 My name is Surya Hota, I'm the Director of Marketing for Thunderex Product Line at Cavium. A number of people have talked about Cavium and Thunderex, Thunderex 2 in the previous presentations. Multiple folks were deploying it in their systems and also ARM, the work that we're doing with ARM. Brief overview of Cavium, we are a leading semiconductor solution provider for high-performance computing. It's one of the key markets that we address and also the cloud and data center. We have a variety of products that go into these segments. The processors are bread and butter. ARM-based processors have, as you have heard, Thunderex and Thunderex 2. And also we have other products that go into the data center including X-pline, which is our Ethernet switching product. And all the connectivity products as well, Ethernet connectivity, Fastlink and LiquidIO. So overall, and also we address quite a bit of security aspects of the data center as well. A lot has been talked about, the work that has been going on for a number of years now with ARM, Cavium and all the partners out there to provide a second, better alternative to the architecture that's out there in the HPC. The key components here are obviously ARM is the most widely deployed ISA out there, open ISA and a much broad ecosystem. So that's the effort of all the work that's going in. And what we are bringing to the table, again, as part of the ARM, as some of the folks from ARM talked about before the break, the tools and the compilers, they're providing specific to HPC is a great value. And we've seen that already when we ran the benchmarks on Thunderex 2, which we will talk about a little bit more. Cavium, what are we bringing to the table? We are bringing in the high-performance compute centric processor. We've been in the process of business for quite some time. And the high-performance is one aspect of it and also bringing in at the better performance per watt and performance per dollar advantages to our customers and to the end users. So I wanted to highlight some of the key press announcements that we made. All of them actually went out today that we have made and also some of our partners and our OEM customers have made. So first thing is the announcements by HPE Apollo 70 system based on Thunderex 2. And the Cray XC50 platform focused on, it's a supercomputer, so that is using Thunderex 2. That's on display at Cray Booth, where you could see XC50 blade with Thunderex 2. That's the ARM-based platform. And then we've also announced a couple of OEM platforms from Gigabyte as well as Ingresses. These are different form factors for a variety of different markets. Red Hat, thanks to John Masters and his team over seven years of work, the Red Hat announced the support for RELL on ARM. That's a big news as well. So let's take a look quickly where we were with Thunderex and then how we got into the Thunderex 2 into our road into the HPC market. So when we started off with Thunderex, we had quite a bit of traction in the system vendors that we were dealing with. Specifically, Cray and Lenovo, all the other ODMs and Oak Ridge and Hartree, they have done quite a bit of evaluations. And then with Thunderex 2, as you can see on the top left, we have improved performance significantly on a variety of different memory made with the memory or the floating point performance or the integer performance, we've improved quite a bit in Thunderex 2. That sort of propelled into the next generation of systems and also made the system bring up and things like that quicker. And as I mentioned in the previous slide, both HP and Cray have announced their system based on Thunderex 2. That's huge for us. And also, you've heard in the morning from Mont Blanc and Simon for GW4, all the work and the performance that they're seeing on Thunderex 2. And then we have a couple of other platforms as well. Atos is one, and then Penguin Computing on the ODM front. Yeah, a little bit more. I've talked about these already. HPE Apollo 70 and Atos Bull, Bull Atos Sikwana platform. They're based on Thunderex 2 and Thundere from Penguin and Cray XE50 supercomputer. So this slide has been shown in the first presentation by ARM. The key aspects, I'll just mention, this is our second generation ARM-based SOC, 223X performance improvement from our first generation ARM SOC. We have 32 cores, fully out of order, multiple threads, multiple issues. And then the memory, you'll see it in the next slide. We have 8 TDF 4 controllers. The memory capacity as well as the bandwidth is quite high compared to our computation. These are the key differentiators, higher core and the thread count. That sort of translates into the overall performance that we bring to the table. And the memory capacity is the other two, Bull Atos Memory Capacity and the bandwidth. When you compare with the Skylake, we are at 33% higher from those metrics. And dual socket support, and we have 56 lanes of PCIe for the connectivity side. So when you take a look at this overall HPC ecosystem, you can all at us that we've come a long way, right? All the way from maybe the hardware side of the fence or the software side of the fence. We have a lot of platforms built as I have discussed before. And also we have quite a few partnerships with the independent hardware vendors where we work together with them. And the OS optimized for ARM, as I mentioned, as we discussed, Red Hat just announced the support. And the other ones already have the support on ARM. So specific to HPC, open HPC framework is already running on ARM and Thunder X2. And some of the workloads, if you take the specific use cases for HPC, whether you have WARF and ECMWF and if you take the CFD, you have the open form. These applications are running on ARM and Thunder X2, right? And also a couple of things I wanted to point out from an application standpoint. We have recently announced that the EDA workload is running on ARM by Cadence. They call it Axelium, which is a parallel simulator that SOC vendors use to build leading edge chips. So Axelium is actually running on ARM, which is Thunder X2 based platform. So that's a great example of another workload coming on board onto the ARM bandwagon. And Hadoop, we have also made the announcement recently. Hortonworks announced that the commercial support of their applications on ARM, which is a big thing as well. So overall, thanks to everybody, including obviously ARM, played a big role. And all the partner ecosystem to get us to this point. So ARM Elenia Studio has been quite useful for us from workload optimization standpoint. So let's take a quick look at some of the performance metrics. Memory bandwidth, as I mentioned, we have more channels and you can actually put in more dims. So if you look at, when we compare with the Intel Xeon goal line 6148 processor, we come out at 30% higher from a bandwidth standpoint. Let's take a look at some of the spec CPU 2017 data. One on the left side is the spec CPU integer performance rate. And the one on the right side is the floating point data. And we have both ICC as well as the GCC numbers on Intel. As you can see, we're very much comparable when we get the optimizations and the compilers on the even scale with GCC. I have spec CPU int rate and spec CPU fp rate. And obviously, as I mentioned, we have been working and the work is still underway with ARM to get the optimized compilers in place. So we are targeting higher performance, 10 to 15% higher than what we are seeing here with the optimizations. So specific HPC workloads, it may be weather or computational fluid dynamics or the EDA, other workloads. We picked some key workloads here, some applications and also some of the mini apps like clover leaf and tea leaf. And then we ran it on Thunder X2. So Intel is normalized to 1x and then if you look at NAMD, we're on par with Xeon gold. And then if you go look at some of the mini apps, tea leaf, we're actually better than Thunder X2. So this is a key metric as we move along. As I mentioned, there's a lot of work done already and some of our team at Cavium is closely working with folks at ARM to further optimize the libraries in the compiler. So we're trying to increase the performance even higher. So that completes my presentation. Overall, it's a big show for us and we are very much focused in the HPC space. And as you have seen, some of the key announcements have come in this morning from our OEM customers as well as the partners and the announcement that we have made. So we have significant presence at SE. Obviously, this is the presentation at the HPC user group. And we've also multiple presentations at different venues at the conference. And then you can always reach us at cavium.com for any further information. Thank you.