Transcript

[00:00:01] Chris: I don’t know if you know this about animals, but mostly they kind of resent us.

[00:00:07] Ned: I can’t imagine why. We’ve done nothing but be kind to them, especially the chickens.

[00:00:15] Chris: I don’t think that that’s even remotely accurate.

[00:00:18] Ned: I was trying to think of which animal we treated the worst, and chickens came to mind immediately.

[00:00:25] Chris: Mean, sadly, the correct answer, unfortunately, is pigs.

[00:00:30] Ned: Yeah, but they’re so delicious.

[00:00:34] Chris: It doesn’t make it better.

[00:00:36] Ned: No, it does, but it’s true. Wow. Yeah. I spent the last few days in Phoenix, and the friends that I was staying with had an orange tabby.

[00:00:50] Chris: Okay. Is that a type of pig?

[00:00:54] Ned: No.

[00:00:55] Chris: Then why are we talking about it?

[00:00:56] Ned: Well, her name was Penelope. All right, I’m back in. But because it was an orange tabby, I kept wanting to call her Jonesey. And if you get that you are of a certain age.

[00:01:14] Chris: Well, or you’re me, who has a cat who is named Jones, who is orange, that might.

[00:01:21] Ned: Have been part of it, too. So maybe it was bleed over from meeting your Jonesy, though. Your Jonesy tries to kill me every time I’m there one of these times. He will succeed, if no doubt, and I won’t even harbor any resentment towards him.

[00:01:40] Chris: Yeah, you’ll just be like, I get it.

[00:01:42] Ned: It’s like when the dog catches the car. What are you going to do now, Jonesy? Speaking of animals, so let me describe an experience that I had to you today, and let me gauge what your reaction would be.

[00:01:59] Chris: I’m already bored.

[00:02:01] Ned: Excellent. So we take our dog, our overactive lab, to daycare twice a week so she can get out her energy with other dogs.

[00:02:12] Chris: Right.

[00:02:13] Ned: Generally. Good plan. And I drop her off in the morning around 08:00. And there’s sort of an unspoken protocol. There’s a daycare door, and the handlers come out of the daycare door and get your dog. And the order in which you park is the order in which you will give your dog to them. Right.

[00:02:37] Chris: Okay. So if you’re, like, fifth, you are the fifth person that hands their dog to the handler.

[00:02:46] Ned: I think that would be the way that it has gone thus far. Now, technically, technically, you are supposed to call them when you arrive and wait in your car and tell them what space you’re in. But nobody does this because everybody drops off in the morning, and they just keep sending people out to grab the dogs. So it seems like it would be kind of a waste of time.

[00:03:10] Chris: Right.

[00:03:10] Ned: Call. Well, Mr. Fancy pants this morning. Not me. Different fancy pants. I was there. Another woman was there. He pulls in, calls the pet resort, I guess, does whatever the thing is in there, and a woman comes out, and instead of saying, oh, these other two people were before me, just gives her the dog and drives away. How would you react to that situation, Chris?

[00:03:42] Chris: Incomprehensible rage.

[00:03:44] Ned: That’s what I felt. There were flames in my eyes. My sight got blurry, and I was the first.

[00:03:56] Chris: I cannot believe that you screwed up that quote. There are flames the side of my face.

[00:04:03] Ned: So sorry. That’s how enraged I have some respect for yourself, sir. I can’t just playing it back in my mind. I’m enraged again. And being the person that I am, I didn’t say anything. I just thought angry thoughts very loud, because I don’t get it. I don’t need that at 08:00 A.m..

[00:04:26] Chris: I don’t I have no notes. I just it seems completely rational.

[00:04:32] Ned: And then the handler came out. Then again, and I was there before the other woman, and the handler started walking over to the other woman, and I’m like, don’t you fucking dare. Don’t you do it. And the woman goes, oh, no, he was here first. And we just nodded each other, like, the knowing nod. Like, I’m not the asshole, and, like, neither am I. All right, we’re cool. That’s my story. That’s my morning. Are you cool, though, in that one very specific scenario with that set of circumstances? Maybe give it a solid maybe.

[00:05:09] Chris: You know what? I’m going to allow it. Okay.

[00:05:12] Ned: Thank goodness. All I wanted to be was cool, man. If you’re trying to be cool, you can’t be cool. Yeah. And Harry Styles’one best album at the Grammys last night. So everything is wrong with the world.

[00:05:31] Chris: Yeah, but I thought we agreed that the Grammys are garbage.

[00:05:35] Ned: They are, but they can be garbage and wrong. And last night they were garbage and wrong. Okay.

[00:05:48] Chris: Just broke my brain a little bit.

