Learning Library

← Back to Library

GPU Basics: CPU Comparison and Cloud Benefits

Key Points

  • A GPU (graphics processing unit) contains hundreds of cores that run computations in parallel, unlike a CPU’s few cores which process tasks serially.
  • This parallel architecture lets GPUs handle compute‑intensive workloads that would overwhelm a CPU, acting as extra “muscle” for demanding applications.
  • Nvidia and AMD are the primary GPU manufacturers, each offering chips tuned for specific use cases such as virtual desktop infrastructure (VDI), 3D CAD, movie rendering, and AI workloads.
  • In VDI and other graphics‑heavy scenarios, GPUs enable high‑performance, low‑latency visual experiences from remote cloud servers, eliminating the need for local powerful hardware.
  • Beyond gaming, GPUs are increasingly critical in fields like financial services, life sciences, healthcare, and artificial intelligence (machine learning and deep learning), making them essential for modern cloud‑based compute solutions.

Full Transcript

# GPU Basics: CPU Comparison and Cloud Benefits **Source:** [https://www.youtube.com/watch?v=LfdK-v0SbGI](https://www.youtube.com/watch?v=LfdK-v0SbGI) **Duration:** 00:07:30 ## Summary - A GPU (graphics processing unit) contains hundreds of cores that run computations in parallel, unlike a CPU’s few cores which process tasks serially. - This parallel architecture lets GPUs handle compute‑intensive workloads that would overwhelm a CPU, acting as extra “muscle” for demanding applications. - Nvidia and AMD are the primary GPU manufacturers, each offering chips tuned for specific use cases such as virtual desktop infrastructure (VDI), 3D CAD, movie rendering, and AI workloads. - In VDI and other graphics‑heavy scenarios, GPUs enable high‑performance, low‑latency visual experiences from remote cloud servers, eliminating the need for local powerful hardware. - Beyond gaming, GPUs are increasingly critical in fields like financial services, life sciences, healthcare, and artificial intelligence (machine learning and deep learning), making them essential for modern cloud‑based compute solutions. ## Sections - [00:00:00](https://www.youtube.com/watch?v=LfdK-v0SbGI&t=0s) **GPU vs CPU Basics** - Alex Hudac explains what a GPU is, how it differs from a CPU in core count and parallel processing, and why GPUs are essential for compute‑intensive workloads, especially in cloud environments. - [00:03:24](https://www.youtube.com/watch?v=LfdK-v0SbGI&t=204s) **AI, GPUs, and HPC in Cloud** - The passage outlines machine learning and deep learning as AI’s core components, describes specialized GPUs for training and inference, explains how GPUs enhance high‑performance computing workloads, and argues that leveraging GPUs in cloud environments provides the necessary performance for compute‑intensive tasks. - [00:06:29](https://www.youtube.com/watch?v=LfdK-v0SbGI&t=389s) **Cloud GPUs: Cost and Performance Benefits** - The speaker explains how cloud‑based GPUs reduce waste by charging only for used resources, deliver superior performance, and allow companies to focus on outcomes, while also covering GPU fundamentals, CPU differences, and key use cases such as VDI, AI, and HPC. ## Full Transcript
0:00Hi, my name is Alex Hudac. I'm an 0:02offering manager at IBM and today I'm 0:04going to talk to you about what is a 0:06GPU. So, I get some pretty basic 0:09questions on GPUs and that's what I'm 0:11going to go over today. First question 0:12is what is a GPU? What is the difference 0:15between a GPU and a CPU? So, I'm going 0:17to represent those here 0:21and then lastly, 0:24why use a GPU? And is it even important 0:26to use a GPU on cloud? So let's start 0:29first with what is a GPU. GPU stands for 0:32graphic processing unit. Uh but 0:35oftentimes people are more familiar with 0:37CPUs. So CPUs are actually made up of 0:41just a few cores. You can think of these 0:43cores as the power or the ability of a 0:46CPU to do certain calculations or 0:48computations. On the other hand though 0:51GPUs are made up of hundreds of cores. 0:56But what difference does it make? So the 0:58thing with a CPU is that when it does 1:00the computation, it does so in a serial 1:03form. So it does one computation at a 1:06time. 1:07But with a GPU, it does it in parallel. 1:12So the importance of these two 1:15differences is that with a GPU, you're 1:18able to do computations all at once and 1:21very intense computations at that. So 1:24oftentimes when you have app codes, a 1:26lot of it's going to be going to the 1:27CPU, 1:29but then every now and then you're going 1:31to have an application that's going to 1:32require quite a bit of compute intensive 1:36uh support that the CPU just can't do. 1:38So it's going to be offloaded to the 1:40GPU. So you can think of a GPU as that 1:43extra muscle or that extra brain power 1:45that the CPU just can't do on its own. 1:48So there are two main providers of GPUs 1:51in industry, Nvidia and AMD. Both 1:54providers manufacture GPUs that are 1:56optimized for certain use cases. So 1:59let's jump into that because big 2:00question I get is why do I even need a 2:02GPU? In what industries and in what use 2:05cases? So the first we'll talk about is 2:08VDI. 2:10VDI stands for virtual desktop 2:12infrastructure. So GPUs are created to 2:16support highintensive graphic 