Intel Corporation (INTC) Earnings Call Transcript & Summary

Hello, everyone, and we are live. Thank you for coming. This is the oneAPI DevSummit for AI and HPC in 2022. And I'm really happy to have you all here. So to start this off, I'd like to first introduce myself. My name is Sri Ramkrishna. I'm the oneAPI Community Manager, and my purpose here is care and feed and grow the oneAPI community. So a little bit about myself. I've been working on technical communities for over 25 years. I really, really enjoyed working with communities, technical communities of all types. And it's one of my big passions. My technical background, I was a systems administrator at Linux System Administrator for many, many years, maybe about 30 years. Just to dig myself, I started working on my first Unix computer in 1986. So -- and I subsequently got in trouble. So don't do what I did, and act at things you should. So with that, I'd like to introduce my cohost, Susan Kahler. Susan, when I come up on the virtual stage?

Hey, everyone. Greetings from Raleigh, North Carolina, where it is a cloudy 51 degrees Fahrenheit here. I'm Susan Kahler, and I'm on the AI product marketing team for Intel. Now I've been in the AI space for several years now. I'm not going to date myself like Sri, but I started with my research studies on intelligent agents. We know it's Q4, and it's a busy time for all of us, and we really appreciate that you chose to spend time with us today. I hope that you enjoy the DevSummit. Happy to see everyone in the chat session. Back to you, Sri.

