WEBVTT - Intel's Tick-Tock Strategy 0:00:00.280 --> 0:00:02.960 Brought to you by the reinvented two thousand twelve Camray. 0:00:03.160 --> 0:00:08.920 It's ready. Are you get in touch with technology? With 0:00:09.039 --> 0:00:18.040 tech Stuff from how stuff works dot com. Hello again, everyone, 0:00:18.079 --> 0:00:20.560 and welcome to tech stuff. My name is Chris Polette 0:00:20.600 --> 0:00:23.160 and I am an editor at how stuff works dot com. 0:00:23.160 --> 0:00:25.520 Sitting across from me as always a senior writer Jonathan 0:00:25.720 --> 0:00:29.160 stir Rickland. Wake up in the morning feeling like p didd. 0:00:29.480 --> 0:00:32.080 I grabbed my glasses. I'm out the door. I'm going 0:00:32.240 --> 0:00:37.040 to hit this city. I don't want to hit something too. 0:00:37.080 --> 0:00:40.479 But that's not what I'm thinking about. Yeah, I I 0:00:40.560 --> 0:00:44.600 despise the source of that. But but at the same time, 0:00:44.640 --> 0:00:47.400 it was too it was too apropos. I could not 0:00:47.479 --> 0:00:50.480 pass it up with all the the TikTok related things 0:00:50.520 --> 0:00:54.960 that you could think of. It was TikTok spoiler. So, guys, 0:00:55.040 --> 0:00:59.080 we're gonna talk today about a an interesting strategy developed 0:00:59.080 --> 0:01:04.240 by into l uh in their micro processor micro architecture 0:01:04.440 --> 0:01:09.160 design work. Right, Yeah, this is this is something that 0:01:10.120 --> 0:01:12.240 chip heads would you would you call them chip heads 0:01:12.240 --> 0:01:15.280 people who really care about the processor speed of computers. 0:01:15.280 --> 0:01:19.199 I call them processor freaks, but with a pH Okay, 0:01:20.160 --> 0:01:22.319 this is something that they would pay attention to, but 0:01:22.520 --> 0:01:25.280 probably the general public doesn't know a whole lot about 0:01:25.440 --> 0:01:29.080 because it's something that goes on somewhat behind the scenes. 0:01:29.080 --> 0:01:32.039 Although Intel doesn't really make a big secret of of 0:01:32.080 --> 0:01:34.280 doing it this way. No, No, they've got they've got 0:01:34.280 --> 0:01:37.080 plenty of information on their own website, completely open to 0:01:37.120 --> 0:01:39.400 the public that explains this, because I mean, really, it's 0:01:39.400 --> 0:01:42.000 just it's showing a process, right. It's not giving any 0:01:42.040 --> 0:01:45.080 proprietary information away, not in the least. It's kind of 0:01:45.120 --> 0:01:48.280 like saying that you know, a car manufacturing plant uses 0:01:48.360 --> 0:01:51.000 an assembly line of some sort that that tells you 0:01:51.120 --> 0:01:53.320 the process or using, but doesn't give you a need detail. 0:01:54.520 --> 0:01:59.560 True enough, So the uh, it's appropriately named the TikTok 0:01:59.680 --> 0:02:02.880 strate g because it's sort of like a pendulum. Uh. 0:02:02.880 --> 0:02:04.880 It moves one way and then the other and then 0:02:04.920 --> 0:02:09.240 back the other way. Um, but that's not really the 0:02:09.440 --> 0:02:12.519 pendulum is not really completely a good analogy because it's 0:02:12.560 --> 0:02:15.079 not just moving back and forth that the process is 0:02:15.120 --> 0:02:18.799 actually moving forward during this time. Yeah, and we can. 0:02:19.000 --> 0:02:21.600 Let's let's before we get into the actual details of 0:02:21.600 --> 0:02:24.720 what TikTok is all about, let's talk about the reason 0:02:24.800 --> 0:02:28.240 TikTok even needs to exist in the first place, and 0:02:28.280 --> 0:02:32.880 that would be from Intel's co founder Gordon Moore. Yeah. So, 0:02:32.880 --> 0:02:36.160 so Gordy, you may remember we've talked about Gordy in 0:02:36.200 --> 0:02:38.640 the past, and uh, you know somebody from Intel is 0:02:38.639 --> 0:02:42.480 listening to that. You know. Yeah, we've had we've had 0:02:42.600 --> 0:02:45.320 some important people email us in the past. Like I'm 0:02:45.320 --> 0:02:49.000 thinking of Vinton Surf who sent me an email and 0:02:49.040 --> 0:02:51.840 that just it's hard for me to remember that that 0:02:52.200 --> 0:02:57.639 important really like people I really legitimately admire and respect 0:02:57.680 --> 0:03:00.240 and who have had an enormous impact upon in the 0:03:00.400 --> 0:03:04.640 technology industries. Um, I can hear what I have to say. 0:03:04.960 --> 0:03:07.040 That just blows my mind because normally I'm just having 0:03:07.040 --> 0:03:08.760 these conversations in a room and we don't have a 0:03:08.800 --> 0:03:12.560 microphone in front of me. But anyway, Gordy, uh night. 