 So we hear the super computing show here in Denver and you have a whole bunch of new announcements, right? Yes. So now Thunder X2 is about to be shipping? Yep. So let's see some of the systems that you're showing off. Yeah, so let's start from here. This is the platform that HPE Hewlett-Beckard enterprises announced at the show actually yesterday and this is part of their Apollo family. Apollo is their high-performance computing platforms and this is called Apollo 70. It's a 2U 4-node high-density platform. This is one of their volume products and this has two Thunder X2 processors with eight memory channels per socket, delivering about 33% higher memory bandwidth and 33% higher memory capacity compared to the computing solutions and you can actually fit four of them here and this delivers very high core count and very high memory density to the end user. The high-performance computing applications that actually benefit from high memory bandwidth and higher core count you know are going to be able to really see the performance differences. There were a number of performance benchmarks announced by our customers HPE as well as Cray and some of the national labs at this conference and it's actually really exciting. So the benchmarks that's the big deal to announce this right? Yes. And 33% better than Intel? Intel, yeah. On the memory. What Intel? The best Intel? Yes, the top Intel. The top Intel. The platinum? Platinum. It's the you know they have six memory channels we have eight and that gives us higher memory bandwidth. So you have higher memory bandwidth is one of the most important things to have high memory. Yes, for high performance. How high, how many gigabytes you do? So the measured performance that we are showing shows about 132 gigabytes per second in memory bandwidth and this is a number that our customers are sharing with their end users and that's about 33% higher than the high-end Skylake platforms. So having a higher memory bandwidth helps with also being 33% more efficient? It helps with a number of applications in the high-performance computing space because they need higher memory bandwidth and we are showing performance benchmarks for those. And HPC, this is a supercomputing conference but the stuff you do is not only for supercomputing but when you announce stuff here it means everything is available for the whole market. Yeah, of course. So we play in the cloud and HPC space so the platforms that we are showing are of course you know for the high-performance computing market but then HPEs also sell similar platforms for the cloud. And that's for example right here Google Cloud, Amazon Cloud, all these customers should potentially use these kinds of platforms. They can buy this now. Of course. And let's look at the other one over there, it's a big one. So this is the platform from Cray. Cray is a very famous supercomputing company. They invented supercomputing. So Cray has a very high-end platform called XC50, this is their flagship product and this is a blade that goes into XC50 platform. It has 800x2 processors. 800x2 processors as you can see here and it ties into their fabric through Aries which is Cray's fabric which provides very high bandwidth and very low latency for the supercomputing applications and this was launched along with a number of Cray customers quoting you know about how great this platform was. So you have eight chips. Eight processors. They work perfectly together. Yes. How do you do it? How do you design a chip that can work together? Cray has a fabric that interconnects these processors. We have dual socket processors so they use four of those dual socket processor blocks in here to deliver the net performance. And lots of RAM. Lots of memory. Yes. And this is their network. This is their network. So this connects to, this connects those processor complexes also ties them into their cluster. Okay so when Cray ships a product they actually ship an entire what they call a cabinet and the cabinet consists of lots of these processors, processor cards. So potentially maybe this one could be number one in the world supercomputer in this year. This is going to, well this is something which is like top 500 class supercomputer right. So Cray has a number one position in terms of the total number of systems in the top 500 at the top 100 or top 10 and you know with the introduction of something like XC50 this really helps us to play in that space. So potentially is it going to be your solution in ARM that could be in the top like beat the Chinese? That I don't know. Of course. Of course. That I don't know whether it can beat the top you know top 10. It's a question of you know customers now have an option to deploy ARM supercomputer which they didn't have before. So this is a volume HPC play. Volume. Okay so this goes to you know high density, high performance, a platform. This one is the real you know scientific academics, labs, national labs, yeah. NASA. Yeah those guys. There was something awesome also at the HPE they were showing the machine. Yes. Are you excited by this? Yes. Yes so machine the only processor that actually supports the Gen Z today with the machine is Cavium Thunder X2 CPU. So the beauty of that is Thunder X2 has a coherent interconnect and they have a chip that connects Thunder X2 into Gen Z fabric. Okay and that is great because now the memory actually can be remote and multiple Thunder X2 blades can actually share the memory across them. So this is what HP calls memory centric computing and this is one of the largest R&D projects in HPE's history. This was launched in May of this year and we're very excited to be a part of that. So you're part of the Gen Z consortium. Yes. Yes. Maybe you could do one thing I wonder is because you have