All right. All right. Thank you. So I'm going to -- before we start off with our headline speaker, I'd like to run through a few slides. Make you comfortable with the user interface so that you could have a problem-free conference attendance, all right? So I have our initial slides here for oneAPI DevSummit Host. And let me go to the next one. So here you'll see a screen shot of the agenda. If you are -- we're currently on the Central time zone, and if you want to change the time zone, you have a little spot here that you can change your time zone. You -- if there are specific talks that you're particularly interested in, and you want to add them to your calendar, there's a thing on the side here that says join the presentation and add to calendar. And one important part of this conference is you have to -- once a presentation is complete, you have to exit out of the presentation and then go back to the agenda page so that you can get to the next presentation. So I think that's really important so make sure to keep that in mind when you're joining -- when a session is done, that you need to close up and come back in again. All right. Great. If you have any questions, of course, there's chat, and where we'll be happy to be online and help you out if you have any problems. All right? The next one is, you're probably looking at this right now as you're viewing this conference. Just to familiarize yourself, you have like 3 active windows that you can move and resize and move around. And the first part here is you'll see the speaker video, that's where you see me, hello. And then if you go down here, you'll see a window at the bottom left, which has 3 tabs. There is a Chat, there's an Abstract and there is Speaker Info. So if you look -- if you click on the Speaker Info right now, you'll see the information for both Susan and myself, probably a little more extensive than what we've said online. And then there's a -- on the chat, there's a chat box, and this is the important part. This is where you are able to comment and ask questions, and those when the speaker is talking, you can ask -- put the questions there, and we'll make sure and monitor them and make sure that the speaker sees those questions and is able to answer them. As well, if you happen to certainly forget what the talk is about, there is an Abstract and Speaker Info in the middle tab. So hopefully, all of that is clear to you. At the bottom, there are some quick links that you can look at. If you hover over them, you'll see what each of them do. If you have problems, especially there is a little window with a question mark, that's your tech support. So if you have problems there, you can take a look at that. To the far right, there's a close captioning for those of you who need that. And I think pretty much everything else is self-explanatory. Of course, if you have any questions, just throw it out there on the chat, and we'll try to help you out. All right. Now the next one, moving on to the next slide. So it's -- this is a good opportunity to earn badges. And if you look here, there's a certain number of badges you can get, there's a free to board. So if you see the badges, you can get an easy 50 points just visiting the Resources tab. So if you go out, there's a Resources tab, easy 50 points there. There's a various other things, Collaboration Hero, Session Attendee Champ, so forth. They're not overly hard to do, but it's a great way to get points and sort of have a reward and compete. So hopefully, you have fun with that. I know we would love -- we'll be monitoring the [ badge reward ] and watching all the fund there as well so. Let's see what we are here. It's we got in under 7 minutes, so make sure I go a little slower here. So I mentioned the resources area. And so if you look here, there's resources. Sometimes we're going to be talking about things like DevCloud or AI toolkits and whatnot. And this is the information that's kind of like a library, and you can find out all the things about like the Intel DevCloud for oneAPI, if want to know what that is. There is also a community forum for accelerated computing, if you want to find out more about that. So these are just kind of a sample list of things if you want to know what those resources are. Then moving on. So no conference is without our social -- I mean this is one of the most important parts of community and being at conferences is the social aspect of it. And we are here to -- we have a social aspect, and we do want to give out some prizes and just sort of build a feeling of community. And so if you -- what's important, of course, is to really show that you're having a great time. And one way to do that is to call it on, on Twitter. So if you're on Twitter, include the #oneAPIDevSummit and throw it out there and you get your opportunity to get like a $20 gift card out there as a price. So you do have to be 18 or older to do this -- to do the social. So make sure you're an adult and post it out there. I also -- if you're also a Mastodon fan, feel free to put it on Mastodon. I'm a big Mastodon fan. And if you're on there, I know there's a bunch of folks on HPC who have their own server. So I would love to see oneAPI DevSummit on Mastodon, and let's see what you guys got, all right? And of course, you want to know more about the social, there's people out there in Discord. And we encourage you to come over to Discord after the end of the conference. Now -- so we have a really exciting agenda today. And so just to call it out, our headliner is Andres Rodriguez and he's going to be talking about AI software and hardware acceleration. Then we have Peter Ma, who's doing -- going to talk about using oneAPI actually in a business context. And there's a short break. Then we have Kurt Keutzer and Zhen on efficient inference and training, and a host of other things. We have a quick lunch break, [ pretty much ], a munch break, and as you can see, there's a whole bunch of great stuff. Make sure you catch Rod Burns' talk, if you're interested in oneAPI community forum, which is going to be an important milestone who oneAPI, especially as moving to an industry standard. So after that, there's a hands-on with TensorFlow, and then we have a special guest, Stephano Cetola, who's coming from RISC-V to talk about accelerating the future of heterogeneous compute. And finally, we have a conclusion and Susan will come and finish the day one of the oneAPI DevSummit. And finally, we have Russel who's going to have a happy hour with the rest of us. So a really fun -- fun packed day today. And again, as Susan said, you can spend your time however you want today, but you're spending it with us, and we're really grateful that you're willing to share your day with us. So... Day 2, we'll talk a little bit more about the day 2, but this is generally the HPC portion, the Day 2. If there's anything I want to call it, I want to really call out the live demo showcase that's going to be there. We have a lot of previous DevSummits. Everybody has been telling us like we want to do more hands on, hands on, hands on and hands on. And we've heard and relisting, and we want -- we would love to have you attend our demo showcase here, so want to call out. But we'll talk about more about all of this the next day for -- on the agenda for day 2. All right. I only got 3 minutes here before we introduce the speakers, so I'm going to go through this quick. Don't worry if you don't see this during the conclusion. We're going to show these slides again. These are questions to the DevCloud. This one is oneAPI meet up community. And we have a meet up at DevCon that I do every 2 weeks talking about various topics around AI and HPC and oneAPI. And of course, you can always hang out with us in Discord even after the conference is over or during the conference, however you want to spend your time and there's the way to get to that. We have a happy hour, and you can join us here to have games and prizes and you get a chance to win some stuff. So hang out with us for that. And let me now introduce our headliner, Andres Rodriguez. So Andres Rodriguez is our -- is the Intel Fellow and Chief AI architect. He actually designs deep learning solutions for Intel's data center customers and provides technical leadership across Intel for deep learning hardware and software products. He was the lead instructor in Coursera course, Introduction to Practical Deep Learning to over 20,000 students. And is the author the popular deep learning -- of the popular Deep Learning Systems book. So -- and, again, to remind you, if you have any questions throughout the presentation, put them in the chat box, and Andres will answer them at the end of the presentation. And I'm now going to hand it over to Andres. Please give a warm welcome to Andres Rodriguez, Intel Fellow, a Chief Architect.