0:03:12.600 --> 0:03:17.440 Back in NINETI wrote this paper about cramming more components 0:03:17.919 --> 0:03:20.959 into a chip. I can't remember. There's something along those 0:03:20.960 --> 0:03:24.360 lines as the title I'm paraphrasing, but I can. I 0:03:24.360 --> 0:03:26.280 can look that up for you if you want. You 0:03:26.360 --> 0:03:28.079 go ahead and look that up while I talk about 0:03:28.160 --> 0:03:30.480 about this. So we're talking about Moore's law here. So 0:03:30.880 --> 0:03:34.560 Gordon came up with this observation, and it was an observation. 0:03:34.560 --> 0:03:36.760 You've got to remember that Moore's law was not some 0:03:36.800 --> 0:03:39.800 sort of fundamental law of the universe. The observation was 0:03:39.880 --> 0:03:43.080 that over a certain period of time, and I believe 0:03:43.360 --> 0:03:47.120 initially it might have been between twelve and eighteen months, 0:03:47.160 --> 0:03:50.040 it's now closer to twenty four months, but over a 0:03:50.080 --> 0:03:53.880 certain amount of time, More observed that the number of 0:03:53.920 --> 0:03:57.560 transistors that one that a company could fit on a 0:03:57.640 --> 0:04:02.680 single square inch of so con wafer material doubled, so 0:04:02.960 --> 0:04:06.080 you could fit twice as many transistors on a chip 0:04:06.640 --> 0:04:11.120 UH within a year or two years um. And that 0:04:11.320 --> 0:04:13.720 he his observation was that this was a trend that 0:04:13.760 --> 0:04:17.560 would continue indefinitely until we started to reach some fundamental 0:04:17.600 --> 0:04:21.680 limits of how small we could make transistors. And at 0:04:21.680 --> 0:04:24.160 the time, no one was really sure you know how 0:04:24.200 --> 0:04:27.120 long that would be. And it's the crazy thing is 0:04:27.160 --> 0:04:29.320 it's held true even today. And you've got to remember, 0:04:29.360 --> 0:04:34.320 this is an exponential uh pattern, right, I mean it's 0:04:34.320 --> 0:04:37.960 it's decreasing, the sizes decreasing by half, the number is 0:04:38.080 --> 0:04:42.960 increasing bye bye, you know, twice as much each time frame, 0:04:43.520 --> 0:04:46.880 So before too long you get an incredible number of 0:04:46.880 --> 0:04:51.359 transistors on a silicon chip. The name of the paper. Yeah, 0:04:51.400 --> 0:04:53.880 And as a matter of fact, Jonathan cited this for 0:04:54.000 --> 0:04:58.880 his really awesome article on War's Law. Um a couple 0:04:58.920 --> 0:05:01.880 of years ago. Wow, that that along, are you uh called? 0:05:01.920 --> 0:05:05.600 It's actually called cramming more components onto integrated circuits. So yeah, 0:05:05.640 --> 0:05:07.640 that's pretty much right. Yeah, it was pretty close. But 0:05:07.680 --> 0:05:11.039 you can find that in Electronics. That's the name of 0:05:11.080 --> 0:05:16.120 the journal. It was April volume thirty eight, number eight. 0:05:16.160 --> 0:05:18.120 If you want to read it, and it's actually it's 0:05:18.160 --> 0:05:20.760 not a dry read. You can actually find a link 0:05:20.800 --> 0:05:23.000 to it on Intel. If you go to Intel and 0:05:23.040 --> 0:05:25.720 you start looking at Moore's law, there is a link 0:05:25.760 --> 0:05:28.159 to a PDF of this paper, and I do recommend 0:05:28.160 --> 0:05:31.159 you read it if you're interested in the original observation. 0:05:31.720 --> 0:05:33.800 One thing that I thought was really interesting was that 0:05:33.880 --> 0:05:37.880 Moore was pointing out this isn't just a technological issue. 0:05:37.920 --> 0:05:40.600 In fact, that was not the main thrust of his paper. 