Jon Masters Thanks to this video, I recently picked up your “Building Embedded Linux Systems”. Good read thus far. I’m looking forward to applying the concepts to PPC, ARM and X86-64 architectures.
@@masternobody1896 CoreTeks talks about what he wants to talk about, he covers interesting topics of the industry, if you want more Nvidia vs AMD talk, you should visit Moore's Law Is Dead or AdoredTV
1 hour video, now the delay makes sense. Tom on his Broken Silicon podcast recently mentioned some stuff you were working on which put my mind to ease knowing something big was coming.
@¡Carajo! What you've just said is so utterly false it's not even funny. This statement right here is enough for anyone to realize you know very little about Intel and AMD.
@¡Carajo! Yea, 2% of customers do more than 1 thing on their PC/laptop/mobile device at a time... 🤦♂️ Even browsers are multi-threaded with each tab being a separate process. What planet are you on?
The only thing which bothers me with all these new companies coming with "innovative ideas" is that most of their executives are actually coming from most of the companies they are actually trying to compete with, which does not translate at first sight with market disruption but filling and exploiting some few spaces/market-shares still vacant in the industry...
There are no vacancies in the compute space that this company is filling, what could you possibly be driveling about? They're directly competing with tech giants, if they don't make something disruptive then they'll be out of business in no time.
@hot100easypack the irony that you don't even realize my response was to the OP, and not yourself, is all too funny here. Dunning-Kruger effect in it's fullest, you're perhaps barely average based on your inability to discern who's comment I was replying to.
@hot100easypack I'm happy to reply to yourself after reading your comment now, let's unpack: 1. Your English sucks 2. "It makes as much or more sense that they can see some opportunity but by which the prior company will not take action on for the reason that it would be cannibalizing their existing market share. So in that sense it is not fill a vacancy" - This is complete gobbledygook, he said they're competing for massive single threaded gains, the stated goal is one that has massive competition and has no applicable "self-cannibalizing" that would prevent the major competitors from being in direct competition with Nuvia for market share. Your "it is not fill a vacancy" isn't even using a real example, just some vague nonsense hypothetical you pulled out your ass to play devil's advocate Mr. "I'm triple digit IQ" 3. So for economical reasons, the companies fighting for data center market-share are unwilling to increase their IPC/single threaded performance, despite that provably being untrue, because that would somehow hinder them in gaining more market-share? Yeah, you're a genius all right.
@@aiGeis What surprises me the most is the entire business models of all these innovative companies is held up competence but is built on business models and product ideas which are very fluid like and can be disagreed and agreed upon easily just like opinions and still they thrive relatively easily. I think everyone can come up with same class of ideas and philosophies and even better to run the world. It's like you can see maze map beforehand but you can make it less hard only if you could navigate and comeback at same spot without your eyes, so you could finally see the image of the maze. LOL well that's my thought.
1. Super startup! - red flag. 2. - -"We will offer real step up on single thread." - "what about NVIDIA?" - - "Market is so huge there is place for everyone =3" I loled.
Worked in enough startups and yeah, this is a red flag. One thing is wishing for great improvement, another thing altogether is delivering a working product.
@@RoonyKingXL The typical vibe in any startup is over excitement about whatever product being developed. It's also known as believing your own hype fallacy. It often reaches cult like levels until few years later reality comes knocking when the funding dries up. So it's difficult yo get excited about promises when I know how these startup companies work internally. They require the hype for funding and building the team, but most startups end up failing.
@@RoonyKingXL Startup - is basically only starting firm, like investment project. Superstartap is like calling your gas station - SUPER gas station. You know, we a not just bank, we SUPER bank. It is sounds really arrogant in regards on startups nature(it is something that only started). Startup - is attempt to start something, calling it super is, eah...
What's never talked about is the dozens of startups in the silicon valley that for years have had millions dumped into them, and have yet to produce anything useful/profitable. Seems the tech industry is well on it's way towards yet another collapse.