Thank you for the introduction, and it's a pleasure to be here with you today. Thanks for making the time. I want to share with you what Intel is doing to accelerate artificial intelligence applications through both hardware and software innovations, powered by oneAPI. So you can see the agenda of the content. First, I want to share with you a quick overview of why AI is important, and then talk about the various products that we are bringing to market, both in hardware and software to accelerate artificial intelligence. And finally, I'll conclude with some of the ecosystem programs that we have. So why invest in AI? So as many of you know, AI is rapidly growing across many different sectors. It's used in language applications. It's used in -- to create images, to detect objects, to detect persons. It's used for recognition. It's also used in the medical field among many other fields. There are multiple articles that are published about the AI goodness every single day. And so what is Intel doing to help developers and data scientists accelerate their AI workloads? So our goal is to provide solutions that are simple to use, that are performing and that increases the productivity of the developers. And to start with, we are developing -- we have multiple hardware platforms that are available for our developers, starting with our on Xeon, Intel Xeon Scalable processors. In our Xeon, Intel Scalable processors, we're adding AI acceleration. We're also bringing to market the Intel data center GPUs that can be used for both AI as well as HPC applications. And for the dedicated AI training workloads, we're bringing to market the Habana Gaudi2 dedicated AI processor. And at the fruit of this hardware, we have oneAPI. And oneAPI has two components. One is the oneAPI specific -- the open specification that is available, not just for Intel, but also for non-Intel hardware and it provides a standard that many hardware vendors can use. And so you can take a workload and you can accelerate it through oneAPI, and it can run on various hardware back ends. In addition, we have a oneAPI product to accelerate Intel's hardware platforms. I'll talk about more details of how this applies to AI in the next few minutes. So to give you an example, here you can see the developers are writing applications using various AI libraries like TensorFlow, PyTorch, Scikit-learn et cetera and these libraries then leverage our oneAPI products. So for example, TensorFlow can leverage oneDNN, Scikit-learn can leverage oneDAL. And these libraries, they are optimized to run across Intel's various hardware backings. So as a developer, you don't have to worry about knowing all day a hardware details of our CPUs or our GPUs, but you can just leverage the high-level libraries like TensorFlow and Scikit-learn, I know that those are going to run with high performance across our hardware back ends. So to give you an idea of the end-to-end software and solutions, that we are offering. You can see on the left-hand side, some of the popular tools that AI developers use for data analysis. In the center, you can see some of the machine learning libraries. And on the right, you can see tools to increase productivity. And all these tools, or most of these tools are leveraging oneAPI acceleration and can be executed across most of Intel's hardware platforms. So in -- particularly in artificial intelligence, we see that models are growing at a fast pace. And they are not just growing in size, but they're growing in complexity. So often, you used to see the models just been a bunch of sequential layers, but now they are growing in much more complexity where the layers are not always just sequentially. Sometimes you have a mixture of [ craftware ] networks, mixed with the traditional multilayer perceptron. And this complexity can be hard for both software and hardware. So what are we doing to meet the demand of this growth in AI and its complexity? So at the center, we have the Intel Xeon Scalable processors. Now the main benefits of this product is that it's widely available. It's simple to use, simple to program, simple to debug. Each core in the Intel Xeon processors is fast. It operates with a high frequency. We have a large memory capacity so that you can store very large models and data sets. The software is robust. We've been working on the software for several years to make sure that a broad set of AI workloads run efficiently on the Intel Xeon Scalable processors. And you can do all both the -- you can do the entire end-to-end on the same processors from the data preprocessing to the training or the inference of your AI models and the post processing. So on Xeon, we have added various hardware acceleration from AVX512 to Intel Deep Learning Boost, via an instruction called VNNI. And more recently, to the next generation of Intel Xeon processors that we're going to be launching in a month. We are the Intel Advanced Metrics Extensions, or Intel AMX, and I'll talk about what this is in a little bit. And in addition, we've worked with the ecosystem community to optimize popular libraries, TensorFlow, PyTorch, ONNX Runtime , XGBoost et cetera., so that you can take these libraries and run them across our hardware -- and you may not even realize that you're using Intel's optimizations. So Intel Xeon processors, as I mentioned, can be used for both the data preprocessing for both the model development, both for traditional machine learning as well as for deep learning as well as for the deployment. And in fact, the overwhelming majority of deep learning inference happens on Intel Xeon processors. Just to give you an example, we work with several companies. So