WEBVTT - TechStuff Looks at Supercomputers 0:00:00.280 --> 0:00:02.840 Brought to you by the reinvented two thousand twelve camera. 0:00:03.160 --> 0:00:08.800 It's ready. Are you get in touch with technology? With 0:00:08.960 --> 0:00:17.600 tech Stuff from House Touffi dot com. Hello again, everyone, 0:00:17.640 --> 0:00:20.000 and welcome to tech Stuff. My name is Chris Poulette, 0:00:20.040 --> 0:00:22.840 and I am an editor at how Stuff works dot com. 0:00:22.840 --> 0:00:26.320 Sitting across from me in a cape and tights is 0:00:26.440 --> 0:00:32.720 senior writer Jonathan Strickland. Hey there, citizen. So, so, Chris, 0:00:32.880 --> 0:00:34.720 if I were to ask you, just off the top 0:00:34.760 --> 0:00:37.839 of your head, how would you define a supercomputer? What 0:00:37.920 --> 0:00:41.040 would you say? Well, if I hadn't already made the joke, 0:00:41.080 --> 0:00:42.760 I would have said it was a computer in the 0:00:42.800 --> 0:00:45.880 Cape and tights, But no, I'm honestly I would say 0:00:46.400 --> 0:00:50.520 supercomputer is a computer that can do a lot more 0:00:50.840 --> 0:00:54.480 calculations in a shorter period of time than the machines 0:00:54.560 --> 0:00:56.560 sitting on our desktop. Yeah. I think of it as 0:00:56.560 --> 0:00:59.640 sort of the bleeding edge of what a computer is 0:00:59.680 --> 0:01:03.840 capable of doing. Something that that still feels fills a room, 0:01:03.880 --> 0:01:07.040 even though typical computers these days don't need to fill 0:01:07.080 --> 0:01:09.440 a room because it's that big, it still has that 0:01:09.560 --> 0:01:14.080 much computing power, right right, And the term comes from 0:01:14.319 --> 0:01:18.040 the nineteen sixties and uh. In order to really kind 0:01:18.040 --> 0:01:22.479 of understand the the the span of this, I think 0:01:22.800 --> 0:01:24.760 I was going to talk a little bit about the 0:01:25.319 --> 0:01:28.360 last computer I could find that was a powerful computer 0:01:28.880 --> 0:01:33.920 that existed before people started talking about super computers, which 0:01:34.000 --> 0:01:39.120 was the IBM seventy thirty stretch. Yes, that was the 0:01:39.160 --> 0:01:45.080 one that was made with elastic. Yes, ye gain a 0:01:45.080 --> 0:01:47.440 couple of pounds. Your computer can still you know fit. 0:01:47.800 --> 0:01:51.120 It was Mr. Fantastic of the computer world. No, because 0:01:51.120 --> 0:01:53.840 it was not a super computer. It took up two 0:01:53.880 --> 0:01:57.280 thousand square feet back in the day, this being the 0:01:57.920 --> 0:02:01.720 early sixties, Higger than my two thousand square feet. It 0:02:01.760 --> 0:02:05.480 cost thirteen million dollars, which if you were to translate 0:02:05.520 --> 0:02:11.920 to today's cash, would be ninety one million dollars. It's 0:02:11.919 --> 0:02:15.080 a lot of cash. So that was the fastest computer 0:02:15.160 --> 0:02:18.120 at the time until a fellow named Seymour Roger Craig 0:02:18.240 --> 0:02:22.200 showed up. I s Mr Craze. Yeah. Craig ends up 0:02:22.200 --> 0:02:25.960 being a big name in supercomputers, particularly in the sixties, 0:02:26.000 --> 0:02:29.880 seventies and up to the mid eighties. That was the 0:02:30.000 --> 0:02:34.959 name in supercomputers. And he was working for a company 0:02:35.000 --> 0:02:39.120 called Engineering Research Associates or E r A, which actually 0:02:39.200 --> 0:02:43.880 grew out of a naval operation, um being the U. S. Navy, 0:02:43.919 --> 0:02:46.800 not belly Button. I was gonna you were looking at 0:02:46.840 --> 0:02:50.400 me like joke. No, No, not that it was a 0:02:50.520 --> 0:02:53.640 Navy project. How about that as in as in the 0:02:53.680 --> 0:02:57.160 military force, not the color It was a Navy project 0:02:57.400 --> 0:03:00.280 that was all about code breaking, all right. So there 0:03:00.320 --> 0:03:03.080 was this project about code breaking that eventually kind of 0:03:03.120 --> 0:03:06.160 spun off and became an actual company all on its 0:03:06.160 --> 0:03:09.720 own called Engineering Research Associates, and it branched out beyond 0:03:09.800 --> 0:03:12.320 