Feedback | SoME

2678x Faster: How GPUs enabled Deep Learning Revolution

Parallel Matrix Multiplication on a GPU using CUDA C.

Analytics

5.55 Overall score*

60 Rank

38 Votes

7 Comments

Comments

More related to computer science and programming than math, bust still a nice topic to learn about. Some concepts are needed to explain because they are used aumming that the viewer knows about it. for example threads in computing.

6.8

2000x or 2000%? Quite detailed.

3.5

While overall there were some good points and education made in the video, I feel that there was one bog issue for me. It felt like there was a high level of assumed knowledge with no indication of where that knowledge comes from, where I could go to learn it, what is necessary to understand the video, etc. I think that it would have been useful to have, whenever a certain concept in introduced, a brief summary along with where to learn it. Matrix multiplication is an exception to this in that you did explain, but I felt it went into more detail than was relevant. It too could have used the same type of summary. E.g. "matrix multiplication is a process which takes two different matrices, performs a whole bunch of different calculations for each row and column, and spits out another matrix at the end. All you need to know for this video is that it requires a lot of separate calculations, but if you'd like to learn more I recommend 3blue1brown's series on linear algebra." Summaries like this would have been useful for a lot of the code shown in the video as well as for some more about the hardware. Despite all this criticism I'm giving, I still think it's a good video! It taught me something and was largely captivating of my attention.

6.5

Very instructive

5.6

Interesting topic, I didn't know that's how the GPU made things so much faster. I didn't care much for the specific code implementation, many others probably would find that interesting though, and got all I wanted in the first bit when you explained the GPU does a gajillion threads.

7.8

The topic intrigued me and created a value addition.

clear exposition. not very inspaiering content. too teqnical at the end (may be good for programing hands on vidio, but not for this compitition). also the neration is a bot to monothon and the acent is a bit too strong.