0:05:41.200 --> 0:05:44.440 It's it's a financial issue because one, you have to 0:05:44.440 --> 0:05:46.880 find the technology to be able to decrease the size 0:05:46.880 --> 0:05:49.719 of these transistors, but too you have to make that 0:05:49.760 --> 0:05:53.480 technology affordable enough to use for it to make sense 0:05:54.160 --> 0:05:57.320 to use it right if you if you develop the 0:05:57.360 --> 0:06:02.120 technology to create a up that has UH an incredible 0:06:02.200 --> 0:06:05.800 number of transistors on it, but it's slow, inefficient, it 0:06:06.240 --> 0:06:09.880 costs a lot of money to create each chip. You're 0:06:09.920 --> 0:06:11.680 not you don't have a good business plan because you 0:06:11.720 --> 0:06:15.480 can't sell those chips to consumers. They would be too expensive. 0:06:15.560 --> 0:06:18.400 You would never recapture those costs. So you have to 0:06:18.400 --> 0:06:20.800 be able to one develop the technology and to develop 0:06:20.880 --> 0:06:24.600 the right procedure so that the technology is efficient and 0:06:24.720 --> 0:06:28.839 you can actually make money off of it. Speaking of which, um, 0:06:28.880 --> 0:06:32.720 some people have predicted the demise of Moore's law, of 0:06:32.760 --> 0:06:36.599 course due to the physical limitations UH for years, but 0:06:36.880 --> 0:06:40.800 also due to the recent UH financial troubles the world over. 0:06:41.200 --> 0:06:44.000 People have said, well, you know, there won't be really 0:06:44.000 --> 0:06:48.400 a need to have faster, faster, faster, faster computers because 0:06:48.600 --> 0:06:52.719 people can't afford to go by them. So UM, but 0:06:52.800 --> 0:06:55.120 I haven't really seen it slow down as much as 0:06:55.320 --> 0:06:58.920 become sort of sort of zag a little bit. And 0:06:59.000 --> 0:07:02.480 that when and the past years we've had single core 0:07:02.560 --> 0:07:05.760 processors and you would see a speed a substantial speed 0:07:05.760 --> 0:07:08.280 boost over the period of a year. Um, you know 0:07:08.320 --> 0:07:11.000 it would go from one gigga hurts to one point 0:07:11.040 --> 0:07:13.480 six gigga hurts, and you know then to two point 0:07:13.520 --> 0:07:16.680 to giga hurts. Well, now we have multi core processors, 0:07:16.880 --> 0:07:20.280 and they don't seem to move as fast because you know, 0:07:20.320 --> 0:07:23.160 we're going from uh, you know, three point two to 0:07:23.280 --> 0:07:26.640 three point six gigga hurts. But then we're also doubling 0:07:26.640 --> 0:07:29.840 the number of cours on that chip, so they are 0:07:29.880 --> 0:07:33.520 getting faster. Just doesn't appear to move as fast because 0:07:33.600 --> 0:07:36.040 of the way it doesn't. At least that's that's my 0:07:36.120 --> 0:07:39.440 personal observer. Yeah, the the jumps, the jumps, an actual 0:07:39.640 --> 0:07:44.880 number of cycles per second that a processor is capable 0:07:45.040 --> 0:07:48.360 of of executing doesn't seem to be jumping at the 0:07:48.400 --> 0:07:51.920 same rate as it was, you know, ten years ago, right, 0:07:52.360 --> 0:07:55.680 But the the multi core approach has created a more 0:07:55.720 --> 0:08:00.720 efficient way to deal with computational problems, and thus ultimately 0:08:00.760 --> 0:08:05.000 you're computing power has has increased, even though the number 0:08:05.040 --> 0:08:07.640 of cycles themselves may not have jumped as high as 0:08:07.640 --> 0:08:09.800 you would have expected. And that's that's part of the 0:08:09.800 --> 0:08:12.880 whole TikTok philosophy actually, which I guess we can kind 0:08:12.880 --> 0:08:16.440 of segue into UH. Until really started to adopt this 0:08:16.560 --> 0:08:21.200 around two thousand seven. Really, was it that long ago? Yeah? Yeah, 0:08:21.280 --> 0:08:23.920 it was. Um it was right around then when they 0:08:23.920 --> 0:08:27.520 had started to develop the core technology. That was when 0:08:27.600 --> 0:08:31.440 the core technology was first starting to be uh UM introduced. 0:08:31.440 --> 0:08:35.600 And at that time, the the number of the the 0:08:35.760 --> 0:08:42.240 discreet elements on a chip, we're around the sixty nanometer scale. 