code breaking, although it took all the code breaking work 0:03:12.320 --> 0:03:16.600 it could get. Yeah, we talked about the Enigma UM 0:03:16.680 --> 0:03:20.560 some episodes back UM and we were talking about the 0:03:20.639 --> 0:03:25.359 bomb UM and yeah, those early uh that was really 0:03:25.360 --> 0:03:30.000 the early application for supercomputers was you know, needing to 0:03:30.120 --> 0:03:33.519 crunch a lot of data very quickly, and there weren't 0:03:34.440 --> 0:03:36.440 There weren't the kind of applications that we have now. 0:03:36.480 --> 0:03:38.360 We'll get into that, I'm sure in a few minutes, 0:03:38.400 --> 0:03:41.680 but but yeah, I mean that was you know, why 0:03:41.720 --> 0:03:44.880 would you need a supercomputer? You know? That was That's 0:03:44.920 --> 0:03:47.720 probably about the only thing I could think of where 0:03:47.720 --> 0:03:50.640 people were needing to crunch that kind of information as 0:03:50.720 --> 0:03:56.680 quickly as possible and defense. Typically, especially with the early supercomputers, 0:03:56.720 --> 0:04:01.880 they were really designed for very specialized suting, so not 0:04:01.920 --> 0:04:05.880 necessarily specialized from the ground up for a one particular 0:04:05.920 --> 0:04:08.120 type of computing, but they were. They were not meant 0:04:08.120 --> 0:04:11.640 to be general computers. They were meant to do tip 0:04:12.720 --> 0:04:16.880 no admiral computers, that's true. Uh No, they were. They 0:04:16.920 --> 0:04:19.640 were meant to do a specific task very very well. 0:04:19.720 --> 0:04:22.640 And that's that's all they were meant to do. Um. Now, 0:04:24.160 --> 0:04:27.200 Craig had an interesting philosophy, he said, and this is 0:04:27.240 --> 0:04:29.760 this is a quote from him. He said, anyone can 0:04:29.800 --> 0:04:32.840 build a fast CPU. The trick is to build a 0:04:32.880 --> 0:04:36.919 fast system. And that was the secret to Craig creating 0:04:36.920 --> 0:04:40.640 the first supercomputer. He realized that if you created a 0:04:40.640 --> 0:04:44.559 processor that was really really fast, that did not matter 0:04:44.680 --> 0:04:48.240 if it couldn't get the data it needed to execute 0:04:48.279 --> 0:04:53.200 operations upon fast enough. So he saw the need to 0:04:53.200 --> 0:04:57.120 create a system that would move data through very very quickly, 0:04:57.200 --> 0:04:59.880 not just processed data, but move it so that means 0:05:00.040 --> 0:05:02.280 needs a lot of memory, It needs a very fast 0:05:02.560 --> 0:05:05.680 pathway from memory to processor. There are a lot of 0:05:05.720 --> 0:05:07.599 pieces that have to be put in place, and he 0:05:07.720 --> 0:05:11.760 saw this very early on, and so using that philosophy, 0:05:11.839 --> 0:05:15.000 he designed a computer back in nineteen sixty two that 0:05:15.240 --> 0:05:19.280 was called the c d C sixty six hundred. Now 0:05:19.360 --> 0:05:24.800 CDC stands for Controlled Data Corporation. Yeah, um, uh e 0:05:25.040 --> 0:05:28.679 R A was taken over by Remington Rand UM and 0:05:28.720 --> 0:05:31.680 that's uh, that's the name I remember because you know, 0:05:32.279 --> 0:05:35.120 UH still remember a lot of those old machine names 0:05:35.720 --> 0:05:40.279 UM from stuff that I found in my uh dad's collection. 0:05:40.320 --> 0:05:43.279 Of course, he was, you know, a mechanical engineer UM 0:05:43.400 --> 0:05:46.239 before he retired, and you know, so he was interested 0:05:46.240 --> 0:05:48.479 in all kinds of machines and I didn't know what 0:05:48.520 --> 0:05:50.120 I was looking at at the time, of course, you know, 0:05:50.160 --> 0:05:53.480 but they were all these UM science and computing magazines 0:05:53.520 --> 0:05:56.880 laying around and that name I recognized also UNITIS because 0:05:57.120 --> 0:06:01.000 Remington Ran became Unities UM, and probably a lot more 0:06:01.000 --> 0:06:03.960 of our listeners are familiar with that name. But he 0:06:04.480 --> 0:06:07.880 partnered with William Norris to start Controlled out of Corporation 0:06:08.480 --> 0:06:12.359 UM back in ninety seven. UM. And really at that point, 0:06:12.720 --> 0:06:18.240 the UNIVAC from Remington Rand and IBM were the computing companies. Yeah, 0:06:18.360 --> 0:06:20.520 you know, IBM has been the heavyweight for so long, 0:06:21.480 --> 0:06:25.719 but CDC was the first uh you know, upstart to 0:06:25.839 --> 0:06:30.400 really make a dent in there, uh stranglehold on the industry. 