live from San Francisco it's the cube covering micron insight 2019 brought to you by micron welcome back to San Francisco everybody we're here at Pier 27 the Sun is setting behind the the buildings in San Francisco you're watching the cube the leader in live tech cover jump date Volante with my co-host David Flair we've been here all day covering micron insight 2019 Tommy Vee is here is the senior vice president and general manager of the compute and networking business unit at micron Tama great to see you again great to see you so you got compute and networking two of the big three you're in your business unit there you go but we're gonna talk about 3d crosspoint today but so anyway you know absolutely we're kind of bringing you outside the swimlane or maybe not but tell us about your bu and what's the update yes we you know we sell primarily memory today DRAM although in the future we see 3d crosspoint it's a great opportunity into the the data center you know both traditional servers and the cloud players pcs graphics and networking yes so you get some hard news today why don't we dig into that a little bit we surely haven't covered much of it but okay yeah so I guess you know a couple couple things of interest probably most directly as we we announced our our first 3d crosspoint storage device it's a it's a it's the highest performance SSD in the world and offers compared to other 3d crosspoint based solutions on the market you know anywhere from three and a half to five times the performance on a range of both sequential and random reads and writes two and a half million I ops bandwidth readin right north of nine gigabytes a second and I'm super fast super fast fast and you know similar similar you know a very positive comparisons up against up against me and SSDs ok and so we're excited about that so where's the fit what are the use cases who you're targeting with sure yeah I mean I think you know that one way to think about it is that anytime you introduce a new layer into the memory and storage hierarchy you know historically it was SRAM caches and then it was SSDs going in between dear and rotating media now this is 3d crosspoint sitting in between DRAM and and NAND and and the reason it is a benefit in terms of another layer is it's you know higher density and and greater persistence than DRAM it's greater performance and and you know you can it can cycle greater endurance than the man and and when you do that you do nibble away at either side of that layer so in this case that nibbles away a little bit from DRAM and a little bit from NAND but it grows the overall pie and it's the only player in the industry that provides DRAM 3d crosspoint in and we think that's a great opportunity at some code to the economics cuz it's more expensive than and less expensive than the DRAM higher performance than the traditional flash short lower performance well under the performance of DRAM so yeah I mean so again I think you know the the the you know the benefits like I said is it's it offers greater density and it offers greater persistence than DRAM and so that's the advantage there and it offers much greater performance on things like bandwidth and I ops and much greater endurance than the NAND and certainly our preliminary results are in in applications like databases in certain AI and machine learning workloads and in workloads that that benefit from low latency I think financial service markets is one specific example you know we think there's a good value bro so so a Colombo question if I may yeah so si P would say no throw it throw everything in memory in Hana and of course sell the DRAM and say ok that's ok with us so you mentioned databases how should we think about this relative to in-memory databases sure I mean I think that if if you can afford it and of course it will be more expensive we would love to provide you know our highest density DRAM modules on on the highest end server platforms and you know put put you know you mentioned you know Hana database in the terabytes and terabytes of the RAM that would be great that is is not free if we refer you to do it right exactly and and so if you have the need for that performance that's will do but we we see there's a you know a an attractive range of workloads that cannot afford you know there's a costume that very high-end solution and so this affords something that that gives you know good benefits a database performance but at a slightly more I know you want to jump in go oh yeah sure I compare yourself with Intel which is obviously got the same raw technology they have gone for consumer type obtain [Music] SSDs but they put all their effort into combining it with a DVD or envied him and have combined that with the processor itself and made a combination which is very good for storage controllers yeah so the quest you can very well in in in the SSD much much much more than they have are you looking to go into that and the dim because he obviously you don't have the processes themselves to to to man yeah I mean you know to be clear the you know what we're offering today you know is a product that runs on standard and yeah and via me and while there may in the future be opportunities to further enhance performance with software optimization it runs you know out of the box absolutely without any software optimization and but I do think that you know there are opportunities both to use this technology in you know more of a storage type of configuration and and looking forward there are also opportunities to use it in a memory configuration you know what what what we're announcing today is our is our first storage and with regard to additional products you know stay tuned so if I think about the storage hierarchy you know the the classical pyramid and forget about let's let's focus on the persistent end of that