eBay published or give a talk in which they showed the acceleration that they were getting by leveraging Intel's hardware and software. So they showed that for their ranking algorithm, which is important to show their end users more relevant research results that were showing around 2.5% improvement in both speed and throughput. Tencent, they have an application for text-to-speech, not just Tencent, but many other companies, of course. And they leverage for the vocoder accelerator, which gives higher quality speech synthesis. And they were getting 4.7x improvement in performance. Also for the -- for other applications like reinforcement learning, another team at Tencent, they use this for a very popular game called the Honor of Kings, and they were able to do distributor training. So they took 16 clusters of Intel Xeon Scalable processors and train their reinforcement learning algorithm on these processors. And they show that they were able to get near linear scaling across the 16 processors. So you can use these processors, not just for inference, but also for training and distributed the training workload across multiple processors to reduce the time to train. So in the upcoming generation of Intel Xeon Scalable processors that we're launching next month, we either specialized hardware AI acceleration to every single core, and this expands the reach of the Intel Xeon processors into the accelerator space. So that for many applications, you can do both the training and inference on an Intel Xeon CPU. And so this accelerator is called the Intel Advanced Metrics Extension, and it has two components. One is tiles, which are essentially two-dimensional registers as well as TMUL, which is a tile matrix multiply. So essentially, you're taking a matrix multiplication accelerator and embed it into every core of Intel Xeon CPU. Now as many of you know, metrics multiplications are one of the most compute-intensive operations in artificial intelligence and is a very common operation. So by accelerating this computation, you end up accelerating the end-to-end workload. To show you some results, so if you compare the performance of Intel Xeon processors in the upcoming generation with the previous generation, you can see 30x improvement by leveraging both oneAPI with software acceleration and hardware acceleration. And so in this slide, you can see by leveraging into TensorFlow flow with oneDNN, which is one of the oneAPI products, you can get a 50% improvement in performance. And then by leveraging the hardware acceleration of that offer in that generation, in the third generation, you get an additional 3.9x performance boost. And then by leveraging the advanced metrics extensions in the upcoming generation, you get another 4x improvement. So this is something that I am very excited that our developers can use their Intel Xeon CPUs for applications that in the past often required more of a dedicated accelerator. So how does this compare to an NVIDIA GPU? So you can see, you can get much faster performance on an Intel CPU by leveraging both the hardware and software accelerator over an NVIDIA GPU. So this is a popular benchmark called ResNet-50, and you can see a significant improvement in performance. Now we'll have more proof points and examples when we launch the product next month. So I won't share many more, but I do want to share one. But before that, what can you use the Intel Xeon processors for then? So for all inference workloads, for small, medium training models as well as for fine-tuning or transfer learning of large and small models. In addition, all your traditional machine learning. So when you're leveraging XGBoost, Scikit-learn those are well optimized for Intel Xeon processors. And even if you're doing a very large Deep Learning training model, you can leverage multiple Xeon processors and distribute the training workload across multiple processors and reduce time to train, So even for large language models. So this is an example of Alibaba. They did some initial testing and then they were excited to share that they were getting a 15.9x performance boost by leveraging both the hardware and software acceleration in this upcoming generation. So one of the -- one of the things to keep in mind is this acceleration works for lower precision. So for a 16-bit precision, specifically a numerical format called bfloat16 as well as for 8-bit precision, so specifically for int8. And often, particularly for int8, strong developers have struggled in the past, taking a model and knowing which layers they can convert to 8-bit precision because some models are sensitive to a loss of accuracy when you moved them to 8-bit. And so some layers in our Deep Learning model can be quantize to 8-bit while others should be kept at higher precision. And it often takes a long time to decide or to know which one is to quantize versus which one is to keep at higher precision. So to help our developers, we developed an Intel neural compressor, which is a tool that you essentially take your Deep Learning model whether it is a TensorFlow model, a PyTorch model or a ONNX Runtime model, and then you put it through the internal compressor, and the output is a compressed model where all the layers may be quantize or only some of the layers may be quantized. And it keeps any layer that is sensitive to loss in accuracy are the higher precision, so that you can get the acceleration from using lower precision for the layers that can get that acceleration and maintain the high accuracy for the layers that would be -- sorry, the high numerical format for the layers that would be susceptible to a loss of accuracy. Now this tool is super