0:06:30.480 --> 0:06:33.840 And and Craig wanted to join CDC fairly early on, 0:06:33.960 --> 0:06:38.920 but apparently he was needed for a project UM that 0:06:39.360 --> 0:06:41.919 would not let him leave exactly what he wanted to. 0:06:42.000 --> 0:06:43.839 So once he did leave, that's when he designed the 0:06:43.880 --> 0:06:49.840 CDC sixty DRED, which was officially announced in nineteen sixty four, 0:06:50.240 --> 0:06:53.400 so designed in sixty two, announced two years later, and 0:06:53.400 --> 0:06:57.039 it was the first commercially successful supercomputer, with a price 0:06:57.120 --> 0:07:00.080 tag of between seven and eight million dollars, sometimes going 0:07:00.120 --> 0:07:02.720 up as high as ten million, depending upon the configuration 0:07:02.800 --> 0:07:05.680 that the customer wanted. UM Now, in today's cash, that 0:07:05.720 --> 0:07:10.280 would equal about sixty million dollars, so thirty one million 0:07:10.320 --> 0:07:16.280 dollars cheaper in today's money than the Stretch computer, and 0:07:16.560 --> 0:07:20.560 it was actually much more powerful. It had four hundred 0:07:20.720 --> 0:07:25.800 thousand transistors and one hundred miles of wiring, and it 0:07:25.880 --> 0:07:28.880 was the size of about four filing cabinets, so it 0:07:28.960 --> 0:07:32.600 was also significantly smaller than the Stretch, didn't take up 0:07:32.600 --> 0:07:35.960 two thousand square feet. The clock speed was around a 0:07:36.040 --> 0:07:41.080 hundred nanoseconds, and it had sixty five thousand sixty bit 0:07:41.240 --> 0:07:44.200 words of memory. So this is kind of an odd 0:07:44.200 --> 0:07:49.360 time in computing. We hadn't really settled on the thirty 0:07:49.400 --> 0:07:54.800 two sixty four bit kind of model. This was before that. UM. 0:07:54.800 --> 0:07:57.280 It also used six high speed drums as sort of 0:07:57.320 --> 0:07:59.880 a temporary storage area. It had a central storage that 0:08:00.080 --> 0:08:03.080 used magnetic tape, and it used the four trans sixty 0:08:03.160 --> 0:08:07.800 six compiler. UM the equivalent to today's machines means that 0:08:07.840 --> 0:08:13.200 it would have about a ten mega hurts processor. Yeah, 0:08:13.440 --> 0:08:16.200 well that could work up to forty mega hurts and speed. Well, 0:08:16.280 --> 0:08:22.600 it could do a three million floating point operations per second. Yeah, 0:08:22.640 --> 0:08:25.360 so three million. That would be a mega flop. Three 0:08:25.360 --> 0:08:28.320 mega flops, right, So we're gonna get into lots of 0:08:28.320 --> 0:08:34.439 different flop terms later as well, and they get incredibly huge. 0:08:34.480 --> 0:08:36.120 H Of course, you have to keep it cool because 0:08:36.120 --> 0:08:39.720 otherwise it breaks out into a flop sweat. That's true. Uh, well, 0:08:39.720 --> 0:08:41.240 not the flop sweat part, but you do have to 0:08:41.280 --> 0:08:44.240 keep it cool. As we know electronics, when you're running 0:08:44.240 --> 0:08:47.199 electricity through them, one of the byproducts is heat, and heat, 0:08:47.360 --> 0:08:50.040 as it turns out, is not a great thing for 0:08:50.360 --> 0:08:56.720 electronic components. It can make stuff expand contacts can lose connections, 0:08:56.840 --> 0:09:00.600 so that stuff starts to malfunction and entire system could 0:09:00.600 --> 0:09:05.840 shut down. So the CDC STRED had a cooling system 0:09:05.880 --> 0:09:11.040 that was provided by a special chemical free on. Really, yeah, 0:09:11.080 --> 0:09:13.280 they used free on to cool the system. In fact, 0:09:13.360 --> 0:09:16.000 it was they would use free on for a while 0:09:16.840 --> 0:09:19.800 before finally having to switch to a different coolant because 0:09:20.600 --> 0:09:24.400 free on just wasn't efficient