0:08:43.080 --> 0:08:45.640 So we're already talking about on the nano scale, which 0:08:45.679 --> 0:08:49.040 is yeah, remember a nanometer, And I got a lot 0:08:49.080 --> 0:08:51.080 of people yelling at me for calling it a nanometer. 0:08:51.520 --> 0:08:53.520 I'm sorry, I get people yelling at me, no matter what, 0:08:53.559 --> 0:08:55.400 we're gonna go nanometer this time, and then all the 0:08:55.400 --> 0:08:57.120 other people can yell at me because I'm sure they 0:08:57.120 --> 0:09:01.360 were tired of being left out. So a nanometer is 0:09:01.440 --> 0:09:05.360 one billionth of a meter, so this is incredibly tiny. 0:09:05.360 --> 0:09:07.240 We're talking about on a scale where even a moat 0:09:07.360 --> 0:09:11.360 of dust on a silicon wafer will ruin an entire 0:09:11.440 --> 0:09:15.280 chip because the mode of dust dwarfs the elements that 0:09:15.320 --> 0:09:19.040 would be printed on that chip. UM. So, sixty five 0:09:19.120 --> 0:09:22.199 nanometer scale for the core technology, now that was technically 0:09:22.640 --> 0:09:26.600 a talk that meant that Intel had already developed a 0:09:26.679 --> 0:09:31.000 chip that could be at sixty five nanometers that would 0:09:31.000 --> 0:09:34.319 be the size of the transistors essentially on that chip. 0:09:35.240 --> 0:09:38.439 So the core technology was a new way of arranging 0:09:38.480 --> 0:09:42.320 those sixty five nanometer transistors in such a way that 0:09:42.360 --> 0:09:47.280 they were more efficiently used, so that they consumed less power, 0:09:47.720 --> 0:09:50.880 they had better output, there was a more streamlined flow, 0:09:51.880 --> 0:09:54.640 so that they were, in a sense an essence, a 0:09:54.760 --> 0:09:57.319 more powerful chip. Because one thing we also need to 0:09:57.400 --> 0:09:59.800 remember about Moore's law is today a lot of people 0:09:59.800 --> 0:10:02.640 in herpret it not as there are twice as many 0:10:02.679 --> 0:10:06.440 transistors now as there were two years ago, but rather 0:10:06.679 --> 0:10:09.920 the chip itself is twice as powerful as it was 0:10:10.120 --> 0:10:13.240 two years ago. This is not exactly the same thing. 0:10:13.240 --> 0:10:16.080 They're related, but you can get more power out of 0:10:16.080 --> 0:10:19.440 a chip just by realigning certain elements and making a 0:10:19.480 --> 0:10:24.160 more efficient workflow. You know, when you start talking about 0:10:24.200 --> 0:10:26.959 more power, I'm starting to get James doing in the 0:10:27.040 --> 0:10:30.080 back of my head. I'm giving her all. She's gone. 0:10:30.679 --> 0:10:33.440 Um so coming to the Jeffreys tubes in your ear. 0:10:33.640 --> 0:10:38.760 Nice Jeffreys too in your PC, trying to a squeak 0:10:38.800 --> 0:10:40.640 out that extra Now, if you guys want to know, 0:10:40.920 --> 0:10:42.200 if you guys want to know more about that, you 0:10:42.240 --> 0:10:44.760 need to read how warp speed works. We actually have 0:10:44.800 --> 0:10:49.200 it on the site. Um the so, yeah, the whole 0:10:49.200 --> 0:10:52.000 basis of tiktalk, right, we have. We've kind of danced 0:10:52.000 --> 0:10:55.480 around it. But Intel's TikTok strategy is that the TICK 0:10:56.240 --> 0:10:58.800 is figuring out a way to reduce the size of 0:10:58.800 --> 0:11:02.560 those transistors, but you bace it upon the previous UH 0:11:02.880 --> 0:11:07.160 chips micro architecture. All right. So so it's like you've 0:11:07.160 --> 0:11:10.040 got the plans for a house. Now you're going to 0:11:10.080 --> 0:11:12.319 build that same house, but you're gonna build it at 0:11:12.400 --> 0:11:15.679 half that scale. I got it, I got it. So 0:11:15.760 --> 0:11:18.560 the TALK strategy is finding out the best way to 0:11:18.760 --> 0:11:22.400 use those smaller components so that it is the most 0:11:22.440 --> 0:11:28.280 efficient effective transistor arrangement. So in that case, what you 0:11:28.360 --> 0:11:31.080 do is you look at the plans for that house 0:11:31.640 --> 0:11:34.120 that you built at half the size of the previous one, 0:11:34.320 --> 0:11:36.680 and you say, all right, I'm going to rearrange this 0:11:36.760 --> 0:11:39.480 now so that this house makes sense at this scale. 