simple to use. We've worked with the [ Stern Lab ] to help them adopt it, and they showed that they were getting 10x productivity. So rather than them trying to find which layers they can quantize, they just use the Intel Neural Compressor and they were able to get an acceleration with a very little effort. So again, one of our goals is to increase the productivity of our developers. So one of the key oneAPI products that we use in Deep Learning is called the oneAPI Deep Neural Network library or oneDNN. And previously, we used to call this library MKL-DNN, but now it's renamed as oneDNN. And it is included into the default versions of TensorFlow and PyTorch. So if you were to do a piping [indiscernible] today or similar with PyTorch, the binary includes oneDNN acceleration. So you can take the default versions and brand your deploying workloads on Intel's processors with oneDNN goodness. And you may not even realize you're getting the oneDNN acceleration because it's included by default, you don't have to change any parameters in your models. Now that's true if you're using some of the recent versions. If you're are using some of the older versions, then there are some parameter settings that you have to set. Some are included here, the versions where oneDNN becomes default into this library. So you know if you're using another version, you can come to our documentation page, which I'll show you the link towards the end of the presentation and set the right parameters as shown in our documentation. But again, if you're using one of the recent versions, then there is nothing that you have to do on your end. Now TensorFlow, for example, by adding oneDNN acceleration, you can get a 3x performance boost for a number of workloads. And oneDNN works not just for Intel Xeon but also for our Intel core products, finding your laptops as well as for our GPUs. And even some of our competitors are contributing to oneDNN such as ARM, so that you can -- for some workloads, you can leverage oneDNN on non-Intel platforms. Now we're for -- with developer community of TensorFlow, which is primarily led by Google, and we have a very great relationship with them. They asked Intel to take ownership, for example, of all future windows build of TensorFlow, which we're doing, and we continue to support, of course, the Linux builds as well. Now oneAPI doesn't just accelerate Deep Learning workloads, it also accelerates traditional machine learning workloads, both in the data pipeline, the data analytics and data preprocessing pipeline, such as with Pandas. And with Pandas, we've introduced a library called Modin that can accelerate Pandas' applications, which is one line of core as shown on the left-hand side. Similar with Scikit-learn, we have introduced a library called Scikit-learn extensions and by adding the two lines of core that are shown in the middle of the slide, you can get significant performance boost, so 38x performance boost in your Scikit-learn applications. And with TensorFlow and similarly with PyTorch, you don't have to do any changes in your core and you can get the acceleration that oneDNN provides. So let me tell you a little bit about our Intel GPU's discrete accelerators. So we've recently introduced what we call the Intel data synergy Max series. And this is a GPU that has built-in AI accelerator -- acceleration into every core. So to show you some of the performance that you can get and support across various numerical formats, you can see here that, for example, if you were to use a popular 16-bit precision, you can get 839 teraflops, which is a significant performance for both streaming and inference. And you -- if you can do inference with 8-bits of precision and that doubles the available compute. Now to fit all that compute, you need a well-structured memory hierarchy, which our GPUs provide. You can see the high bandwidth between cases and the large case sizes our Intel GPUs have. And this is -- we're bringing this to market in different platforms from an OEM through a subsystem with 4 GPUs to one with that's built in with our fourth generation Xeon scalable processors. And this is going to be available through multiple OEMs. I'm showing some of them here in the slide, from HPC, Dell and others. Now the last hardware product that I want to briefly mention is the Intel Habana Gaudi. We recently announced a Gaudi2 accelerator, deep learning accelerator. Now this accelerator, we competed in [indiscernible] benchmark, which is a benchmark dedicated for the deep learning workloads. And you can see that on the left-hand side, the blue is the Gaudi2 result, and it's bidding A100 for both BERT and ResNet-50, comfortable bidding ResNet-50 for NVIDIA A100, the results are using different performance. So the results are not necessarily apples-to-apples, but they are with the A100. And this is something that we're very proud of and to share that product with our developers so they can reduce the time to train significantly by leveraging the Habana Gaudi2 accelerator. And in addition, one of the main benefits is not just the raw performance, but the much better cost, and I'll show you in a few slides, the significant cost advantages that this product has over the A100 from NVIDIA. But just to give you a high-level architectural view, this is built on 7-nanometer processors. And one of the things that -- one of the many things that is a -- some very strength for this product, it's not just the performance of single Gaudi2 accelerator as was shown in the previous slide, but also the ability to do distributor training across multiple Gaudi processors. So we have [ 24 times ] 100 gigabit DRAM mix. So that you can