enough. Eventually, now at the 0:09:25.040 --> 0:09:29.560 hundred it was still doing the job. So Kari was 0:09:29.880 --> 0:09:33.079 also an innovator in another way. UM, the stretch IBM 0:09:33.120 --> 0:09:38.320 stretched the UM was sort of a hybrid machine. They 0:09:38.360 --> 0:09:42.600 had transistors and vacuum tubes in it UM, and that's 0:09:42.800 --> 0:09:45.720 I think why one of the reasons why craze machines 0:09:45.800 --> 0:09:49.400 were smaller. The sixteen O four, which proceeded the sixty 0:09:49.480 --> 0:09:53.600 six hundred UM, was the one of the very first 0:09:53.640 --> 0:10:00.120 to use transistors only. So there was also a transistor 0:10:00.160 --> 0:10:02.520 machine and so it would take up a lot less 0:10:02.520 --> 0:10:04.360 space than the vacuum tubes. And I would imagine that 0:10:05.000 --> 0:10:07.439 based on my knowledge, my personal knowledge of vacuum tubes, 0:10:07.640 --> 0:10:10.240 might have been a little cooler simply because of that. Yeah, 0:10:10.240 --> 0:10:12.000 I would imagine that they wouldn't have had to have 0:10:12.280 --> 0:10:15.960 as dramatic and a c system to keep the the 0:10:16.040 --> 0:10:19.560 room bearable because vacuum tubes do put off a lot 0:10:19.600 --> 0:10:25.040 of heat. Um Another interesting IBM CDC connection here is 0:10:25.080 --> 0:10:28.880 that Thomas Watson Jr. Which was IBM s C. He 0:10:29.080 --> 0:10:32.760 was IBM CEO at the time, wrote a famous memo 0:10:33.520 --> 0:10:38.240 that time two IBM employees, and he said, last week 0:10:38.440 --> 0:10:42.760 Controlled Data announced the sixty system. I understand that in 0:10:42.800 --> 0:10:45.960 the laboratory developing the system, there are only thirty four people, 0:10:46.080 --> 0:10:49.760 including the janitor. Of these fourteen our engineers and for 0:10:49.920 --> 0:10:55.319 our programmers. Contrasting this modest effort with our vast developmental activities, 0:10:55.440 --> 0:10:57.840 I failed to understand why we have lost our industry 0:10:57.920 --> 0:11:00.439 leadership position by letting someone else offer the world's most 0:11:00.440 --> 0:11:07.000 powerful computer. Craize's response was a reportedly, well, there's your problem. Essentially, 0:11:07.000 --> 0:11:10.960 Craig was saying that, you know, perhaps IBMS approach it 0:11:11.000 --> 0:11:17.360 was a little burdened by size that IBM had grown 0:11:17.440 --> 0:11:21.840 so large that to manage a project like this was 0:11:22.160 --> 0:11:24.199 very difficult to do because it was just too big. 0:11:25.440 --> 0:11:28.960 So that's an interesting idea that an organization needed to 0:11:29.040 --> 0:11:32.760 be kind of small and nimble in order to pull 0:11:32.800 --> 0:11:36.760 something off like creating the world's fastest computer. He followed 0:11:36.840 --> 0:11:41.320 up the c d C sundred with which had a 0:11:41.440 --> 0:11:44.839 sixty five thousand, five hundred thirty six sixty bit word 0:11:44.920 --> 0:11:47.240 memory and a clock speed of twenty seven nano seconds 0:11:47.600 --> 0:11:51.720 uh and actually in practice ran about five times faster 0:11:51.880 --> 0:11:55.599 than the six. But then Cray left c d C 0:11:56.280 --> 0:12:00.960 and he formed his own company, Cray Research two and 0:12:01.040 --> 0:12:04.760 in nine six he introduced the Kray one, which if 0:12:04.760 --> 0:12:08.040 you've ever heard the craze supercomputer, that's what this is. 0:12:08.120 --> 0:12:10.720 It's the crazy one was the first of those. It 0:12:10.840 --> 0:12:14.360 had a clock speed of a well, it's processor ran 0:12:14.440 --> 0:12:18.800 at eighty mega hurts and back at this time these 0:12:18.800 --> 0:12:23.520 supercomputers were still using a single CPU, so that was 0:12:23.600 --> 0:12:26.440 kind of interesting to these were single CPU systems. So 0:12:26.440 --> 0:12:30.079 it had eighty mega hurts processor, sixty four bit system, 0:12:30.120 --> 0:12:32.720 it ran at a hundred thirty six mega flops, so 0:12:32.800 --> 0:12:36.160 one or thirty six million floating operations per second, and 0:12:36.200 --> 0:12:39.080 it had one thousand, six