[00:05:53] Ned: Hello, alleged human, and welcome to the Chaos Lover podcast. My name is Ned, and I’m definitely not a robot. I am able to traverse well defined terrain with relative ease, and when I stumble gracefully over previously unscanned objects, I am able to write myself without assistance. I barely ever fall into easily avoidable bodies of water. Yes, I am an ambulatory organic miracle, just like you. With me is Chris, who’s also here. Hi, Chris.

[00:06:28] Chris: I am such a child.

[00:06:31] Ned: You couldn’t help yourself.

[00:06:34] Chris: I’m trying so hard not to laugh out loud at making you say that out loud.

[00:06:41] Ned: You know, it’s deserved I’m doing really good. You really are. One time, Chris. One time.

[00:06:51] Chris: Dude, stop it.

[00:06:52] Ned: Okay, three times.

[00:06:54] Chris: It was, like, ten times.

[00:06:57] Ned: What can I say to the creeks? They beckoned me. The lake has its eyes on me.

[00:07:05] Chris: That’s why I talk about some tech garbage.

[00:07:08] Ned: No, definitely not. That’s why I went to the desert. Guess what? There wasn’t bodies of water, so you.

[00:07:16] Chris: Were safe for a short period of time.

[00:07:19] Ned: There was one creek we had to cross, and I was like, It’s only 2ft wide, but crap, it’s got my number. I will fall off this two foot wide bridge. Not a problem. Oh, yeah. All right, we’ve prevaricated enough. Let’s get into the tech garbage.

[00:07:37] Chris: I’m not sure that is a word.

[00:07:40] Ned: Look it up. You look it up. Don’t look it up. Risk five is a safe bet. See what I did there?

[00:07:50] Chris: Finally we agree on something.

[00:07:54] Ned: I had three other titles for this and that was the one that I thought would make you grown the least. You’re welcome. So we’re going to talk about the architecture that is risk five. But before we do that, we need to go over a little bit of history on why risk is even a thing that we need to think about. And this is risk spelled RASC for reduced instruction set computation, which we’ll get to. But to understand why that exists, you have to understand a little bit about what came before it and the X 86 instruction set. So are you ready? Azure? You firmly strapped into your chair.

[00:08:36] Chris: Do you have I have never been more prepared.

[00:08:42] Ned: Do you have your cheese helmet firmly in placed?

[00:08:46] Chris: It’s actually in the other room, but I’ll get it if I need it.

[00:08:50] Ned: Okay. Because nothing goes with chips. AWS.

[00:08:54] Chris: Well, as KCEL I’m going to kill you.

[00:08:58] Ned: It’s going great. All right. So if you’ve installed software before, you may have seen this pesky folder or directory or something along those lines. That’s called X 86.

[00:09:11] Chris: Right?

[00:09:12] Ned: And the reason it’s called that is because it refers to the X 86 instruction set that was developed by intel through its series of chips in the late seventy s and early 80s that started with things like 80 88 and 80 86. I believe 80 88 came first for some bizarre reason.

[00:09:32] Chris: You are correct.

[00:09:34] Ned: And it’s sort of the beginning of what we could consider the modern desktop world that along with the IBM PC, really sort of set the standard in terms of how desktops would be laid out for a long time. The instruction set itself was extended over time to support 32 bit computing and then 64 bit architectures, as well as things like coprocessors and floating point units. Yeah.

[00:10:06] Chris: And this is something that is not necessarily completely understood, but for example, in the early days of operating systems you could only ever have file names that were eight bits long.

[00:10:20] Ned: Yeah.

[00:10:20] Chris: Meaning eight characters. That was it. That was tied to the fact that it was an eight bit CPU. So as we went through time and went through 32 and then 64 bit processors, these things got what we could describe as more rational.

[00:10:45] Ned: Yes. But you would occasionally see, especially for older programs, that they could not deal with the longer file names and that they would truncate paths with weird Tildas and other markings because they couldn’t deal with that longer path. So that’s another a little bit of trivia if you’re ever plumbing the depths of your computer. It doesn’t come up much now that we’re all on 64 bit operating systems and even most of the essential command line utilities have all been updated to no longer be that problem. But yeah, every once in a while it’ll crop up. The other folder you’ll sometimes see is AMD 64, which that’s because, fun fact, AMD actually beat intel to the punch when it came to 64 bit instruction sets, which is why it’s called AMD 64 and not something else. Intel also introduced the Pentium Linode of processors because numbers like their previous processors, the 286, 386 and 46, couldn’t be trademarked, so other CPU manufacturers could name it the same thing. Isn’t that fun? Just like Sirix Hooray IP Law said. No one ever So the X 86 and AMD 64 instruction sets focus on CISC, or complex instruction set computers.

[00:12:15] Ned: And the goal behind a complex instruction set is to reduce the number of instructions that are required for you to accomplish a given task. Now, if you’ve ever dabbled in assembly language, and who hasn’t, honestly? Wild times really good. It’s one step above machine language. You will know that writing anything in assembly is extremely laborious.