0:11:39.600 --> 0:11:41.560 It's I'm not going to change the scale. I'm just 0:11:41.559 --> 0:11:44.320 going to change the layout of the house. So what 0:11:44.360 --> 0:11:48.000 you're saying is they they create a blueprint for a 0:11:48.080 --> 0:11:52.640 chip the next UH and then the next cycle they 0:11:52.640 --> 0:11:59.400 work on making the components using that blueprint uh more efficient, right, 0:11:59.679 --> 0:12:02.000 and then and they take the efficiency that they learn 0:12:02.080 --> 0:12:05.160 on that cycle and build a new architecture on it 0:12:05.240 --> 0:12:07.880 for the next cycle. And so on one side they're 0:12:07.920 --> 0:12:12.440 working on making everything more uh you know, reducing the 0:12:12.480 --> 0:12:14.920 size of it, and the next cycle is all about 0:12:14.960 --> 0:12:19.880 the actual design of it. Right, So exactly goes size 0:12:19.920 --> 0:12:22.400 design size design. And this is, by the way, a 0:12:22.520 --> 0:12:26.600 never ending research process. It's not like you've got people 0:12:26.679 --> 0:12:29.960 researching how to reduce the size and then they stop, 0:12:30.040 --> 0:12:32.040 and then they switch to figure out how to make 0:12:32.120 --> 0:12:34.640 it more efficient. You've got teams working on both of 0:12:34.679 --> 0:12:39.320 these things simultaneously. And so um, it's interesting. You know 0:12:39.400 --> 0:12:41.680 we look back and the core technology being the talk 0:12:41.679 --> 0:12:46.600 at SS Well, the next one was the pen Ryn chip. 0:12:47.160 --> 0:12:49.400 That was that was the tick. Yeah. Now that one 0:12:49.600 --> 0:12:53.160 used the same micro architecture as the core processors, but 0:12:53.240 --> 0:12:56.520 this time they had reduced the size of the elements 0:12:56.559 --> 0:13:01.480 to the forty five nanometer scale. Now, after Pentrin came 0:13:01.520 --> 0:13:05.600 one of the chips that I wrote about, Yes, the Haleem, 0:13:05.679 --> 0:13:10.360 yes Nehalem microprocessor micro architecture. That was the next talk, 0:13:10.800 --> 0:13:15.160 and the Haleem had introduced a lot of interesting features 0:13:15.200 --> 0:13:19.800 like multi threading and and uh arranging memory in such 0:13:19.840 --> 0:13:22.360 a way so that it would uh that the various 0:13:22.400 --> 0:13:25.600 cores could share memory certain parts of memory very quickly 0:13:25.640 --> 0:13:28.080 to make it more efficient. That's what we were talking 0:13:28.120 --> 0:13:31.120 about here. They're actually rearranging the elements that are on 0:13:31.160 --> 0:13:34.880 the microprocessor chip in such a way that that the 0:13:35.000 --> 0:13:39.360 data flow is faster just because it makes more sense, right, 0:13:39.840 --> 0:13:42.600 And it doesn't necessarily it's it's it's an arrangement that 0:13:42.640 --> 0:13:45.320 would not have necessarily worked at the larger scale because 0:13:45.320 --> 0:13:50.200 you could not physically find that same configuration with the 0:13:50.280 --> 0:13:52.960 elements being larger. They had to be that size for 0:13:53.000 --> 0:13:55.200 you to be able to kind of shift them around. 0:13:55.240 --> 0:13:57.400 It's almost like one of those puzzles where you have 0:13:57.520 --> 0:14:00.440 one piece missing and you have to slow the other 0:14:00.480 --> 0:14:03.240 pieces around until you make the right picture. It's kind 0:14:03.240 --> 0:14:05.439 of like that. You know you've got you've got this 0:14:05.679 --> 0:14:08.520 certain amount of space that you are allowed to use, 0:14:09.000 --> 0:14:11.400 and you have elements of a certain size, and you 0:14:11.440 --> 0:14:13.160 have to find the right way to fit all those 0:14:13.160 --> 0:14:15.839 elements together so that it's the most efficient possible. Well, 0:14:15.880 --> 0:14:18.319 with the larger ones, you just don't have as many configurations. 