There are many familiar examples of technological marvels. For example, cramming billions of transistors onto a computer chip is quite amazing. But do you know how cool streaming video is?

Intel's Tom Petersen recently took a deep dive into video compression intuition for Gamers Nexus. For starters, without video compression technology, the one billion hours of video delivered per day on YouTube alone would consume more than 100 times the global Internet bandwidth.

To put this in numbers, according to Petersen, the world's current Internet bandwidth is 1.2Pbits per second; assuming streaming at 1080p SDR, YouTube would need 155Pbits per second to stream content in uncompressed video needed.

Full HDR 4K video can multiply this number by about 10. In other words, if YouTube were to stream all of its content in uncompressed 4K HDR, it would require about 1,000 times the available Internet bandwidth in the world. Wild.

The solution, of course, is video compression, and Petersen describes the five main components of compression. First is color downsampling, which takes advantage of the fact that the human eye is more sensitive to luminance than to color, and more sensitive to one color than another.

This means that some color data can effectively be discarded without changing the subjective appearance of the image. This allows for up to twice the compression.

Then there is spatial and temporal compression. One aspect of this means updating only those pixels whose colors change from one frame to the next. It also allows the use of vectors to move pixels, rather than simply updating the color data completely. This results in up to 20 times more compression.

From there, things get more complicated and, frankly, harder to understand. Following spatial and temporal compression comes error correction. Then comes frequency quantization.

This is the most difficult step in the process to understand, but in the end a large number of pixels can be discarded without affecting image quality. See the video for more details, but this step alone allows for an additional 40x compression.

Finally, there is symbol encoding. This is basically abbreviated encoding of frequently repeated patterns in the binary image data itself. This allows for up to twice the compression.

This completes the entire compression pipeline; the exact value of bandwidth used by YouTube after compression depends on the video content and resolution being streamed.

However, YouTube's own guidance for sustained speeds for 4K video is 20 Mbps and 5 Mbps for 1080p. That's 2.5MB and 0.625MB per second, respectively. So how does that compare to raw, uncompressed video?

At 8-bit color, regular SDR 1080p requires roughly 185 MB per second, while full 4K HDR clocks in at a whopping 2 GB per second. Both of these numbers are quite mind-boggling.

In any case, to decompress and view the encoded video, the compression steps are done in reverse order: all modern GPUs, including Intel's CPU-integrated hardware and Arc graphics cards, have a dedicated media engine with fixed-function hardware for each of the five compression steps media engines that can complete the task on the fly while providing very high throughput.

But digging into the details, video compression is arguably regarded as a less-sung-about but absolutely stunning technological achievement.