hundred sixty two printed circuit 0:12:39.080 --> 0:12:43.880 boards that made up the components of this computer. It 0:12:43.960 --> 0:12:46.400 costs between five and eight million dollars, depending on how 0:12:46.440 --> 0:12:48.839 you wanted it set up, and in today's cash, that's 0:12:48.880 --> 0:12:52.240 about twenty five million dollars. So we see that the 0:12:52.360 --> 0:12:56.679 processor speed is increasing and the price is coming down. 0:12:57.080 --> 0:13:00.080 Often the size of the computer is decreasing as well, 0:13:00.120 --> 0:13:04.920 although that that also flip flops over the years because 0:13:05.360 --> 0:13:09.200 while the solid state electronics definitely brought the size down, 0:13:09.880 --> 0:13:16.640 eventually the way we pack in more speed requires more space. 0:13:16.880 --> 0:13:21.319 But we'll get into that. Okay, So after the Cray 0:13:21.400 --> 0:13:26.959 one came the Cray x MP in Yeah, this is uh, 0:13:27.000 --> 0:13:30.640 this is interesting because Crai realized also in addition to 0:13:31.360 --> 0:13:34.480 the fact that he knew that the components, the all 0:13:34.520 --> 0:13:38.719 of the components, the entire machine was important and not 0:13:38.800 --> 0:13:42.679 just a processor, he also realized that, uh early on, 0:13:42.720 --> 0:13:47.480 that parallel processing could also speed things up. Um. Now 0:13:47.840 --> 0:13:50.680 it's common for us to have multi core processes in 0:13:50.720 --> 0:13:54.640 our desktop machines or laptops, or in fact, now we're 0:13:54.640 --> 0:13:57.880 starting to see them in our mobile devices. Um. But 0:13:59.120 --> 0:14:01.720 you know, at the time in the seventies and eighties, 0:14:01.760 --> 0:14:05.920 this was still something sort of newish, um, and it's 0:14:05.920 --> 0:14:09.280 not something that everybody realized. Uh. So the x MP 0:14:09.679 --> 0:14:15.679 actually was to Cray one computers linked together um and 0:14:16.280 --> 0:14:20.920 using those two machines together in a multiprocessing effort UM 0:14:21.000 --> 0:14:26.000 they could triple the performance of just one Cray one UM, 0:14:26.040 --> 0:14:31.160 which is something interesting to note. And uh, the Create 0:14:31.240 --> 0:14:34.440 two had four processors in the same machine, and that 0:14:34.520 --> 0:14:38.040 was the first to exceed one billion flops as Britainic 0:14:38.080 --> 0:14:40.920 it tells me. Yeah, uh, it actually could have up 0:14:40.960 --> 0:14:46.920 to eight CPUs c um. The these machines often over 0:14:47.000 --> 0:14:51.080 time were upgraded, so the initial step specs you would 0:14:51.120 --> 0:14:53.400 get when they were first released were one thing, and 0:14:53.440 --> 0:14:56.440 then by the end of the run of production they 0:14:56.440 --> 0:14:58.680 would be better. I mean, which makes sense. I mean 0:14:58.760 --> 0:15:00.720 we see that in computers all the time. We definitely 0:15:00.760 --> 0:15:03.400 we tend to call them different model numbers now, but 0:15:03.640 --> 0:15:07.000 the same sort of thing happens. So back in two 0:15:07.000 --> 0:15:10.880 you had this Cray XMP with a hundred five mega 0:15:10.920 --> 0:15:16.040 hurts CPUs running around two hundred megaflops each. Uh, and 0:15:16.600 --> 0:15:18.640 if they had up to four CPUs you could get 0:15:18.680 --> 0:15:22.680 eight hundred megaflops going and that was pretty impressive and 0:15:22.760 --> 0:15:25.320 had the equivalent, by the way of a hundred and 0:15:25.320 --> 0:15:29.040 twenty eight megabytes of RAM. So yeah, you think about 0:15:29.080 --> 0:15:33.400 that one or twenty eight megabytes of RAM in that 0:15:33.520 --> 0:15:37.720 was considered bleeding edge for a supercomputer. UM and the 0:15:37.880 --> 0:15:41.080 storage units for the Cray XMP were the size of 0:15:41.120 --> 0:15:43.160 a file cabinet and they could hold up to twelve 0:15:43.160 --> 0:15:46.640 gigs of storage because they have a flash in my 0:15:46.680 --> 0:15:49.960 bag with me eight gigs. Yeah, and you can find, 0:15:50.280 --> 0:15:52.880 you can find, you can find twenty gig or more 0:15:53.560 --> 0:15:55.520 