[00:12:43] Chris: It’s one really mean.

[00:12:47] Ned: You really can’t assume that the processor is going to do anything for you. Nothing’s built in, nothing is assumed. Each step from loading the data out of memory to putting it into a register, to then processing that data with an operation and writing it back to the register and then back to memory requires an explicit instruction from the instruction set. And each of those instructions takes a clock cycle to complete. So the more instructions that are needed to complete an operation, the longer it’s going to take, because it’s going to take more clock cycles. So the goal behind CISC is to take an operation and streamline it by reducing the number of instructions that need to be performed. A good example that was provided by Stanford on their website is the process of, say, multiplying two members, two numbers that are currently held in memory. In the olden days, you would have to load each number into a register, perform the multiplication operation, store the result in another register, and then finally write the result back to memory. Since this is something that you’re probably going to do fairly often, there exists a CISC instruction that could be called Mult for multiply and it would accept two memory locations as arguments, and then it would just do the rest for you in a single clock cycle.

[00:14:19] Ned: And that’s pretty neat.

[00:14:22] Chris: And that’s the thing, you’re taking a complicated conversation, mathematically speaking of, you start with two separate numbers, you multiply them, you put them into a different category, and all that happens at one time, right?

[00:14:41] Ned: So it could be considered a little more efficient in that regard.

[00:14:46] Chris: Or as a great philosopher once said, pretty neat.

[00:14:50] Ned: Pretty neat. The alternative approach is Risk, aka reduced instruction set computers, and those came after CISC, which is kind of implied by the reduced portion of the name. The core idea behind Risk is to have fewer, less complicated instructions and put the onus back on the assembly language programmer to include all the necessary instructions for accomplishing a goal. So multiplying two numbers with Risk might take four instructions. Two to load, one to multiply, and one to store. Now, each approach has its trade off, so you don’t get something for nothing. CISC requires specialized transistor configurations into logic gates to handle the expanded instruction set. We are literally at the hardware level here, where we have to have dedicated transistors and logic gates to handle this particular instruction. In fact, there are 81 different integer operation instructions in the X 86 spec alone. Add on to that floating point smid, cryptographic instructions, virtualization instructions, and a bunch of fun undocumented instructions, because intel likes to put little Easter eggs in there, and each of those instructions can be executed in a single clock cycle. But the trade off is the hardware gets massively more complex to design, and it generally requires more power because you have more transistors.

[00:16:31] Ned: The risk approach involves fewer and simpler instructions, which require less sophisticated hardware and fewer transistors to implement the same functionality. Risk also makes less assumptions about how a series of operations should be completed, which to a certain degree, puts control back in the hands of the developer. That might be good if you can trust your developer, but not always. If so, CISC and X 86 work great for large systems that have room for the larger, more power hungry processors, while risk is generally a better approach for smaller systems and embedded devices. And that’s part of the reason Arm processors arm processors have done so well. In case you haven’t guessed, Arm uses a risk based architecture, which makes it way more power efficient. And for that reason, Arm processors did really well in the mobile space for a while. And now they have started to make their way into workstations and servers with the m one and m two chips for the Max and Armbased server class hardware like AWS’s Graviton chips.

[00:17:49] Chris: Right. And also things like raspberry pi.

[00:17:54] Ned: Yes, the raspberry pi. I have many of them that sit around and do mostly nothing, but it’s comforting just to have them around, Chris.

[00:18:06] Chris: I just want them to be in the room.

[00:18:10] Ned: I use one as a pie hole, and I use one as the controller for my 3D printer. And they’re awesome. And they do something that doesn’t need a ton of computation, and they do it well without sipping down a whole bunch of power. So I appreciate that. Now, due to the fundamental differences between the two instruction sets, anything that interacts directly with the hardware needs to be written to work on a specific architecture. And that’s why an X 86 operating system like Windows Eleven won’t run on an Arm processor. There is an Arm version of Windows. It’s terrible. That’s because porting a no, go ahead.

[00:18:57] Chris: Yeah, I was just going to say the operating system barely works, but none of the applications work.

[00:19:06] Ned: That’s right. That’s why porting a CISC friendly operating system and device drivers over to a risk platform is really no small feat. It is very difficult, and I think it’s one of the reasons that risk didn’t catch on sooner than it did. You need a walled garden, let’s say, like the Mac operating system, where they can just say, now we have this new hardware and it matches up with our operating system, which we rewrote to work on the risk architecture.

[00:19:40] Chris: Yeah. The thing is that Apple, when they moved in that direction how do I put this? That was brave. They knew that they had an opportunity to create a different world for their own applications and they were just like, well, either application people are going to adapt or they’re just not.