take a very large model and you can train it across hundreds or even over 1,000 Gaudi processors. And in that way you can reduce the time to train and the scaling across the processors is nearly linear because of the high bandwidth between processors. Now this is available through various OEMs like Super Micron, through Inspur and through Vivin. Now how to leverage the acceleration? It's quite simple. It only takes, for TensorFlow remodels, just 2 lines of core in order to tell the model to leverage the Habana acceleration. For PyTorch, it takes only a few lines of core, again, just to tell the model to leverage the Habana accelerator. Now if Intel were a Vertex for [ imported ] performance. If you go to the Intel Habana GitHub page, you can see a number of performance that we've measured and that you can produce across various versions of TensorFlow, PyTorch and across many models. And so you can tell all the models that have been well optimized, and the performance is not just limited to the models that we're showing. The performance coverage has a much broader set of models. But this can give you an idea of the performance for the types of models that you may have. As far as the cost, this -- the first generation is already available at AWS. And so if you were to leverage AWS DL1 instances, which are the ones that have the Habana accelerator, Habana Gaudi accelerator, you can see -- on the left-hand side, you can see the Gaudi cost is much cheaper than both A100, the high-end GPUs, the 80 gigabytes, both -- and also 40 gigabytes as well as AV100 for a number of models. So again, in addition to the higher raw performance of Gaudi2, I think the main benefit to our developers is going to be the cost savings as well as the ability to distribute a model across hundreds of [indiscernible] Gaudi instances. So this is something that Sundar, the Head of Machine Learning at AWS highlighted the nice advantages and nice things that developers can get by leverage in these instances, and highlighted that for -- that you can take a [indiscernible] model and you can do inference with 15 milliseconds across multiple per large models. We work with the community as well to -- with many companies to make sure that they can take advantage of Habana Gaudi acceleration. So we work with Lidar , which is an R&D company, they were doing COVID research and time to train is important for them. And they saw significant cost savings, over 60% by leveraging the L1 instances over the GPU instances. We work with Mobileye. They saw over 40% in cost savings for training their models. And Gaudi cannot just be used for training, but also for inference. We did a fun demo using the efficient models. So these are models where you pass a text description and it generates an image based on the text description. You can see on the right-hand side, the performance of Gaudi2, where lower is better. You can see across various workloads that Gaudi2 comfortable outperformance NVIDIA A100. I mentioned one of the key advantages is this ability to do near linear scaling across hundreds or even over 1,000 Gaudi instances. And we, in partnership with Microsoft Research, have done various examples where we're leveraging the Gaudi processors. So we did some training with 512 Gaudis for multimodal understanding. We've seen some of the latest transformer-based models as well as using some diffusion -based models. So we are not just working with companies, but we're also working with academia. We want the developers that are pushing the state-of-the-art developing new models to be able to leverage the acceleration because we believe that this is going to accelerate the state-of-the-art. One of the popular companies that is at the forefront of transformers, it's called Hugging Face. And so we partner with Hugging Face to make sure that the optimizations that are offering the optimum library has acceleration for both Intel Xeon Scalable processors and Habana Gaudi processors. So if you go to the Hugging Face to the optimum library, then you can leverage Hugging Face models accelerator for both Xeons and Habana Gaudi processors. You can get access to both the latest generation of Intel's Xeon scalable processors as well as the Gaudi2 processors as well as our Intel GPUs through our Intel Developer Cloud. And if you do a Google service for Intel Developer Cloud, it will bring you to the page and you can request access to run a number of your experiments. So we work with the various -- with many ecosystem partners. We have 2 programs that I just want to quickly highlight. One is called the Intel Disruptor Initiative. This is where we partner very closely with a number of companies that are pushing the limits of innovation, and we provide them support -- technical support so that they can more rapidly innovate on Intel's platforms, both software and hardware. And for the larger ecosystem, we have the Intel AI builders. We partnered with many, many companies and you can see a few of them in the slide. Across various vertical, it's both like retail, health care, FSI, et cetera, and horizontal partners. So if you are a developer in one of these -- or in a particular company, we'd love to partner with you. So what's next? We want to invite here to visit our developer website at developer.intel.com/ai and you can see there, all the documentation for oneAPI and how you can leverage oneAPI for AI. You can learn more about the work we've done on TensorFlow, PyTorch, Scikit-learn, Pandas, et cetera. and we are excited to partner with you to accelerate your workloads and give you additional productivity. So thank you for your attend today. I'm happy to take some questions.