flash drives, which you know, you think about that, that's 0:15:55.560 --> 0:15:57.960 something that is small enough for you to carry on 0:15:58.000 --> 0:16:00.600 a key chain. While back in nineteen eighty two you 0:16:00.640 --> 0:16:03.840 had a file cabinet sized device that could hold twelve 0:16:03.880 --> 0:16:07.920 gigs and that was considered massive, like a massive amount 0:16:07.920 --> 0:16:12.760 of information. So, yeah, time really does change things, doesn't it. 0:16:12.840 --> 0:16:15.760 So yeah, the Cry two, that's when they switched from 0:16:16.200 --> 0:16:21.720 free On to Flora Nert as their coolant. I'm sorry, 0:16:21.720 --> 0:16:24.600 but that sounds like a made up alien name from 0:16:24.640 --> 0:16:27.800 a from a an animated movie. Technically all names are 0:16:27.840 --> 0:16:35.800 made up. I know, that's I just blew your mind. 0:16:35.920 --> 0:16:41.760 What if there were no hypothetical questions? Turn on the yes. 0:16:42.280 --> 0:16:45.320 So the Floria Nert the reason why they switched was 0:16:45.360 --> 0:16:48.200 because they had at that point packed the components so 0:16:48.320 --> 0:16:50.920 tightly together that free on was not efficient enough to 0:16:50.960 --> 0:16:53.360 cool them, so they switched from free on to Flora 0:16:53.400 --> 0:16:58.200 Nert and it's a little Floria Nert I've had around somewhere. 0:16:58.360 --> 0:17:00.760 Then they also had to figure a new way to 0:17:00.880 --> 0:17:04.000 access the memory on the Create too, because at this 0:17:04.040 --> 0:17:06.560 point they had reached that that point that Krey had 0:17:06.560 --> 0:17:10.880 mentioned earlier about creating a CPU that can process information 0:17:10.960 --> 0:17:15.119 faster than it can pull information in. So they found 0:17:15.240 --> 0:17:19.639 they would actually dedicate processors to getting data from memory 0:17:19.680 --> 0:17:26.080 and funneling it into the central processing units. And UH, 0:17:26.280 --> 0:17:28.760 this was this was really important. It was what kind 0:17:28.760 --> 0:17:32.680 of led the way into threading and and loading memory 0:17:33.520 --> 0:17:37.199 CPUs that have that capability to load information from memory 0:17:37.320 --> 0:17:40.600 preloading things that kind of came out of this work. 0:17:40.840 --> 0:17:43.480 In fact, a lot of the UH, the advances that 0:17:43.520 --> 0:17:47.240 we see in personal computers UM are really possible because 0:17:47.240 --> 0:17:49.600 of the pioneering work that was done in supercomputers. It 0:17:49.680 --> 0:17:52.960 was stuff that that found its way from the engineering 0:17:53.040 --> 0:17:59.160 of supercomputers into personal computers, often a completely different sense 0:17:59.200 --> 0:18:03.320 of scale, but a similar approach. Now after the Create too, 0:18:04.160 --> 0:18:09.879 that's when Japan started to produce some supercomputers that were 0:18:10.000 --> 0:18:13.000 UH that were actually faster than anything that was being 0:18:13.000 --> 0:18:16.200 produced in the United States. So up until this point, 0:18:16.280 --> 0:18:19.840 it was all the US that was they dominated that, 0:18:19.840 --> 0:18:24.440 that country dominated the supercomputer industry. But in n six, 0:18:24.840 --> 0:18:28.920 so this is you know again, the Kray craze, if 0:18:28.960 --> 0:18:31.760 you will, lasted from the sixties all the way into 0:18:31.760 --> 0:18:35.480 the eighties. Well ninety six, Japan introduced the s R 0:18:35.680 --> 0:18:40.040 twenty two oh one, which had two thousand forty eight processors. 0:18:40.320 --> 0:18:43.760 So remember create two. That was up to eight processors. 