0:14:18.320 --> 0:14:21.520 You don't have as much freedom because the elements themselves 0:14:21.520 --> 0:14:23.560 are bigger. You don't, you can't, you know, there's only 0:14:23.600 --> 0:14:28.280 so many configurations you can use. So after Nehalem came Westmere, 0:14:28.840 --> 0:14:31.400 and Westmere was another tick. So it was using the 0:14:31.480 --> 0:14:35.720 same micro architecture as Nehalem. But now we have gone 0:14:35.760 --> 0:14:40.640 down to the thirty two nanometer size, right, And here's 0:14:40.680 --> 0:14:44.240 where another challenge comes in, because when you get down 0:14:44.240 --> 0:14:47.920 to this size, this this nanometer scale, you're starting to 0:14:48.000 --> 0:14:54.000 encounter some pretty funky quantum physics problems, quantum mechanics problems. 0:14:55.000 --> 0:14:58.360 You're talking about electron tunneling. That would problem, that would 0:14:58.360 --> 0:15:00.400 be a big one. Yeah, because I mean, of course, 0:15:00.440 --> 0:15:02.520 and we've talked about electron tunneling before as well, so 0:15:02.680 --> 0:15:04.960 long time listeners will know what we're talking about here. 0:15:05.320 --> 0:15:09.920 Electrons have this wacky little way of apparently defying the 0:15:10.000 --> 0:15:13.000 laws of physics as we understand them. Uh. Actually it's 0:15:13.000 --> 0:15:16.640 not entirely true quantum physics. It makes perfect sense. Classic physics, 0:15:16.680 --> 0:15:19.000 it makes no sense at all. So on my scale 0:15:19.120 --> 0:15:21.480 where I you know, if I drop something, it falls 0:15:21.560 --> 0:15:24.320 and then it hits something and it stops. That makes 0:15:24.320 --> 0:15:27.280 sense to me, Right, that's the world I grew up in. 0:15:28.160 --> 0:15:30.560 If I were to drop something and it would pass 0:15:30.720 --> 0:15:34.000 through the floor beneath me without making a hole and 0:15:34.040 --> 0:15:38.520 then continue to go down, I'd say, huh, that's strange. 0:15:39.000 --> 0:15:42.120 But on the quantum world, not so much. Those wacky 0:15:42.120 --> 0:15:45.280 electrons Tuesdays at eight. Yes, so electrons, if if a 0:15:45.320 --> 0:15:48.480 barrier is thin enough, and we're talking about a couple 0:15:48.520 --> 0:15:52.320 of nanometers wide or thick, I should say if if 0:15:52.760 --> 0:15:56.840 if it's thin enough and electron can tunnel through, and 0:15:56.920 --> 0:16:00.400 that it passes through that barrier as if the barrier 0:16:00.440 --> 0:16:03.280 we're not there. It doesn't make a hole. And in 0:16:03.320 --> 0:16:05.840 a way, it almost is like it's on one side 0:16:05.840 --> 0:16:07.600 of the barrier at one moment and on the other 0:16:07.640 --> 0:16:10.720 side of the barrier the next um. But that's one 0:16:10.760 --> 0:16:13.360 of the problems. And then what Intel has found is 0:16:13.360 --> 0:16:16.400 they found that by using different materials, by by switching 0:16:16.800 --> 0:16:21.360 their transistory gates to other kinds of elements, that they 0:16:21.360 --> 0:16:25.200 were more resistant to electron tunneling. And you don't want 0:16:25.200 --> 0:16:28.840 electron tunneling, by the way, because leaking electrons means that 0:16:28.920 --> 0:16:33.600 transistor is pretty much useless. Transistors are all about governing 0:16:33.800 --> 0:16:36.480 the flow of electrons and if you can't stop them, 0:16:36.600 --> 0:16:42.720 then the transistor is always open essentially. Sorry. I think 0:16:42.720 --> 0:16:46.040 two on semiconductors, which we've discussed in the past. Two 0:16:46.040 --> 0:16:51.520 because semiconductors are uh you know, essential to running pretty 0:16:51.600 --> 0:16:55.520 much all kinds of electronics, including computers, um and it's 0:16:55.560 --> 0:16:59.400 all about controlling using certain materials to control the flow 0:16:59.400 --> 0:17:01.240 of electron So you knew, in order to have a 0:17:02.080 --> 0:17:05.320 computer processor to function as it needs to, it also 0:17:05.480 --> 0:17:07.959 it needs to be able to control when and where 0:17:07.960 --> 0:17:14.720 the electrons and the uh inside the actual transistors are going. Yes, 0:17:15.000 --> 0:17:17.240 So I mean that's it's it's crucial. And if you 0:17:17.359 --> 0:17:21.439 start having electrons going willy nilly and put it, your 0:17:21.480 --> 0:17:24.000 processor is just gonna have errors. It's not going to 0:17:24.040 --> 0:17:27.880 be able to compute things because it can't. You know, 0:17:27.960 --> 0:17:30.720 the data that's working from the operations that's performing are 0:17:30.800 --> 0:17:34.720 all going to be affected by that electron leakage. So 0:17:35.080 --> 0:17:36.679 you have to find a way to reduce that and 0:17:36.760 --> 0:17:39.639 Intel has been doing that by experimenting with different materials. 