Hi! Andres, this is Susan. Can you hear me okay?

Yes.

All right. So we have 3 questions that have come in during your presentation. So the first one is from [ Shreayns ] and [ Shreyans ] wants to know when you were on the slides where you were talking about the fourth gen Xeon, the question is would this accelerator removes the need for any other hardware accelerator? We are using TPU, GPU?

Yes, that's a great question. So let me see if I can go back to the slide, and I don't know if the audience can see it. But essentially -- okay, thank you. So essentially, the acceleration built into the Xeon Scalable processors does expand into the accelerator space. So I would recommend that developers start with Xeon. Because I think -- well, I'm confident that Xeon will be able to cover the majority of your AI workloads. So all your traditional machine learning should be done on Xeons and most of your deep learning workloads. Certainly, all the inference workloads and the small training models. Now if you are a developer that is constantly training very large models, then it makes more sense to have a dedicated accelerator. And that's why we have both a Gaudi2 scalable processors -- sorry, the Gaudi2 AI processors and our discrete GPUs. So we do see a need for those -- for those products, of course, so -- but we see Xeon being able to meet the majority of your AI needs.

Okay. I have a question -- sorry, go ahead. Go ahead, Sri.

Yes. Yes. So we had another question, and it was -- I'm sorry, I forgot to copy the name. It's from -- [indiscernible] -- it went sack while I -- it's from [ Stefan Tran ], and he says, "Hi, Xeon acceleration seemed very promising. Is there any notable hardware or software acceleration for the core CPU families that would have a significant impact on the training and deployment? Any tips and tricks to use the most of the core CPU family?"

That's a great question. So we are adding hardware acceleration to our core. I cannot share all the details at this time. But we have added software acceleration. And so even though I highlighted all the goodness on the fourth generation of Intel Xeon Scalable processors, both previous generations of Xeon processors as well as our core CPUs that is -- that are using client and some work stations, those can leverage from the software acceleration. So if you aren't -- seeing some older version of TensorFlow or PyTorch, I would encourage you to try the newer version because we're constantly adding more and more acceleration to all these products. Many developers are leveraging older CPUs and [indiscernible] develop on client that don't just go straight to servers. So of course, we want the developers to take advantage of the performance boost. And so oneAPI, DNN, our deep learning library for software acceleration is not just targeting the latest hardware, but it's also optimized for other hardware products or other hardware CPUs so that you can get a performance boost. But again, I would invite you to use one of the more recent versions of these libraries, TensorFlow, PyTorch or XGBoost?

And Andres, we have one more question. This is from [ Yusaf ], who wants to know, is Habana available for Julia via the oneAPI library?

Yes. So Habana is targeting TensorFlow and PyTorch. And as long as you are using those two libraries, then you can take your workloads and run them on Habana Gaudi processors. There is no support for other libraries outside of TensorFlow and PyTorch.

Great. So I've just asked, Andres, I've just posted the chat to see if there's any more questions. So if you'll stick around for a little bit longer, just let me see if anybody comes up with any more questions, please. Okay. So we do not -- I don't see any more question appear in the chat. Over to you, Sri.

All right. Well, Andres, since there are no more questions, we want to thank you for coming to the oneAPI Summit and presenting for us. And if there's -- if there's any chance, if you want to hang on to Discord at the end of the conference and answer any questions. That would be fantastic. Other than that, this concludes our presentation. And so please, as of our [ core ], close this session and then joining us for the next session starting at 10 p.m. -- 10 a.m. sorry. I'm not always confused about the time, but I guess it's just in my mind at p.m. So I will -- we will see in the next time slide. Thank you for attending this session.