spectrum yeah this is at the tip right is that how we should think about this or not necessarily I mean it is at the storage tip yes but I think we 10 to think a little bit more holistically that you know that that triangle extends from you know from DRAM traditionally to SSDs to rotating and we're now inserting a 3d crosspoint based layer in between and and so from that perspective it is it is the tip of the storage triangle right but it does sit below it does sit below DRAM so in the overall and the reason for my question was sort of a loaded question because if you eliminate the the DRAM piece now you've got that tip sewn and benefits from the volume of consumer thoughts on how you get volume with 3d crosspoint sure you know again I think there are you know at a at a lower performance point you know you can get higher density you know more cost effective storage solutions with that um and we certainly don't see you know NAND going away or we're quite bullish on that you're like man you know it's both a both a SATA and a nvme 96 layer TLC nan based products today so that's that continues to be a major area of investment but you know from a you know from a from a value and opportunity point of view we see a better opportunity you know applying this technology again into this layer in the you know in the in the server or datacenter hierarchy um you know as opposed to what one might be able to do in the consumer space and your OEM say bring it on right I mean they they want this we're talking about the server manufacturers data center yeah I mean I think we're in you know we're in we're in limited sampling with select customers so you know more to say about our go-to-market you know at a at a future date but certainly we we see that there is you know we're we're bullish about the opportunity the marketplace so just asking a question about volume making sure you if you look at the marketplace it's arm has been incredibly successful and it's driven a huge amount of memory and and Nan for yourself then that seems to be where the volume is growing much faster than most other platforms are you looking to use this technology 3d crosspoint as in in in that environment as even memory as in DRAM itself as memory itself at a much lower level I'm just thinking of ways that you could increase volume sure I mean so to be just just to be clear you're talking about what's driven overwhelmingly by by the cell phone market right obviously it's it's proliferating into IOT you know I guess again our our our view of the of the first and best opportunity is in the data center which is still today an x86 dominated world I would say you know in terms of opportunities like I said for you know memory based solutions in the data center um and for how we apply this in other areas you know stay tuned let's talk about this forward next acquisition so it's really interesting to see micron making moves in an AI why the acquisition tell us more about it sure yeah so it's a it's a it's a small small start-up you know handful of players although you know fairly experienced as as I believe sanjay mentioned they're on their their fifth generation of their architecture and so what we've acquired it's both it's both the hardware architecture that currently runs on FPGAs along with the supporting software that supports all the common frameworks the tensorflow is the the PI torches as well as the range of the network architectures you know that that are necessary to support again primarily on the inference side you know are we see the best opportunities in edge in fencing but in terms of what's behind the acquisition first of all there is there's an explosion of opportunity in machine learning we see that in particular on you know on edge inferencing and we feel that in order for us to continue to optimize and develop the best solutions both over all of a deep learning platform that includes memories but also just memories that are best optimized we need to understand you know when you noticed in the workloads we understand the best solutions and and and so that's why we made this acquisition we integrated it with our team that has for some time developed FPGA based adding cards and it's actually the basis of the technology for some of the dialog that used to offer example with OHSU when you talk about edge inferencing we're envisioning this sort of massively scalable distributed system that of course comprises edge you want to bring the compute to the data wherever the data lives obviously don't want to start moving data around now you're bringing a eye to that data which is the data data ai cloud all these superpowers coming together uh-huh so our premise is that the inferencing is going to be done at the edge much much of the data if not most of the data is going to stay at the edge yeah so this is what you're enabling through that integration provision heterogeneous combination of technologies correct I mean you know to use the extreme example that we talked about you know on stage earlier you know CERN has this massive amount of information that comes from the I think it's 40 million collisions a second or I may have my figures wrong and you cannot possibly store nor do you want to transmit that data and and so you you have to be applying AI to figure out what the good stuff is and there's no stream it's exactly and that solution exists in a myriad of applications you know the very you know simplistic one you're not going to send you know the picture of who's at your front door you know to a core data center to figure out if it's somebody in your family yeah you don't want to be doing that maybe not in the camera but certainly a lot closer because you just you know the network simply will not can't handle the capacity all right we got to go but but last word you know what are the takeaways from today what do you want our audience to remember from this event well I think you know I think it's just we continue to build on our memory and storage base to to move up the stack and add values in way that maybe storage subsystems like our our NAND SSD and 3d crosspoint that you know go a little further up the stack in terms of our gaining greater expertise in you know machine learning solutions or or the example with authentic of providing you know a broader solution including key management for how we secure the billions of devices they're gonna be at the edge touching all the bases Tom all right congratulations on all the hard work and it was great to see you again thanks guys Dave and Dave thank you and you keep right there but it will be back to wrap micron insight 2019 right after this short break from San Francisco you watching the cube

