WEBVTT 00:00:00.000 --> 00:00:04.400 If you're a PC enthusiast, you probably know that the main way that our PCs, 00:00:04.400 --> 00:00:08.480 our phones, and our consoles have gotten faster over the years is through shrinking 00:00:08.480 --> 00:00:14.480 transistor sizes on our processors. Indeed, we've gone from using 250 nm chips in the 00:00:14.480 --> 00:00:21.120 mid-1990s to just 7 nm models today, and that number is expected to keep shrinking. 00:00:21.120 --> 00:00:25.440 But this was not achieved simply by taking the exact same transistor design 00:00:25.440 --> 00:00:30.000 and shrinking it every couple of years. Instead, engineers have had to get creative 00:00:30.000 --> 00:00:34.720 with how to build transistors in order to pull this off. As we touched on in our video about 00:00:34.720 --> 00:00:39.840 processor nodes, which you can check out up here, one major innovation was FinFET. 00:00:39.840 --> 00:00:45.360 This is a type of transistor in which the channel the electrons move through is raised up off the 00:00:45.360 --> 00:00:51.040 substrate, like a shark's fin, increasing the surface area and allowing more electrons to flow. 00:00:51.040 --> 00:00:56.400 But now, we're getting to the point where FinFET is reaching its limits in terms of how tightly 00:00:56.400 --> 00:01:00.320 we can pack the transistors together, which means the industry is going to need 00:01:00.320 --> 00:01:05.520 to move to a new transistor design for future generations of high-performance chips. Enter 00:01:06.560 --> 00:01:12.720 Gate All Around, a super boring name for a super cool update to the FinFET concept. 00:01:12.720 --> 00:01:18.160 Similar to FinFET, the channel is raised up off the substrate, but instead of being one large 00:01:18.240 --> 00:01:23.200 fin-like piece, it's broken up into several pieces called nanoribbons. 00:01:23.200 --> 00:01:27.280 To find out more about these nanoribbons, we talked to our friend Jack Cavalleros at Intel, 00:01:27.280 --> 00:01:32.400 and we'd like to thank him along with Bruce Feinberg, Marissa Ahmed, and Ben Benson for helping 00:01:32.400 --> 00:01:37.920 us with this episode. So Gate All Around design increases the surface area so that the gate that 00:01:37.920 --> 00:01:44.160 controls the current flow surrounds the nanoribbon on all sides, hence the name Gate All Around. 00:01:44.160 --> 00:01:49.600 This is hugely important for a few reasons, actually. You see, the gate's ability to control 00:01:49.600 --> 00:01:55.600 current depends on the physical size of the channel, and a single bulky fin can be hard to control. 00:01:56.160 --> 00:02:01.440 At present transistor sizes, it's not a huge problem, but if we kept shrinking FinFET transistors 00:02:01.440 --> 00:02:06.640 even further, the lack of control would introduce leakage, which would interfere with, and possibly 00:02:06.640 --> 00:02:11.520 corrupt, your data flow. But with Gate All Around, the smaller nanoribbons means the gate has a much 00:02:11.520 --> 00:02:17.200 easier time controlling the current flow, allowing for more tightly packed transistors. On top of that, 00:02:17.200 --> 00:02:22.240 you can stack the nanoribbons, see what we did there, which takes up less space than adding 00:02:22.240 --> 00:02:26.400 more fins in FinFET, which required you to put the fins next to each other instead. 00:02:26.400 --> 00:02:30.960 Another big advantage is that if a certain processor design requires more power, 00:02:30.960 --> 00:02:35.920 you can simply make the nanoribbons wider, laterally. With FinFET, each fin could carry a 00:02:36.000 --> 00:02:41.600 discrete amount of current, so if a design needed, say, 5x current, but each FinFET could only carry 00:02:41.600 --> 00:02:48.080 2x, you'd have to have 3 fins with a capacity of 6x, a design that wasted space. However, 00:02:48.080 --> 00:02:52.960 you can make a nanoribbon just as wide as you need, saving valuable room on the transistor, 00:02:52.960 --> 00:02:57.680 and the best part is that you don't need super exotic materials in order to pull this off. 00:02:57.680 --> 00:03:02.800 It can be done with a silicon-germium alloy, which is way more basic than it sounds, 00:03:02.800 --> 00:03:07.440 meaning that it shouldn't be too difficult for large chip makers to produce gate all-around chips 00:03:07.440 --> 00:03:12.240 in bulk. On that subject, when should we see gate all-around transistors hitting the market? 00:03:12.800 --> 00:03:17.840 Well, it should be within the next five years across all vendors, but it'll probably appear 00:03:17.840 --> 00:03:22.800 in data center and server applications before it trickles down to consumers, like you and me. 00:03:22.800 --> 00:03:26.560 But once it does, we're going to see the expected benefits that we're used to with 00:03:26.560 --> 00:03:31.280 smaller transistors, like faster computing, better battery life, lower power consumption, 00:03:31.280 --> 00:03:36.080 and an ability for devices like laptops and phones to better handle tasks without having to 00:03:36.080 --> 00:03:41.200 rye as much on cloud computing. I wonder how much better it might make our games as well, 00:03:41.200 --> 00:03:47.280 assuming we can get nanoribbon GPUs from someone other than an eBay scalper when the time does come. 00:03:47.280 --> 00:03:50.880 So thanks for watching guys, if you liked this video, hit like. If you subscribed to this video, 00:03:50.880 --> 00:03:54.640 hit subscribe. What? And be sure to hit us up in the comment section with your suggestion 00:03:54.640 --> 00:03:56.880 for topics that we should cover in the future.