字幕列表 影片播放 列印英文字幕 - Hey guys, this is Austin. This little box might not look like much, but inside is one of the strangest stories in Intel's history. So this is the Intel NUC8i3CYSM. (bell ringing) Intel have been selling these little NUCs for quite a while, and some of them are pretty cool. Last year we took a look at the Skull Canyon NUC, which had a proper amount of gaming power in a tiny package. With a lowly Core i3 inside, this does not look nearly as impressive, and that's because it's not, at least on paper. So traditionally, Intel has used what is known as the tick-tock model, not to be confused with tick tock. The tick, tock. So essentially there are two different ways of building a new chip. First of all, you just start with the big stuff. The architecture of the things will actually legitimately make it faster year-on-year. Once you do that, the next iteration will just take that design and shrink it to a smaller process node. So if we get a little bit of Michael's Toys action, let me show what happened just a couple years ago. That is a really bad square even though I have a ruler. Anyway, so say that this is the previous generation. This would be 14 nanometer, which is pretty much what all single Intel processors use today. This strategy worked great until they hit the 10 nanometer node. Now this is what the Cannon Lake chip inside is based on, and unfortunately, well, they hit some road bumps. So the advantage of going to a smaller process node is that you can cram more and more transistors into a smaller and smaller area. So if we have a quad-core processor, and this is very simplified, what you're getting here is basically the same number of cores but just in a smaller area, which means that not only is the chip either cheaper to produce or you can actually fit more cores. So in theory, if you had a 10 nanometer chip which is the same size as the old one, it could be a six-core design as opposed to a quad-core one. So when I say 14 nanometer and 10 nanometer, what I'm referring to is the process technology used to actually build these chips. So think of it like this. If I'm trying to draw something really precise, it's the difference between using a Sharpie and a pen. It is a much finer-grained tool which I could be much more precise with. The smaller the tip, the more precise and the more stuff you can cram into these tiny, tiny little lines. This is absolutely key to the reason why computers have advanced over the last 50 years. Every time you get a smaller and smaller process node, it means you can get more densely pack it with transistors which can be spent on things like the CPU, the GPU. Really, this is absolutely the reason why things are 100 million times faster than they were in 1972. Now, the downside here is that the smaller and smaller you get, the more difficult it is to actually pack all these transistors. I mean, seriously, at a certain point, you hit the atomic level where just things don't really get much smaller. From 2006 to 2014, every two years, Intel were able to shrink the process over and over and over again. That is until 2014 where they hit a huge, huge wall. Now this forced Intel to get very creative with their updates without the advantage of being able to constantly change their process. Instead, they did things like add more CPU cores to pretty much their entire lineup, which brings up to the Cannon Lake CPU inside this NUC. Now originally it was supposed to ship as a 7th Generation part all the way back in 2016, but as you might imagine, things didn't quite go that smoothly. So why should anyone care about this little NUC? Because inside this is running Intel's 10-nanometer Cannon Lake processor, and up until this point it is the very first and only device that's shipped with it. So to compare, I have, well, maybe not a perfect apples-to-apples comparison. However, this is an Aspire E5, and it does have a very similarly-specced Core i3, just a 14-nanometer version as opposed to the 10-nanometer inside the NUC. Put these two side-by-side, and they're about as close as it gets. They're both dual-core processors, they both have Hyper-Threading. They both have a 2.2 gigahertz base. Really, the only difference is that the laptop has a slightly higher turbo with 3.4 versus 3.2, but besides that, it's gonna be about, well, this is about as close as I can get a comparison. So first of all, let's start out with Geekbench. Now I'm not expecting any huge performance differences, but what's interesting about this is that the Cannon Lake chip inside is pretty much entirely undocumented. Now, yes, on the ark.intel site, you can see that it does confirm at least that it's 10 nanometer, but beside that, this is an entirely new CPU architecture which has basically never been talked about before. It's unusual because usually when Intel brings out a new series of chips, they're way more on top of talking about the different specs and everything that they've done to improve it, but with this, it's like it was a weird half-step that they never actually really fully acknowledged? Okay, so very, very close. The NUC is very slightly quicker in single core, and a little bit quicker multi-core, but honestly, that's close enough that I would consider it to be basically a draw. Next, let's give Cinebench a try. In any case, it definitely does not look like Cannon Lake is any kind of major improvement as far as IPC goes. It's very, very close. So in this one, the laptop wins by a little bit, 353 versus 331. I'm pretty sure that's entirely just that slightly higher turbo. I want to give a huge shout out to Ian from AnandTech for not only helping me research for this video, but also doing a terrific write-up all about the incredible saga that was Intel's 10-nanometer process. The fact that this is a product that is in my hand in 2019 is kind of incredible 'cause for all intents and purposes, Cannon Lake has been completely shelved at this point. So where are things today? Well, Intel has pretty much given up on the Cannon Lake chips in favor of going straight to Ice Lake, which in theory will be shipping later this year on a 10-nanometer process with improved architecture. Essentially, it's a tick and a tock rolled into one. Meanwhile, we have the NUC as what is arguably one of the only examples of an Intel future that we never got to experience, although judging by the performance, wasn't that all exciting to begin with.