertz, but still providing over 50% raw Integer performance the floating point performance would be even higher and specialized AI type workloads would be much much more than that So what that really does is provides you with a substantially better Performing processor which can support large screen compute form factors But still maintain the efficiency that is expected for a smartphone Is this targeted towards 7 nanometers 8 nanometers 10 nanometers or any? so firstly all the IPV provide is Portable, so you could take it to any process note of your choice We expect for the premium mobile side. This will first show up in 10 nanometer But realistically it would be still a very good fit in 16 nanometer or 14 nanometer as well And if you look at the a 55 which is more mainstream you could see it in 16 You could see it in 20 or even 28 for that matter. So really this is a very wide set of solutions Will it go into 7 nanometer? There's nothing stopping it from it and So those are two new 64 bit chips and they are Coming with arm v8.2 Reinstalled or how do you say that is good? So they're built on the arm v8.2 architecture so that builds on v8, which has 64 bit capability But adds in a few key benefits beyond that firstly, we provide Atomic operations that are more targeted towards enterprise type applications where large Sets of clusters can work even better and more efficiently together We have advanced virtualization support well beyond what we had today so make transitions between guest operating virtual systems faster but v8.2 adds in a lot of advanced features beyond that Firstly for reliability. It has more features It also adds more enterprise features to enable data movement and cache stashing that can pre populate what the Accelerators work on and what this in the cores work on and Gene able more Innovation in accelerators on the SOC people can put more stuff now Absolutely, so this is actually more where dynamic comes in because it allows the accelerators to plug right in and what we've done is improve the Latency and the bandwidth that it's capable of addressing these accelerators in a much better way So you could see neural net type accelerators plugging in for AI type solutions You could see deep packet inspection or other cryptography accelerators for networking or infrastructure type solutions and All of this in a more efficient fashion So it feels much more tight-knit in a heterogeneous computing environment And so there's a heterogeneous computing kind of consortium and some other consortium You kind of you you announced that you part of to get all the industry players to work together on this Interconnect solutions, right? Yes, so we have worked with Heterogeneous computing before we continue to do so and if anything right now For the networking server world an infrastructure world There are a lot more open NFV open DP DK all those types of platforms coming in as well that allow The main software stacks to run on the core, but also allow for differentiated solutions to under to link into Acceleration or specialized hardware underneath So there might be I mean there will be some awesome laptops. They will be this is this is potentially good for servers also indeed, so Certainly as we said Laptop class or notebook class performance now available in a smartphone on the look so you could certainly see that in The infrastructure world again Efficiency is still key. You have more threads on the same amount of dye Compute density is really what people are looking for and with the advanced performance Advanced floating-point capability advance AI capability, which still stays within a much more tight constrained budget It really changes how infrastructure solutions can be delivered. It improves the TCO versus incumbent architectures and Chipmakers are allowed to take your designs and modify and there's new these new licenses that this been announced, right? That they can take and tweak and improve So let me clarify this first So if you're talking about the built-on ARM Cortex technology, it is not meant for partners to take and modify We work with the partner. We take a standard product work with a partner to modify it to The partners needs, but we still deliver the product all right, so This is 2018 every time you announce and it comes 2018 but every time there's some people shipping before the end of 2017 anyways, so that's maybe possible again so Naturally, there's always going to be prototypes before volume silicon we expect a 75 and a 55 base platforms to actually hit devices and market shelves Certainly by the second half of 18 probably sooner Which naturally means that there will be prototype platforms before that some of them may be early production as well so we do expect that in the not too distant future and a 75 sounds awesome, but maybe the big big big volume is The successor of the a 53 which is the successor of the a 7 right which is a huge volume And so what have you done to improve that that that market? So if you look at the Cortex a 55 as you say it has a lot of Great history a 7 has been one of the most popular course that has gone into the mainstream as well as the entry-level market The Cortex a 53 did the same thing three generations of products If not more that were designed for the mainstream and the entry-level market what a 55 does is now Improves the performance Improves the efficiency and gives more capability so you could see more AI type workloads being run on an a 55 and that Changes the way you see mainstream smartphones today plus it puts that into Edge compute into single board computing into industrial and home automation Into Edge and IOT so really a 55 takes what a 7 and a 53 have done and takes it to the next level And it's tiny right is such a small part of a very compact part so compact, but a whole bunch more so yes We continue to focus on mainstream with a 55 for a greater performance greater capability While maintaining or substantially improving the efficiency for that segment You