0:17:39.760 --> 0:17:43.240 So we left off at Westmere, which was the tick 0:17:43.359 --> 0:17:46.480 we talked about going down to thirty tick we talked about. 0:17:46.640 --> 0:17:49.679 It was the tick we talked about right and the 0:17:49.760 --> 0:17:52.960 next one you actually wrote about yourself again. Yeah, it 0:17:53.320 --> 0:17:55.399 turns out I write a lot about the talks. I 0:17:55.400 --> 0:17:58.520 haven't written about the ticks. But the next talk is, 0:17:58.520 --> 0:18:01.520 of course, as a time we're recording this podcast, the 0:18:01.560 --> 0:18:07.600 most recent Intel processor the sandy Bridge processor, and sandy 0:18:07.640 --> 0:18:10.480 Bridge is again it's at that thirty two nanometer scale, 0:18:10.480 --> 0:18:12.359 because remember it's going to be on the same scale 0:18:12.359 --> 0:18:15.520 as the Tick before it, but it's got a different layout. 0:18:15.600 --> 0:18:18.760 It's no longer based on the new halam Mark micro architecture. 0:18:18.800 --> 0:18:22.160 It's got its own micro architecture, which includes a section 0:18:22.359 --> 0:18:26.439 on the chip specifically dedicated to graphics processing, which that 0:18:26.560 --> 0:18:29.440 was the big thing that set it apart from its predecessors. 0:18:29.760 --> 0:18:34.600 It also has a very small bowling alley. Alright, I 0:18:34.640 --> 0:18:36.360 don't even know where you're going with that joke. I'm 0:18:36.359 --> 0:18:39.159 just gonna I was just imagining I was going with 0:18:39.200 --> 0:18:43.560 your metaphor earlier in thinking about the building. You know, 0:18:43.600 --> 0:18:45.960 it's got its own graphics processor and a bowling alley. Okay, 0:18:45.960 --> 0:18:48.240 I'm gonna I'm gonna call that on a gutter ball 0:18:48.359 --> 0:18:53.959 right now. Yes, the sandy Bridge micro architecture is is 0:18:54.240 --> 0:18:56.640 brand new as of the time we're recording this. Yes, 0:18:57.000 --> 0:19:00.760 so new that there are problems with other chips associated 0:19:00.800 --> 0:19:03.880 with sandy Bridge, not sandy Bridge itself, we should point out, 0:19:04.240 --> 0:19:05.960 but we can talk a little bit about that. It's 0:19:06.000 --> 0:19:09.160 only it's only vaguely related to what we're chatting about 0:19:09.200 --> 0:19:12.480 in this in this podcast. So the neat thing about 0:19:12.480 --> 0:19:15.800 the graphics processing, of course, is that this means that 0:19:15.840 --> 0:19:18.760 if you have a computer with a sandy Bridge processor, 0:19:18.840 --> 0:19:20.400 especially if you have one of the faster ones, because 0:19:20.400 --> 0:19:23.119 they'd come in different flavors. You know, there's some that 0:19:23.200 --> 0:19:27.040 can have there's some that have two cores and can 0:19:27.119 --> 0:19:30.320 run up to four threads of data. And then there 0:19:30.440 --> 0:19:32.200 it goes all the way up to I think four 0:19:32.240 --> 0:19:35.399 cores that can run eight threads of data. Um, it 0:19:35.440 --> 0:19:37.040 may even go higher than that. I have to look 0:19:37.040 --> 0:19:40.880 it up again. But the the you know, you get 0:19:40.920 --> 0:19:43.520 one of the faster ones and you get this graphics 0:19:43.520 --> 0:19:46.480 processing built onto the chip, it means that you may 0:19:46.520 --> 0:19:51.320 not need a dedicated graphics processing unit to add to 0:19:51.320 --> 0:19:53.040 your computer in order to play some of the more 0:19:53.160 --> 0:19:58.400 advanced video games or to do things like video editing. Um, 0:19:58.440 --> 0:20:03.080 you you might not need more additional power because you've 0:20:03.119 --> 0:20:05.960 got everything you need on that one micro chip, which 0:20:06.000 --> 0:20:08.720 is pretty fascinating. I mean, that chip is tiny, and 0:20:08.760 --> 0:20:11.760 to think that it does the the equivalent of a 0:20:11.800 --> 0:20:17.320 dedicated graphics processing card is a phenomenal. Now, granted, I'm 0:20:17.359 --> 0:20:20.000 sure there are going to be games out there that 0:20:20.040 --> 0:20:22.080 if you crank them to the highest setting, you're still 0:20:22.160 --> 0:20:25.120 gonna want your own graphics processing card because it's not 0:20:25.800 --> 0:20:28.399 it's not able to you know, take over the entire load. 0:20:29.680 --> 0:20:32.560 I had a feeling when you said that that someone 0:20:32.680 --> 0:20:36.680 will write in to to tell you that that, uh, 0:20:36.760 --> 0:20:38.320 you know, that is not the case that you're if 0:20:38.320 --> 0:20:41.080 you're going to be running a high frame rate first 0:20:41.119 --> 0:20:43.199 person shooter or you know, something with a lot of 0:20:43.200 --> 0:20:46.399 detailed games like that, uh, that you're not going to 0:20:46.480 --> 0:20:49.040 want that. And yes, we're we're aware of that. Yeah. 0:20:49.040 --> 0:20:52.760 But if you're playing you know, Crush the Castle, yeah, 0:20:53.440 --> 0:20:56.359 uh well, and and it's funny too. I was thinking 0:20:56.920 --> 0:21:00.720 about the most recent release of the Mac os ten, 0:21:01.080 --> 0:21:03.359 which as at this point was snow Leppard and uses 0:21:03.800 --> 0:21:06.960 the Grand Central technology, which sort of coopts your graphics 0:21:07.000 --> 0:21:10.320 processor if it if it isn't busy with something. Uh. 0:21:10.520 --> 0:21:16.440 So it seems like the uh operating system manufacturers and 0:21:16.600 --> 0:21:19.240 the chip manufacturers are both sort of aware, you know what, 0:21:19.280 --> 0:21:22.639 we could probably be using you know, one chip to 0:21:22.720 --> 0:21:25.480 do multiple things, and if you have multiple chips, you 0:21:25.480 --> 0:21:28.360 can have them sort of you know, pinch hit when 0:21:28.400 --> 0:21:31.680 required in other areas. So it seems to make the 0:21:32.200 --> 0:21:36.919 architecture more computer itself architecture, not the the chip architecture 0:21:37.000 --> 0:21:40.679 more flexible because that way, say you have a Sandy 0:21:40.720 --> 0:21:44.960 Bridge chip which has the onboard ability to graphics process 0:21:45.119 --> 0:21:47.760 or process graphics. Sorry. Uh, and you have a GPU 0:21:47.880 --> 0:21:50.439 as well, it seems that that you would have a 0:21:50.480 --> 0:21:54.879 lot of ability for your computer to use those computing cycles, 0:21:55.359 --> 0:21:57.480 you know, in both if it's if the operating system 0:21:57.520 --> 0:22:00.520 is capable of handling or you know, routing those instructions 0:22:00.520 --> 0:22:03.080 to different places like that. Yeah, and you've got some 0:22:03.320 --> 0:22:07.560 GPU manufacturers that are looking into doing CPUs now. So 0:22:08.040 --> 0:22:11.320 you know, while we're seeing Intel kind of push its 0:22:11.320 --> 0:22:14.960 way into the graphics processing unit world just by incorporating 0:22:14.960 --> 0:22:17.520 it into the chips, we're seeing the opposite from the 0:22:17.560 --> 0:22:21.000 graphics processing world as well. So uh yeah, there's a 0:22:21.000 --> 0:22:23.480 lot of competition in this space, and that's one of 0:22:23.480 --> 0:22:25.959 the things. As you make these elements smaller and smaller, 0:22:26.000 --> 0:22:28.520 you can you can really diversify what they can do. 0:22:29.000 --> 0:22:30.719 So let's let's look a little bit into the future, 0:22:31.160 --> 0:22:33.160 all right and look at some of the chips will 0:22:33.200 --> 0:22:36.400 be coming out over the next few years. So following 0:22:36.480 --> 0:22:40.200 sandy Bridge will be ivy Bridge, which of course will 0:22:40.200 --> 0:22:43.160 be another tick. So we're talking about a reduction in size. 0:22:43.720 --> 0:22:46.440