0:18:44.760 --> 0:18:47.520 The s R two two oh one two thousand forty 0:18:47.560 --> 0:18:51.840 eight processors. I count two thousand forty more processors with 0:18:51.880 --> 0:18:55.480 that computer. Then with the Cray, do my math could 0:18:55.480 --> 0:18:58.399 be off in an English major and it could it 0:18:58.400 --> 0:19:02.200 could have up to six hi flops of processing. That's 0:19:02.280 --> 0:19:05.400 kind of crazy. Um yeah. I also I also feel 0:19:05.400 --> 0:19:07.520 like we would be remissed to mention the efforts of 0:19:07.840 --> 0:19:13.200 Danny Hillis um W. Daniel Hillis was a grad student 0:19:13.240 --> 0:19:16.320 at m i T. Massachusetts Institute of Technology when he 0:19:16.800 --> 0:19:20.560 realized that distributing computing was the way of the future, 0:19:20.600 --> 0:19:24.359 if you will. UM he started thinking Machines Corporation in 0:19:26.040 --> 0:19:29.000 UM and this CM one, which was the first of 0:19:29.080 --> 0:19:32.080 his machines to come out in eight five. Um it 0:19:32.200 --> 0:19:40.320 had sty six one bit processors grouped sixteen to a chip. Interesting, 0:19:41.240 --> 0:19:47.320 that's a that's a really interesting approach tiny tiny processors. Yeah, 0:19:48.600 --> 0:19:53.040 so you know, but yeah, I didn't come across that 0:19:53.200 --> 0:19:55.399 my um my, my research, which is why this is 0:19:55.440 --> 0:19:58.400 actually really like I'm my mind is really as I'm 0:19:58.400 --> 0:20:00.880 thinking about that sort of archetype. Sure, that's really an 0:20:00.880 --> 0:20:03.800 interesting approach. Well, it's interesting too to see how different 0:20:04.480 --> 0:20:07.640 See Jonathan and I do our research separately on purpose 0:20:07.720 --> 0:20:10.679 so that we uh come up with different things on 0:20:10.720 --> 0:20:13.560 the cases. And um, so it's funny that that I 0:20:13.560 --> 0:20:16.119 would have come across that. Also. Well, I think of 0:20:16.600 --> 0:20:18.560 Danny hillis because I've seen his name a lot in 0:20:19.119 --> 0:20:22.399 things like along Now Foundation and people with he he 0:20:22.720 --> 0:20:25.640 hangs out with people like Stewart Brandon, Kevin Kelly, UM, 0:20:25.720 --> 0:20:30.200 fascinating people. But um anyway, yeah, that that's uh, that 0:20:30.280 --> 0:20:32.159 was one of his contributions. And you see that in 0:20:32.320 --> 0:20:35.959 again in today's machines. I mean, we have this, you know, 0:20:36.040 --> 0:20:39.160 with us every day. But you know this is uh, 0:20:39.240 --> 0:20:41.160 this is when we started to realize that you don't 0:20:41.200 --> 0:20:44.520 necessarily have to go buy More's Law and wait until 0:20:44.520 --> 0:20:46.879 next year's chip comes out with twice as many processors 0:20:46.880 --> 0:20:50.680 on it. You can you can do this by dividing 0:20:50.720 --> 0:20:53.119 up the work. Yeah, and and in fact, that's another 0:20:53.560 --> 0:20:57.280 good point about the SR two oh one, the computer 0:20:57.320 --> 0:21:00.600 from Japan, because, uh, in order to who use these 0:21:00.640 --> 0:21:04.520 two thousand forty eight processors, there was a new development 0:21:04.560 --> 0:21:08.320 in computer science which was called multiple instruction multiple data 0:21:08.440 --> 0:21:11.400 or m I m D. Now, this is the idea 0:21:11.520 --> 0:21:15.480 of being able to solve problems by pulling in information 0:21:15.640 --> 0:21:19.679 from from memory and feeding it to different processors that 0:21:19.760 --> 0:21:23.680 are all using different operations on that data to come 0:21:23.720 --> 0:21:28.320 to a single solution, not necessarily a single solution, but 0:21:28.400 --> 0:21:30.640 that's I'm using that as as an example for this 0:21:31.280 --> 0:21:35.760 for this explanation. So this m I m D approach 0:21:35.960 --> 0:21:40.520 is what allowed us to develop multi core processors, because 0:21:40.640 --> 0:21:43.760 in this case we're still talking about single processors that 0:21:43.800 --> 0:21:46.800 are all grouped together. Eventually we will get to the 0:21:46.800 --> 0:21:49.399 point where we have multi core processors where a single 0:21:49.440 --> 0:21:52.879 processor has multiple cores and each core can work on 0:21:53.040 --> 0:21:58.200 part of a problem or separate problems to solve things faster, 0:21:58.320 --> 0:22:01.560 to to get to a inclusion faster than they would 0:22:01.880 --> 0:22:05.560 if it was just one single processor, even if it 0:22:05.640 --> 0:22:09.919 was a really really fast processor working on a series 0:22:09.920 --> 0:22:12.639 of problems. So I always I always use this analogy. 