so many things in the SOC but some you have networking too right? Well you have a network in there but you can also have it externally. So this so our networking is industry standard networking right so you know for high performance computing they want ultra low latency interconnects and the Cray fabric actually enables ultra low latency interconnects into their cabinet. So it's great because it gives customers in the high performance computing space a choice in terms of you know having a very high performance interconnect which allows them to have higher scalable performance for their applications. Over there is this the Gigabyte? So this is the Gigabyte channel product this they announced this at Computex and now they're actually launching this into production. This uses the BGA version of Thunder X2 Silicon. These are all LGA so they have socket socket so you can actually just like X86 you can replace the CPU right? Yeah. So when you go to a BGA version if you're a cloud customer and you really know exactly what you want you can actually have a lower cost platform in BGA. There's no it's not a socket you don't need to take it out. You don't have to because you know now that way you can save the cost on the motherboard you can save the cost of the socket so if you know exactly what you want you can actually you know just socket the part and that actually saves you money. Gigabyte did an interesting thing over here right? You can add a whole bunch of... Yeah so they have a they have a riser card so they take advantage of all the PCIe lanes we have in the chip so they can actually support lots of different kinds of PCIe cards including full-height adapters and this platform can be used for storage because it has lots of drives. And but we're here at the supercomputing this is not necessarily a supercomputing product right? Oh it could be. It can be okay so this is more like a traditional 2U 2 socket server which is which could be cloud it's like the workhorse of computing right? One of the most popular form factors in the server market is a 2 socket server. Super computing has more high-density platforms which are targeted to more compute but this one gives you storage and compute. So this one compared to an Intel solution it's like 2x better or how do you compare? So it depends on the application and you know what we are showing is I think we talked about memory bandwidth we are showing integer performance we're showing floating point performance we are showing some of the HPC application performance okay and you know we are actually very very competitive relative to the latest Intel processors. There's even some other announcements that have been here to show. It's been very very busy show right? Yes it is. So we have this platform this is from Foxconn Ingressus and it's the same kind of 2U 4 node system and these are more cloud oriented platforms you know some of the cloud guys buy directly from the ODMs they don't go through OEMs so they have platforms that they can get these are all tier one ODM partners so we have announcement from them we also have announcement from our partners for example MicroSemi announced support of their RedAdapter which goes into this platform for storage so we have had a number of announcements from our partner the other exciting thing is that RedHat actually announced production support for the HP platform over there right? This is the first time RedHat is actually announcing production rel support for ARM. And that's a big deal? It's a very big deal because lot of the enterprise customers and lot of the HPC customers actually use standard RedHat and it gives customers a choice in terms of you know running exactly the same set of applications that they used to run. The other exciting thing is that Bull who is a European company won an award for the best collaboration with their Mont Blanc project for the Sequenoblade which is based on Thunder X2 ARM CPU so they won HPC Insider Award. We're very excited about that they have a liquid cooled system and they're going to talk more about it at the birds of feather session tomorrow. It sounds like crazy crazy busy right now. Yes it is. Yes it is. It's exciting. Thunder X1 was great but it was like a stepping stone for what's happening now right? Exactly. Exactly. Now you see all the OEM platforms you see broad ecosystem we are showing performance which used to be you know so and not just us our customers are showing performance on this platform and if you notice at this show it's our customers who are actually talking about their products it's not Cavium talking about our product and that's how we want it to be. Beautiful companies like Cray showing this amazing stuff. Exactly. And the HPE getting very excited with the machine. The machine could change a lot in the future right? Exactly. It could change like the whole architecture of super computing. Exactly. Well it's not just super computing it's also enterprise computing right because you know by having memory that is shared across multiple compute nodes you actually have a much better utilization of memory and you can actually with storage class memory now you can share that that changes the computing paradigm and we're very excited about our partnership. And there was Foxconn ingress stuff too. Yeah. So that's this one right there. So all the big guys are part of this. Yes. Everybody's excited. And hopefully right there we're going to actually announce that they deploy data centers with this. It's ready. It's ready right? Yes. Our goal is to actually make the platforms ready. Have the ecosystem there. Have the performance there so that you know it makes it it removes all the barriers to adoption of this architecture in the data center. And for high performance.