There is no question about it. GPU Computing is a popular topic right now. All you you need to do is look at this graph from Google Trends and you’ll see how interest in this technology has jumped in the last year in the search volume index.
So you’ve heard about it, but you don’t know where to start when it comes to learning and understanding the technology.
For a minute, I am going to assume that you have already heard about the c410x PCI Expansion Chassis from Dell Data Center Solutions. This monster is a 3u chassis that holds up to sixteen NVIDIA Tesla 2070 GPU’s giving you a whopping 16.5 tfps in a very small package!
The idea here in that when you remove the GPU’s from the compute chassis and you make each one of them independently serviceable, you remove the issue of having to down the entire compute platform by allowing someone to just work on the one card that needs servicing.
So back on point, CUDA is NVIDIA’s parallel computing architecture. Using CUDA and the GPU you can decrease the computing time it takes to do any number of tasks that would have taken days or even weeks to run using standard CPU computing down to a few hours potentially, depending on the workload. Let me give you an example.
One of my customers crunches numbers for the oil and gas exploration industry. Before they went to a GPU solution they were using standard CPU compute power (thousands of CPU’s) and a normal data run would take roughly three weeks. They then implemented a GPU solution which took that same data set from three weeks to about 3.5 hours. I could go on and on with examples, but suffice it to say, GPU computing is only scratching the surface as to what is possible.
So where do you go to learn about GPU computing and the CUDA architecture? The folks at NVIDIA have some of the best online content out there, and a lot of it is freely available via a webinar or other online means. The first place I will point you to is the CUDA Zone. This is a site so rich with information that you’ll likely be here a while. I know that it was one of the first places I landed when researching GPU and CUDA many months ago, and I still go back frequently to see what else is up there. The next place I will point you is the GPU Computing webinars. Some of these are open to everyone, while others require you to be a Registered Developer. In any case, there is still so much there you can’t help but learn as you browse through their materials. Finally, you should take time to visit the Resources section that will give you access to MANY great videos on different topics related to CUDA and GPU Computing.
GPU is really picking up steam, and the possibilities of what it can do for us are endless. Thanks to the folks at NVIDIA, the learning curve will be much smaller.
Until next time, I’ll see you in the Cloud!