>> Live from San Francisco, it's theCUBE, covering Micron Insight 2018. Brought to you by Micron. >> Welcome back to San Francisco everybody. You're watching theCUBE the leader in live tech coverage. I'm Dave Vellante. He's David Floyer, and we're covering Micro Insight'18. It's all about bringing together artificial intelligence and the memory and storage requirements. We're here on the embarcadero. We've got treasure island that way. We've got the financial district over there. We've got Golden Gate bridge behind us. Tom Eby is here as senior vice president and GM of Micron's booming compute and networking business unit. Good to see you Tom. >> Great to be here. >> And Permal Savla is here. He's the director of deep learning at NVIDIA. Welcome. >> Thank you. >> So obviously some of these new emerging work loads require collaboration between folks like Micron and folks like NVIDIA. But Tom why don't you kick it off. What are some of the big trends that you're seeing in some of these alternative work loads that's driving this collaboration? >> Well a lot of what we're talking about here today is the drive of AI and machine learning work loads, and the implications for memory. Certainly there's a host of them, natural language processing, photo and image recognition, applications in medical research, applications in optimizing manufacturing like we're doing in our fabs, and there's many many more. And of course what's exciting for us is that to support those in an optimized way really does require the mating of the optimal processing architecture, things like GPUs. With the right high band width with low latency memory and storage solutions. That's what leads to great partner ships between partnerships like Micron and NVIDIA. >> David was explaining at our open the intensity of the work loads that you guys are serving, and how much more resources that requires to actually deliver the type of performance. Maybe you could talk about some of the things that you're seeing in terms of these emerging work loads. >> Yes, so at NVIDIA, we build systems for X rated computing. AI and deep learning is a very quickly expanding field at this point which needs a lot of CP horse power. What we are seeing is that different applications like you said there's image processing, whether it's video, whether it's natural language processing the amount of data that is there, that is required to do deep learning and AI around it, we break it up into two work flows. One is the training where you actually train the software, and make it intelligent enough to then go and do inference later on. So that you can go and get you results out of it at the end of it. We concentrate on this entire workflow. That's where when we are looking at it from a training perspective, the GPU gives it the processing power. But at the same time all the other components around it perform at the peak. That's where the memory comes in. That's where the storage comes in, and we need to process that data very quickly. >> Yeah, so we know from system's design that you got to have a balanced system or else you're just going to push the bottle necks around. We've learned that over the years, but so it's more than just slapping on a bunch of storage and a bunch of memory. You're doing some other deeper integration, is that correct and what is that integration? >> Yeah, I think the two companies have had a great relationship, just to talk about a couple examples. We essentially co-defined a technology called GEDR 5X, which greatly enhanced the speed of graphics technology. We gently introduced that to the marketplace with NVIDIA about 18 months ago. And then worked with them again very closely on a technology called GDDR six, which is the next generation of even faster technology. We were their launch and ran partner for their recently announced G-force RTX line of cards. It's a very deeply engaged early in the process, define the process, define the standards, jointly develop the solution. Very intimate sharing in the supply chain area. It's a great relationship for us. We're excited about how we can continue to expand and extend that relationship by going forward. >> So obviously there's the two parts of it. You said the learning part of it, and the inference part of the computing. What do you think is the difference between the two? I mean obviously at the end of the day, the inference part is critical. That's got to be the fastest response time. You have to have that in real time. Can you talk a little bit about what you're doing to really speed that up, to make that micro seconds as opposed to milliseconds? >> So from an NVIDIA perspective we build the entire end to end tools steps for training and inferencing. We have a set of libraries that we have made it openly available for all of our customers, all our partners, and all users. So that they can go download it, and do the training so they can use the different frameworks and libraries to accelerate the work that they're doing. And then