[00:20:11] Ned: Pretty much. Yeah. And they have the sort of ecosystem where they can make that sort of decision by fiat. We’re making this move, it’s happening, deal with it.

[00:20:21] Chris: Right. Which, spoiler alert, we will talk about that again in the lightning round.

[00:20:27] Ned: Excellent. Wow. It’s like we planned this sort of thing. We definitely don’t, no.

[00:20:33] Chris: I’ve never met you before in my life.

[00:20:35] Ned: Happy accidents. While Arm and Risk chips have become somewhat synonymous with each other, it isn’t the only game in town. And risk five is one potential competitor to Arm. And the main reason traces back to that old intellectual property thing I may have mentioned earlier. So the X 86 instruction set, or Isa, for instruction set architecture, is a proprietary instruction set and it is closely guarded by intel and AMD. You can’t just take a look at the instruction set, implement it on your own hardware and start selling it. You would get sued into oblivion or worse. And as far as they know, they don’t license it out either. So if you want to run X 86, you’re pretty much stuck with buying it from one of these two companies, forming a duopoly that’s a little lopsided at that. It’s been getting better.

[00:21:38] Chris: I mean, we’ll see.

[00:21:39] Ned: We’ll see indeed, now that Sapphire Rapids is out. I called it Sapphire Lake last week and then realized that I called it the wrong thing. But honestly, I don’t care.

[00:21:51] Chris: I think I’ve been calling it the same thing.

[00:21:54] Ned: I don’t want to get into all the other processors that exist out there prior to X 86, like Sun Spark processor or IBM’s Z series mainframe with their custom processor, or even Intel’s failed attempt to ditch the X 86 instruction set with the risk based Itanium. Basically, X 86 turned out to be good enough for 90% or more of all compute needs and operating systems like Windows Nt and Red Hat Enterprise Linux. And they were designed specifically for that Isa. It just has to be good enough. Chris?

[00:22:32] Chris: Well, I think that’s the gist of it is that was the end of Moore’s Law is CPUs got fast enough and efficient enough that they could do literally what we needed. And that was fine. It didn’t have to be efficient anymore because the CPUs were so goddamn good.

[00:22:55] Ned: That is true, and that has been the bane of programming in recent years as well. Is, it doesn’t matter how inefficient your code is, we can just throw more horsepower at it and it’s probably going to be fine, right? Until it’s not. Well, then came mobile devices and everything effectively changed overnight. So intel and AMD were not all that interested in developing chips for the nascent mobile market at the time. And the X 86 Isa was not very well suited for small devices that had limited battery life. Their disinterest created a market opportunity, and that was mostly capitalized on by Arm. So chips that use the Arm design implement a risk instruction set that meets the needs of mobile and embedded compute devices. And as cell phones, and then especially smart devices began selling like hotcakes, companies like Qualcomm started making boku bucks. Frankly, embarrassing amount of money.

[00:24:07] Chris: Like a ton. Yeah, like just hand over fist.

[00:24:12] Ned: Yeah. Now, it’s easy to conflate Arm chips with the company Arm itself, but it’s important to note that unlike intel and AMD who design and produce the chips they sell, granted, AMD leverages other facilities to actually create the chips, but they design and pay for the manufacturer and sell them directly, right? That is not how Arm does it. Arm creates reference designs for chipmakers and licenses the design to them. And then those chip makers can expand on the core instruction set and reference designs that Arm provides. But Arm gets paid a license fee. A sweet, sweet license fee. So Qualcomm, Apple, and AWS are free to implement their reference design however they want as long as they keep paying Arm that’s sweet money.

[00:25:16] Chris: How come we don’t license things?

[00:25:18] Ned: That’s a good .1 of the main benefits that Nvidia was going to see by trying to buy Arm was the cash savings from not paying their licensing fees anymore. Oh, and getting those licensing fees from everyone else also. It was a terrible idea and I’m kind of relieved and a little amazed that it got stomped on by legislators or regulators. Really. I kind of thought that was going to go through and create a new Hellscape.

[00:25:51] Chris: Sometimes the Hellscapes just can’t abide. Even themselves.

[00:26:00] Ned: The hellscape abides. Great, we have an episode title now. While I think the Arm approach is definitely an improvement over how intel and AMD do things, when it comes from a licensing and reference architecture perspective, there are still problems inherent in the proprietary scheme. The biggest concern is that of cost. Chipmakers have to keep paying Arm a license fee and Arm can kind of charge whatever it wants. It can strong Arm, if you will. I’m sorry, they aren’t. I’m not sorry. Arm isn’t the only game in town when it comes to these mobile chipsets and increasingly workstation chipsets, but they’re pretty close. The other thing is governments can prevent Arm from licensing their tech to countries, as is the case with China at the moment. You also lose out on the benefits of open source, where anyone can make suggestions, review code, and contribute to the reference instruction set. Now, 20 years ago, it was pretty infeasible that any small organization would be able to participate in chip design, test out that chip design and fabricate it. But things have changed tremendously since then. You can now, as a relatively small outfit, go through that whole process and send your reference design off to one of the chip fab companies like TSMC, and they will happily make those chips and send them back to you as long as in a decently sized batch.