0:22:13.080 --> 0:22:18.240 Imagine that you have one super smart math genius taking 0:22:18.240 --> 0:22:22.240 a math test, and the math genius is going through 0:22:22.280 --> 0:22:25.159 and solving all of these problems, and he or she 0:22:25.840 --> 0:22:30.560 is able to do this flawlessly, able to solve all 0:22:30.560 --> 0:22:32.280 the problems, but it takes a certain amount of time 0:22:32.320 --> 0:22:34.760 to get through the test. Then you get that same 0:22:34.800 --> 0:22:39.399 test to four above average math students. They're not geniuses, 0:22:39.440 --> 0:22:42.840 but they're there. They can hold their own. And you 0:22:42.960 --> 0:22:45.320 divide it up, say all right, you take this this 0:22:45.600 --> 0:22:47.840 one fourth of the test, you take this quarter, you 0:22:47.880 --> 0:22:50.119 take this quarter, and you take that quarter, and the 0:22:50.160 --> 0:22:53.199 four students together start to work. Those four students are 0:22:53.280 --> 0:22:55.600 very likely going to be able to finish the entirety 0:22:55.600 --> 0:22:57.800 of that test much faster, each of them working on 0:22:57.840 --> 0:23:00.840 a quarter of it, rather than the genius who is 0:23:00.840 --> 0:23:02.960 working on the full thing at the same time. Even 0:23:02.960 --> 0:23:05.199 though the genius is smarter and can work faster on 0:23:05.240 --> 0:23:09.359 each individual problem, collectively those four students are going to 0:23:09.400 --> 0:23:14.360 solve that test faster. That's the philosophy behind both grouping 0:23:14.640 --> 0:23:19.119 cores together and making them a parallel processing unit or 0:23:19.240 --> 0:23:22.320 taking a multi core approach to a CPU. Yep, and 0:23:22.400 --> 0:23:25.040 you can. You can thank Danny Hillis for figuring out 0:23:25.080 --> 0:23:29.919 the idea of massively parallel computing UM. But you know 0:23:29.960 --> 0:23:33.879 that that's a problem though too, because instead of having 0:23:34.240 --> 0:23:37.800 two machines running side by side and linked together, now 0:23:37.840 --> 0:23:39.960 you have to figure out how you're going to parse 0:23:40.040 --> 0:23:43.720 all those instructions between all those different processors. So you 0:23:43.760 --> 0:23:46.960 have to have the software or the operating system that 0:23:47.119 --> 0:23:52.760 will give instructions to each of the processors actively and 0:23:52.800 --> 0:23:55.840 direct essentially directing traffic. Yes, this is this is kind 0:23:55.840 --> 0:23:57.679 of like, it's not. It's not a simple thing to 0:23:57.680 --> 0:24:01.119 figure out. It reminds me of Intel's to talk approach 0:24:01.240 --> 0:24:05.800 to developing processors. You think of the TICK being the 0:24:05.840 --> 0:24:10.679 physical machinery that's going to do the processing, and you 0:24:10.720 --> 0:24:13.879 think of the talk as the software that's optimized to 0:24:13.960 --> 0:24:17.000 work on that physical hardware to make it really live 0:24:17.080 --> 0:24:19.520 up to its potential. And then you have another tick 0:24:19.640 --> 0:24:22.879 where you've got an advancement in the physical hardware, but 0:24:22.960 --> 0:24:26.119 perhaps the last generation of software isn't really optimized to 0:24:26.160 --> 0:24:28.840 work on that, so you have to make new software. 0:24:29.119 --> 0:24:31.359 This is a continuation. In fact, that's one of the 0:24:31.400 --> 0:24:36.560 things that people say is a barrier to artificial intelligence 0:24:36.560 --> 0:24:38.879 to the point of having a a computer that has 0:24:38.920 --> 0:24:43.200 self awareness. It's not necessarily that we can't reach the 0:24:43.240 --> 0:24:47.679 physical uh requirements we would need in order to have 0:24:47.680 --> 0:24:50.639 a computer be able to have some form of self awareness. 0:24:50.920 --> 0:24:55.000 It's the idea that we could throw as much horsepower 0:24:55.119 --> 0:24:57.480 at the problem as we wanted to. Without the software, 0:24:57.520 --> 0:25:02.400 it just won't happen anyway. Get back to supercomputers specifically, 0:25:02.440 --> 0:25:05.840