2:18applications. Uh so for think about if 2:21you're a construction worker, right, and 2:23you're out in the field and you need to 2:25access a very high graphic intensive 3D 2:28CAD program. So rather than having the 2:31server right next to you, right in the 2:32field with you, you can have a server 2:34that's in a country away in a cloud data 2:37center and be able to view that 3D 2:40graphic as if that server was right with 2:42you. and that's going to be supported by 2:44the GPU because the GPU supports graphic 2:47intensive applications. Another example 2:49of this would be movie animation or 2:51rendering. So in fact GPUs actually 2:55first got their name mainly with the 2:57gaming industry. Oftentimes they were 2:59referred to as gaming processing units 3:01because of this ability to provide uh 3:05end users with low latency graphics. But 3:09gaming is no longer the focus in 3:11industry anymore. It's big piece of it, 3:13but now financial services, life 3:15sciences and even healthcare are 3:17starting to get into it with artificial 3:19intelligence. 3:22So artificial intelligence has two big 3:24pieces to it. There's machine learning 3:26and there's deep learning. So now there 3:28are also GPUs that are optimized and 3:30created specifically for those 3:32applications. So there are some that are 3:34created for inferencing for machine 3:36learning purposes and there are some 3:38that are created to help data scientists 3:40create and train neural networks. In 3:43other words, they're trying to create 3:44these algorithms that can think like a 3:46human brain. That's something that a CPU 3:48can simply not do on its own and 3:50requires GPU capabilities. 3:53And then lastly, let's talk about HPC. 3:58HPC is a buzzword that's been going 4:00around. stands for high performance 4:02computing. Uh while a GPU is not 4:05absolutely necessary for HPC, it's an 4:08important part of it. So high 4:09performance computing is a company's 4:11ability to spread out their compute 4:14intensive workloads amongst multiple 4:16compute nodes or in the case of cloud 4:19servers. Oftentimes though, these 4:21applications are very compute inensive. 4:23It could include rendering. It could 4:25include AI and that's where a GPU comes 4:28in. You can add a GPU to these servers 4:30that are spread out amongst an HPC 4:32application and utilize those in that 4:34manner. So this is a nice little segue 4:37into why should we use GPUs on cloud? If 4:40HPC is such a big piece of that, what 4:43else is important about cloud? 4:47So the first part of that is you get 4:49high performance. 4:52You need cloud for that. The GPUs are 4:54great but not on their own. So back in 4:58the day and even still today there are 4:59companies that use a lot of on-prem 5:01infrastructure and they utilize that 5:04infrastructure for any of their compute 5:05intensive applications. However, 5:08especially in the case of GPUs, the 5:10technology is everchanging. In fact, 5:12there's typically a new GPU coming out 5:15almost every single year. So it's 5:17actually very expensive and nearly 5:19impractical for companies to keep up 5:20with the latest technology at this 5:22point. So cloud providers actually have 5:25the ability to continually update their 5:28technology and provide GPUs to these 5:31companies to utilize them when they need 5:33them. So on a more granular basis though 5:37cloud technology can often be broken 5:39down from an infrastructure perspective 5:41between bare metal and virtual servers. 5:44So let's talk about the differences. 5:49There are advantages of using a GPU on 5:52both types of infrastructure. If you 5:54utilize a a GPU on a bare metal 5:56infrastructure, the companies oftentimes 5:58have access to the entire server itself 6:01and they can customize the 6:02configuration. So this is great for 6:04companies that are going to be really 6:06utilizing that server and that GPU 6:08intense application on a pretty 6:11consistent basis. But for companies that 6:14need a GPU, maybe just in a burst 6:16workload uh scenario, the virtual server 6:19option might be even better. And the 6:21nice thing about virtual is that there 6:23are often different pricing models as 6:25well, including hourly. 6:27And the cool thing about cloud is that 6:29you only pay for what you use. 6:33So if a company is using on-prem 6:36technology or infrastructure, but 6:38they're not utilizing it at the time, 6:39that technology is depreciating and it's 6:42essentially a waste of money for that 6:43company. So it just makes a lot more 6:45sense from a cost perspective. And then 6:47because the GPU is so great at 6:49performance, it just makes sense from a 6:51performance perspective as well. So 6:53companies are able to focus way more on 6:55output than they are on keeping up with 6:57the latest technology. So in summary, 7:00what we covered is what is a GPU graphic 7:03processing unit, what the differences 7:05between a GPU and a CPU, the use cases 7:08for GPUs being in VDI, AI, and HPC, and 7:12why is it even important for GPUs to be 7:14used on cloud. Thank you guys for 7:16joining me here today to learn about 7:18what is a GPU. If you'd like to learn 7:20more about GPUs, click on the links 7:22below. And we're always checking the 7:24comments, so feel free to throw one in 7:25there for us. And be sure to subscribe 7:27for future