 at Mobile World Congress 2016 and who are you? Hello, my name's Neil Wordman, I'm the Product Manager for the Cortex R8 Processor. And last week you announced a new Cortex R? Yes, so that's the Cortex R8 which is a new processor from ARM that's been developed from the Cortex R7. So it doubles the performance of the Cortex R7, it has a number of other additional features such as larger TCMs, a fast path per core to talk to peripherals such as DSPs or hardware accelerators. So again, it's giving a lot more performance, especially for our partners that are looking to develop LTE Advanced Pro and 5G modems in the future. So this is R, it stands for real time? Yes, so at ARM we have three different types of processors. A for application processors, R which are the real time and the M's which are the smaller devices for microcontrollers. So I focus on the Cortex R series and again, yeah, very much, very separate, often not as public as many of the other processors because they're often deeply embedded inside specific devices, be they storage, modems or many other real time devices. So you're in hard drives and SSD? Yes. We're in hard drives, we're in all of the leading hard drives now, have been for many years. Also in SSD is a very rapidly growing market where again we're working with all of the key manufacturers. How come you are the architecture of choice for that kind of segment for the market? I think one of the key reasons for storage is the real time aspect. So certainly in hard disk drive, when you're controlling a servo motor with these heads moving very, very fast, then the real time aspect is critical. When you move to SSD then what we find is a wide spectrum of devices from the small devices controlling the flash and managing the errors, all the way up through to consumer devices that typically use a Cortex R5 to manage the host interface, but also the flash translation layer, managing all of the tables that keep up to date with where all the different blocks are kept. And in enterprise storage we've seen success with Cortex R7 already, where there the IOPS are growing very, very rapidly every year. So very quickly people are needing more and more performance and also lower and lower latency. For financial applications, if people are using SSDs in those devices they really want to have a very, very low latency and it has to be the worst-case latency is what their customers are looking for. So if they're looking at real-time share data, they can't have it, but every now and again there's a slight delay on some of them, they need to be there very, very rapidly. So from R7 to R8 is double the performance? So the double the performance primarily comes from moving from a maximum of two coherent cores to four coherent cores, but also in the real-time world the tightly coupled memory is very critical. So the tightly coupled memory is directly attached to the processor and it's typically zero weight state. So in those memories you can keep all of the critical routines and all the critical data so that when an interrupt arrives you can immediately start running it and access it. With the Cortex A processor they're very good at running Linux and for overall average performance very, very high power, but what may happen is that maybe instructions and data could be evicted from the cache. So when that interrupt arrives then actually you've got to go through the memory system and fetch those instructions. TCM's prevent that and with the Cortex R8 processor now we've increased the size so that each core can have up to two megabytes of TCM. With Cortex R7 that was limited to 256 kilobytes per core. And the Cortex M doesn't do the R stuff? Well the Cortex M is typically, it's a very small, very, very low power processor. So something like the M0 plus is perfect for the back end for managing flash. But the Xtreme M, Cortex M7 now does have TCM's and caches but that's the first M processor that has caches and TCM's available. So they're typically a shorter pipeline and typically only run on, sorry, thumb instructions. For Cortex R there is a lot more sophisticated programmers model with special routines to handle any exception that happens. So it's just a more sophisticated processor. So the R has its own kind of like programming world. Is it also Linux or what happens on R? So with Cortex R a key differential of the A class is they have a memory management unit and to run Linux you need that memory management unit and it's, the memory management unit presents to the software a really like a huge flat memory space so that applications can just allocate memory and work really well. But in the background there are things where you may have to swap pages of memory in and out. So doing a page table walk can take a thousand cycles or more and then in a real time environment that you can't have that MMU because that just breaks the memory. So what OS do you have? You have real time OS of some kind of, who's making them? So we have a whole range of partners in our ecosystem that provide RTOS's. So all of them from, you know, Nucleus, Red X, Green Hills, pretty much you name it. It works on Cortex R. And so they're all very happy about the R8 announcement because that's going to give them new work to do? Definitely, definitely. I think it's a very exciting area. Quite often these are very sophisticated applications. So in a modem, say for 5G, where you're handling multiple carriers, you need a sophisticated RTOS that can really reduce that latency and provide the best overall user experience for the guys using the handsets. So you say that 5G is going to be enabled by this? It's not possible to do with the previous R? What does it actually do in the system? Where do you feel the double performance? Where does it go? Okay. So what we've seen as we've moved through from initial GSM handsets all the way through Edge 3G and to LTE where we are now, moving into LTE Advanced. Now at this point, there's really a maximum in LTE Advanced, there's a maximum of 4 carriers that are available. So this means perfectly 4 separate data paths and these are all LTE carriers. As you move to LTE Advanced Pro, which is coming up very shortly, the number of carriers increases but also those carriers could be Wi-Fi or could be unlicensed. So there's a whole range of new carriers and to be able to manage each of those carriers is very time critical. In some cases you're talking to two cell towers at the same time and you need to then effectively duplicate that handling. So you just need more and more performance and that's really where Cortex R8 comes in. It's the fastest real-time processor available in the market at the moment by quite some considerable margin. So you announced it, how long does it take before people can buy modems and maybe also hard drives with R8 or other stuff with R8? How long does it take? Realistically, we do expect Silicon this year from some of our lead partners, for modems my suspicion is that some of those early devices will probably be test silicon, well they'll be proving it. To develop a modem typically takes maybe two years and that's really for hitting hands heads because it's not just developing the silicon. So once ARM delivers the technology for the processor, our partners develop that into a system on chip for the modem or maybe including all of the application processes. When that's ready you still then have to go through the type approval process with all of the network operators globally. So modems do take some time to come to market. But in other ARM markets such as storage, I think you're going to see it very very rapidly deployed. The storage is also looking forward to R8? Yes, definitely. What can they do with that? So again it's really about performance but there's also some additional features we've added as we move to Cortex R8 which enhance the error management, the error containment and really develop a system for certainly for enterprise storage where you really must make sure that the user's data is never corrupted and Cortex R8 has some very neat features that ensure that that can never happen. So you didn't just increase the performance, you added more stuff? Exactly, yes. So there are some areas where we really enhance the error detection and error containment and there's also other areas where we just done some extra enhancements to the Cortex R8 just to improve this. Nice. So now you just go around, you talk with all your partners and help them implement it? Exactly. And again we launched it today but we've been talking to our partners for a very long time. And actually Cortex R8 has been driven by our customers requirements and what they're looking for moving forward.