[00:27:43] Ned: Plus you have I mean, you’re not.

[00:27:45] Chris: Going to be able to make like, ten no.

[00:27:48] Ned: But if you do want to go small scale to test things out, they have field programmable gate arrays or FPGAs that allow you to basically create the chip that you want using those programmable field arrays. Make sure it works as you expect it, and then you can send that updated and tested design off to the chip fab to create 1000 of them.

[00:28:12] Chris: Or like 10 million.

[00:28:15] Ned: Or like 10 million. So with that all in mind, the stage is now set for Risk Five. Risk Five is an open source instruction set architecture that can be implemented by anyone with no royalties or licensing fees or restrictions. You can view the entirety of their instruction set on GitHub right now, and if you are a computer engineer by training, you could even design a chip that implements the reference. And that’s fine. They will not charge you anything for it. Risk Five is overseen by the Risk Five international nonprofit. I believe that’s based out of Switzerland, which is actually a recent change that was spurred by the way governments are starting to use license embargoes as a cudgel against each other. The hope is by locating in Switzerland, they can stay neutral.

[00:29:14] Chris: So wait, are you trying to imply that governments are being mean to each other for no reason?

[00:29:21] Ned: Won’t say it’s for no reason, but they’re definitely being mean to each other. This is my ball and you can’t have it.

[00:29:27] Chris: It’s so weird. All right, go on.

[00:29:30] Ned: It’s frustrating. Risk Vibe was released to the community in 2015. Seven years ago. All right. Eight years ago. My God, what year is it?

[00:29:41] Chris: You are not good at math.

[00:29:42] Ned: No, definitely. And since then, hardware manufacturers have been hard at work creating chips that implement the standard. The history of Risk Five is pretty interesting, so I’ll include a link in the show notes if you want to read up on it. And you can also check out a debate around proprietary versus open standards when it comes to Isa I’ll include a link to that one as well. It’s pretty interesting. It has someone that worked on AMD chips and worked at intel for a while in the Pro proprietary category and then the person debating them worked on Risk Five and a bunch of IBM stuff back in the day. So both deeply steeped in chip design but coming from it with very different attitudes. Now, in order to adopt a Risk Five architecture, you’re going to need stuff like, I don’t know, an operating system, device drivers, applications that know how to use this instruction set and that is.

[00:30:40] Chris: A whole bunch of minor things.

[00:30:41] Ned: I know minor details. Sweep it under the rug. Unfortunately, that is no small task. Yeah, not for that reason. Stuff like embedded devices and smaller simpler computing devices have been the main target for Risk Five so far. These systems tend to have relatively simple operating systems, or in some case, no operating system at all, with the application written to talk directly to the hardware following the application execution environments to application binary interface model that is outlined by the Risk Five privileged software stack documentation. It’s a barn burner. It’s fascinating reading it’s 110 pages and you will not be able to put it down.

[00:31:31] Chris: It kept me up at night.

[00:31:36] Ned: Oh dear. If you need to fall asleep on a plane, I highly recommend it. Simple devices also tend to be very cost sensitive in nature because you’re making a lot of them. And if a manufacturer can save on licensing fees that they would pay to arm normally by rewriting the code to run on Risk Five, it might make a lot of financial sense to do so, especially once the devices are being manufactured at any significant scale. The next big hurdle is developing chips and developing software to support more complex devices like servers and workstations. And in a weird kind of way, this is actually easier to do for servers because it turns out serverless tend to do only one or two things, unlike your workstation, which has to run a bunch of off the shelf software from possibly hundreds or thousands of different vendors. We ask a lot of our desktops 1000. Well, not one individual workstation. Probably not. But if you think about a fleet of desktop, I need to hit the.

[00:32:46] Chris: Brakes on this right now because 1000.

[00:32:49] Ned: Yeah, I am not kidding.

[00:32:53] Chris: You probably run like twelve things. Seven of those.

[00:32:58] Ned: Twelve things, azure Pornhub eight, but who’s counting? If you look at a large Enterprise 10,000 seed Enterprise, that’s like the standard, right? They’ve done surveys of the application sprawl at these organizations and found that the average organization has over 1000 applications that are used. Not used well, not used efficiently, but they’re there.

[00:33:29] Chris: All right, so let’s just say that that’s an accurate number. Holy shit.