transform it onto the inference spot. We have something called denser RT, which is basically denser real time. That gives the capability to get these answers very quickly. So on our D4 of the tuning, Chip said that we just announced. We can get a very high performance for our image. So any kind of image recognition or image processing that we need to do, we can do that on the systems very quickly. And we can meet, rebuild entire architectures. So it's not just about one piece. It's about the whole end to end architecture of the system. >> So we heard earlier today in the analyst briefing, the press briefing that Micron certainly in the last 40 years has changed. We're seeing a lot more diversity. Usually it'd be all about PCs. Now there's just so many alternative work loads emerging. Clearly NVIDIA is playing there as well with alternative processing capabilities. What do you guys see as some of the more exciting, emerging work loads that are going to require continued collaboration and innovation? >> Yeah, well I think to build a little bit on some of the other comments about the need for real time inference, one of the things in the area of diversity that we've found interesting. The relationship between Micron and NVIDIA in high performance memory really started around their graphics business. But we are seeing in other markets closer to the edge, in automotive, in networking and in other areas where there's a need for that real time performance. Yet there's also a need for a degree of cost effectiveness. Perhaps a little more so than in the data center. That we're seeing technologies like GDR six being applied to a much broader range of applications like automotive, like networking, like Edge AI, to provide the performance to get that real time response but in a form factor and at a cost point that's affordable for the application. >> Anything you'd add to that Permal? >> So I would also add you talked about applications, different applications that are changing right? Today we announced a new set of libraries and tools for the analytic space. That's again a big work load in the enterprise data centers, that we are trying to optimize and accelerate with machine learning. So we announced a whole set of tools which take in these large data sets that are coming in, and applying it in the data centers and using it to get answers very quickly. So that's what NVIDIA is also doing is expanding on these capabilities as we go in. And as these components and as these technologies get better it just gets our answers much more quickly. >> As exacts in the space and you guys both, you're component manufacturers, and so you sell to people who sell to end consumers. How do you get your information in that sort of pull through? Obviously you work with your customers very closely. >> Mm-hm. >> How do you get visibility to their customers? Just going to go to shows, you go do joint sales calls, how does that all work? >> Certainly some of that is in discussions with our customers and their marketing groups about what they're seeing from a customer point of view. But certainly there's other paths. One of the reasons behind the hundred million dollar venture fund that we announced today, is one of the best ways to get that advanced insight, is to be working with some of the most innovative start ups that understand what some of those end users needs might be and are developing some unique technologies. So there's a range. Working with our customers through eventually finding others, but it's important that we understand those needs because the lead time to developing the solutions both memory and processing architectures is quite well. >> Of course everybody wants to work with NVIDIA, you guys have an inundated like come on oh no we're the most. We're tied up now. Of course there's not a lot of choices here when you're talking about the levels of components that you're selling. But what's life like at NVIDIA? I mean they've been knocking down your doors to do partnerships. >> I think we've grown from being just the component to now being a complete system and an architecture. We don't only just build just a chip that the GPU was. We also build full SLCs. We also build the libraries, software, and the tools that are required to make this complete end to end solutions. We also do a lot of open source technologies because we want our customers and our end cast partners to build and take what we have and go beyond what it's capable of. That's where we end value at the end of the day. Yes, it's all of us together. We need to work together to make that much more faster as we go. >> The tuning is incredibly important. This is complicated stuff. It doesn't just work out of the box, right? So you need an ecosystem as well. >> Yes. >> Yes. >> That's what you guys have been out building. Tom, well give your final thoughts. >> Yeah well I guess to build a little bit. Certainly NVIDIA is moving up the stack in terms of the ecosystem, the software, the complete solution and I think Micron does as well. Like you commented, traditionally it was a component play. And increasingly, we're going to be building subsystems in memory and storage that occurs today on the storage side. I think we'll increasingly see that in memory, and with some of the future, very promising technologies like 30 Cross Point. >> Yeah it's the dawn of the days where everybody just gets piece parts and put them all together. They need you you guys to do more integration, and more out of the box like you say subsystems. So guys thanks very much for coming on theCUBE. Really appreciate it. >> Thank you. >> Thank you. >> Alright you're welcome, keep it right there everybody. We'll be back in San Francisco, you're watching theCUBE from Micron Insight 2018, accelerate intelligence. We'll be right back after this short break. (music)