[00:33:41] Ned: I know. I’ll dig up the study at a later date. But my point is it would be very hard to get all these different vendors on board to rewrite their applications to run on risk five. But it’s not that challenging to just rewrite Apache or NGINX to run on risk five. That’s doable.

[00:34:03] Chris: They both already run on risk five.

[00:34:04] Ned: See, that’s what I’m saying. You don’t have to rewrite anything at all. So the point here is a lot of servers, yes, they run different applications on top, but they’re really all just running like a web server stack, basically. So that makes things a little bit more straightforward as opposed to doing it for a workstation. One such company was covered on the website, the Next platform, and it’s called Ventana Microsystems. Why is it called Microsystems? I think it’s a nod to the old to Sun Microsystems of your because they don’t make microsystems. Sure, why not? They emerged from stealth mode in 2021 and they’re founded by industry veterans who are no strangers to chip design and product development. Both of the co founders have spent the last 20 years doing some level of chip design and marketing. So they’re aware of not only what the chips need to be able to do, but how to market it effectively to sell those chips, because that is a big portion of it. So their chip is called the Veron. We’ll go with that. Veyro NV one. There’s a v one and a v two. The V Two hasn’t actually been produced yet, but the V One, they’ve actually got mocked up AWS using the PaaS.

[00:35:29] Ned: It is going to be fabbed with a five nanometer process by TSMC and it’s going to be clocked at 3.6. It should be available for customer testing mid this year. This is a server class processor that supports virtualization standard I O hooks and memory interfaces. In terms of performance. The ventana has run the spec int 2017 benchmark against their V one processor, and it surpassed the performance of Intel, Ice Lake and AMD epic Milan chips, as well as AWS’s Graviton three processor. The next platform did some estimates of what the Sapphire Rapids and Epic Genoa will do on that same benchmark and it was at parity with both of those. But that’s honestly just a guess from the Next platform’s perspective because they haven’t actually been able to run the benchmark tests. But it’s competitive, I guess is my point.

[00:36:32] Chris: Which then it becomes a question of what is the cost basis.

[00:36:40] Ned: Absolutely. And initially it’s going to be relatively expensive because they won’t be producing them at scale.

[00:36:47] Chris: And that that’s the hard part. Right. Because that’s one of the reasons that intel and AMD can do what they can do with their CISC models, is they make 100 million CPUs at a time.

[00:37:01] Ned: Yes.

[00:37:02] Chris: And that’s not an exaggeration.

[00:37:03] Ned: No, that’s true. And they charge a premium for those chips, obviously. Intel especially intel yeah, I was going to say especially intel arm is a little bit cheaper. Right. Because they’re just licensing it. Out to whatever chipmaker is going to produce it. And if you decided to bring that chip design in house, hey, that’s something that AWS has done, then you save on that fee that you would normally pay to the chipmaker. Now you’re just paying the licensing fee to Arm. But you know what would be even better? Not paying the licensing fee. You know what AWS likes to do? Make money.

[00:37:48] Chris: Pay license fees.

[00:37:49] Ned: Yeah, they like to make money and repurpose open source stuff.

[00:37:52] Chris: Right. And claim that it was their idea.

[00:37:56] Ned: Right. So this really seems like right up their alley now. In truth, it took a really long time for Arm to make any headway in the data center, and it really was not until the Cloud Megascalers got tired of Intel’s shenanigans and said, we’re just going to make our own chips and run our stuff on that. I think Risk Five will succeed in the long term for two important reasons. One is that the Cloud hyperscalers have been through this already with Arm, so they already have a process in place and it’s not totally foreign to them, the idea of moving to a different architecture. So I think they’re primed to do it again if it will save them enough money, and I think in the long term it will. And the other big reason I think it will be successful is China. China is being increasingly cut off from the rest of the world when it comes to technology through embargoes and trade sanctions and all that kind of jazz. Being able to adopt Risk Five without worrying about the west cutting them off would be a win.

[00:39:04] Chris: Yeah, I think you’re right on both points. I think, first of all, the success of Graviton in AWS proves your first point.

[00:39:17] Ned: Yeah.

[00:39:19] Chris: Those virtual machines are the fastest growing thing that AWS has going on right now for a reason.

[00:39:27] Ned: Yes. And they are significantly cheaper than the X 86 equivalent EC Two instance.

[00:39:33] Chris: And the fact that they are built by AWS probably drives everyone at intel completely insane, I would guess. Part of the thing that I really.

[00:39:47] Ned: Enjoy about them, and I will point out that although Azure and GCP are not making their own Arm chips, they are already running some portion of their paths on Arm today because it’s easier to do that with a stack that you completely manage and only expose a small portion to. And they’re both moving to offer an Arm based or maybe they already have. I haven’t checked the news recently, but they’re moving to offer a virtual machine that uses Arm as well on the cloud platform. So I don’t think it’s that much of a stretch to see in the next three to five years them starting to adopt Risk Five for platform services and then slowly roll that into offering virtual machines that also have Risk Five. And they will provide the operating system images to run on those virtual machines.

[00:40:48] Chris: Yeah. And that’s really the only thing that’s missing is the applications. And Microsoft has been resistant, but for reasons that I don’t necessarily 100% understand, because Microsoft could very easily put their thumb on the scale and say, we need to support risk five. I don’t know why they haven’t done that yet, because if they did that and they twisted, for example, I don’t know, Adobe’s Arm. You’re seriously going to sit here and look at me and say, Adobe would not agree?

[00:41:34] Ned: I don’t know. I think they would. And it could just be a matter of time until Microsoft gets behind it full force once it realizes the cost savings that it would see from no longer paying Arm.

[00:41:48] Chris: Right.

[00:41:49] Ned: So I think we’re looking at least three to five years before we see any hyperscaler true, hyperscaler adoption, and another five before it becomes a relevant thing for your average administrator. But I think it’s safe to say risk five is going to play a huge role in the future of Compute, both in the data center at the edge and also in our pockets. Honestly, I think it’s going to roll out to mobile devices in the not too distant future as well. Maybe Google has something in the skunkworks. It would not surprise me. So that’s my risk five.

[00:42:27] Chris: I’m not going to lie. I’m going to take the under.

[00:42:30] Ned: You think it’s going to be sooner?

[00:42:32] Chris: Yes.

[00:42:33] Ned: All right, we will put that on the prediction spreadsheet. I’m putting you down for two years. 2.52.5. You got it. Lightning round.

[00:42:46] Chris: Lightning round. Apple looks to release even more expensive phone options because thanks. Thank God.

[00:42:57] Ned: Yeah.

[00:42:59] Chris: Inflation, constant layoffs, demanding, damaging the customer’s will to spend what ever. Apple already a $2.36 trillion company, needs more gosh. Damn it. Reports are coming in that the big fruit will be releasing an ultra model of their already world dominating and super expensive iPhone. This is hot. On the heels of the staggeringly successful ultra model of their increasingly world dominating watch, which clocks in at literally twice the price of the standard model, apple’s current highest priced option of a phone, the iPhone 14 Pro Max, clocks in at a frankly embarrassing $1,600, which holy shit. What on earth are the ultra versions going to cost?

[00:44:11] Ned: I think you just said it’s going to be like $3200.

[00:44:15] Chris: Tim Cook seems to have no problems with any of this, saying that people are, quote, willing to really stretch to get the best phone that they can, which frankly seems insane to me. Now, maybe I should be speaking in eye statements here, but there is no chance that I will ever pay four figures for a fucking cell phone. Ever. According to the Blue Book, my car is not even worth $1,000 anymore. Now, maybe that just means that I’m cheap, which fair. Yeah, but this is preposterous. It’s a phone. You cannot tell me that a phone needs to cost north of $2,000.

[00:45:11] Ned: You just can’t they will stop making it when people stop buying it, Chris. And apparently people always have a little more plasma to donate.

[00:45:21] Chris: Whose side do you want?

[00:45:23] Ned: I’m on the side of capitalism. More cloud outages, you say? Microsoft proves me right. Remember when I said that there would be more cloud outages in 2023? And also that no one would really care? Boy, is my back sore from all that padding. During a planned maintenance update on January 25, microsoft made a whoopsie doozy with their routers, which caused a multi hour outage for teams Azure and Microsoft 365 Services. According to their post incident report, the command sent to the router caused it to tell all adjacent routers in the wan to recompute their adjacency and forwarding tables. While that computation was happening, the routers could not properly handle forwarding traffic. Amazingly, the root cause was not actually DNS for once, it was BGP, which messed up DNS. So it’s still kind to DNS. Microsoft has assured everyone that they’ve taken steps to prevent this issue from occurring in the future, which is what every company always says. And entropy simply laughs maniacally because there are always infinitely more ways for a thing to be broken than fixed. It’s why I never made my bed growing up and why the universe will eventually descend into a heat death from which no escape is possible.

[00:46:50] Ned: So when you put it in that context, I guess a four hour outage doesn’t seem that bad, now, does it?

[00:46:58] Chris: Elon Musk, freed from one lawsuit, still has the other 80 to deal with. So here’s one that legit shocked me. Y’all remember 2018, right? We were so young then. Elon, clearly high, went on Twitter and announced that he was going to take Tesla private with the help of anonymous Saudi Arabian backers. The price was a dead giveaway, that he was lying, and also high. He said, and I cannot believe that I have to remind people about this, that the buyout was going to happen at $420 per share. Get it?

[00:47:50] Ned: Yeah, you get it.

[00:47:55] Chris: Look, nobody has ever accused Elon of being clever.

[00:48:02] Ned: No.

[00:48:05] Chris: This week, that lie finally got its result in court, and somehow a San Francisco jury declared Elon Musk not liable for the damage he did to Tesla’s stock. This is frankly unbelievable to me, as the facts of the case seemed obvious. Musk’s lead lawyer said that Musk was doing, quote, considering taking Tesla private and that a fraud cannot be built on the back of a simple consideration. This sounds to me like the worst kind of bullshit, but I’m only one man. I can’t really imagine how Elon can live with himself regarding basically every part of the way he lives his life, but I just assume this is one more sad brick in his sad, sad wall.

[00:49:15] Ned: I wish. Advertising company Google now offers free malware quick woo. You need to retrieve and install a piece of software. How do you do it? Now, if you’re like most folks, you probably open up a browser. Who am I kidding? You open a new tab on a browser you already had Open, and you type in the name of the software, and you click on the first viable link. And this is exactly what the Malware advertisers are hoping for. Better known by their portmanteau malvertizing malware distributors have been finding new and interesting ways to infiltrate the Google Ads served up to the unsuspecting public. Spam House has discovered a surge of these ads based off a pseudorandom collection of searches. One example cited by AWS Technica was a search for a Visual Studio download, which could be considered by some to be malware in its own right 12GB for a standard install, but the top ad shown went to www dot downloadstudio. Net, which attempted to deliver a malware package. Google does try to stay ahead of the latest techniques that Malvertising uses to disguise the source of the download and the domains involved. As always, you should ignore all Google Ads and verify the link before downloading any software.

[00:50:44] Ned: Better yet, try using a package manager like Chocolatey or Homebrew.

[00:50:50] Chris: You think you know about Ice, but you have no idea. If you ever want to go down a crazy rabbit hole, I recommend looking into the history of Ice and how ice was basically responsible for literally all of life on Earth. I’m not overstating this the fact that frozen water floats is damn close to unique in nature. Interestingly. Another thing that floats when frozen? Alcohol.

[00:51:36] Ned: So we could have been alcohol based. Think about it.

[00:51:38] Chris: Think about it. So what’s new? Well, recent research was able to show a new kind of ice, which they have defined as VIIT. This research was done using a quote, diamond anvil to squish water into a tetragonal crystal structure. God, I can’t believe it took me that much work to get that out. This delirious technology that I barely understand is something I am still somehow fascinated by. If there are listeners who do work in this field who would like to tell me more, my DMs are open.

[00:52:34] Ned: Developer survey confirms Devs love, rust and hate going anywhere. JetBrains has released the results of their State of Developer Ecosystem Survey for 2022, and the results are interesting. JavaScript continues to be the most popular language, despite being terrible to work on and universally reviled. It was also the least favorite programming language of the bunch. Python is a close second in popularity, probably because it topped the favorite language chart even though it uses white space. The two most popular languages developers are planning to learn are Rust and Go, with Rust having a noticeable rise in usage over the scope of 2022. A rather depressing portion of the survey showed that 91% of respondents are male across all countries surveyed, with Argentina and South Korea having the highest showing of female identifying programmers at 11% and 13%, respectively. That’s just pathetic. Also, from a demographics perspective, 72% have no kids, and 54% listed programming as a hobby. I’m not sure what to glean from that, other than maybe these folks should take an occasional break from programming. It should also be noted that 44% of respondents are between the ages of 21 and 29, so that might result in a skew towards people that are still firmly entrenched in hustle culture.

[00:54:10] Ned: I azure you, all of you developers out there. It is safe to put down the bullet coffee, close the lid on your MacBook, and try interacting with organic beings every once in a while. You might even enjoy it and make children. Well, yeah, I mean, if you must. Hey, thanks for listening or something. I guess you found it worthwhile if you made it all the way to the end. So congratulations to you, friend. You accomplished something today. Now you can head outside on your hoverboard, don your UV reflected raybans, and blast Hughie Lewis till the lactose symbiotes trundle home. You’ve earned it. You can find Miracris on Twitter at ned 1313 and at heiner 80 respectively. Or follow the show at ksunderscore lever if that’s the kind of thing you’re into. Show notes and the sign up for our newsletter are available@kslever.com if you like reading things. But you shouldn’t podcasts continue to be better in every conceivable way. We’ll be back next week to see what fresh hell is upon us. Tata for now.

[00:55:15] Chris: All right.

[00:55:17] Ned: What shit z IIT sounds like.

[00:55:21] Chris: Hold on. I am angry with you, sir. Why do you have an unkind we’re her to say about Hughie Lewis?

[00:55:29] Ned: I was not implying any unkind because.

[00:55:32] Chris: These are fighting words. I will elbow drop you sports top.

[00:55:37] Ned: Ten albums of all time without question. That’s the power of love, Chris. You’re lucky.