As an academic-minded professional, it is great to see names like John Hennessy, David Patterson and Onur Mutlu being mentioned. As far as the SoC goes, while some of the aspirations sounded quite ambitious, it is great to someone openly addressing the real targets, benefits and challenges of architectural improvements rather than hiding behind buzzword-oriented marketing. Also, all of your questions were spot on. This is definitely my favourite video from this channel now. Thanks a lot for making this. Looking forward to more content like this!
You would have thought they would have avoided clashing names with the top of UK search results for Nuvia which is a 'International nuclear engineering, project management and service provider'.
@@LoveScreamTrue As far as my searches tell me there is only one brand or commercial entity simply called Puma. What you find in search results for a brand name is important. For Puma I expect to find results for the sportswear brand and the whole first page results on Google and Bing are for that brand, except for one on Bing for the animal.
What a fabulous interview. There is much to take away from it. I really appreciate the time you put into this one, Celso. I would love to see more interview videos in the future, for the plebs that are tech hobbyists, such as I, love the insights gained from talking to others within the silicon/coding space. Thank John for us as well. I'll be keeping my eye out for Nuvia news. I think this is easily one of my favorite videos that you have done. Your channel deserves high praises and you deserve all the success that comes your way (if not more). Interview or otherwise, I will always look forward to your videos.
This left me with a large question. If most performance gains are to be found in software optimization, what incentives will compel companies to optimize their software? I am thinking of companies like Adobe, which practically has a monopoly in the art world with Photoshop. There are only a few large software packages for a given field, and slapping on poorly optimized extras seem like the favored approach to keeping consumers happy. But, we need a lot of this software to practically be rewritten from scratch.
Linux servers are a special case there, in that 99% of the software is shared, you can make the other 1%, and the end user never sees any of it - just an API. All you're writing is there 1%, and you tie that into the other 99%.
RoonyKingXL GIMP is ages behind Photoshop, but that is irrelevant. He focus is on features over performance. Blender 3D is an amazing software package, but it still emphasizes features and lags behind commercial products in performance.
Laurel Sporter I am not sure what you mean. Just because something is open does not mean that it is optimized. On the topic of Linux, we could use a new OS built from scratch on modern technology.
@@Austin1990 Most optimization goes into a tiny fraction of a percent of any given program's code, and your own code is also pretty tiny. So, you have service X. You build it on top of FOSS software. You only need to identify and optimize what really needs it in X, and then find what's bottlenecking it from there outside - which is mostly code you can analyze, too, if need be. You get to produce a small code base that does complicated work, by using massive amounts of other code to build on top of. In many cases, a lot of that code, if it needs to be, such as for computation, will typically be well optimized from the start, and just needs config tweaking. For what isn't, or can't be, you can write a caching layer in top of it, swap to different software used for certain tasks (without customers ever even seeing anything change), and more. A client workstation app takes far more manhours to produce than a server-based service. A team of a handful of good devs can make the next big cloud app/service, or any niche business software, and have the available time to optimize it, whatever exactly that entails. There are a plethora of frameworks for all manner of things, different types of DBMSes, and so on, relieving a lot of the tedious work, while allowing a lot of flexibility, and the ability to just use them, so you can focus on your marketable solution to a problem some bureaucracy probably created. A decent GUI *alone* can take many people a long time to develop, and GUIs are tedious AF to code up. On top of that, MS has priced themselves out of small business and startup server/service development customers, over the past several years. Windows server software is increasingly becoming a nonstarter, no matter how compelling some features may be, and without running that software, the expensive VS licenses aren't worth it, either - it's not free to spin up hundreds of complete containers or VMs on demand, like with FOSS systems. Linux is the serious option, today, excepting a few FreeBSD weirdos. If another FOSS OS ecosystem took over, it would apply for that, too. It won't apply to Windows or OS X, though, nor any desktop program development. IoW, nothing these guys are doing, whether they succeed or not, really will matter at all for desktop apps. Totally different market. For their potential customers, it's about TCO in datacenters, for which more efficiency means more profit, and there's plenty of room for improvement. For client side commercial apps, you need competition, and there's a high barrier to entry. FI, Da Vinci has kicked Adobe into making huge performance improvements - good old competition.
Underestimating ARM chips is one of the biggest mistakes you can make right now, in the silicon industry. Some of the most powerful chips that will come out in the next 2 years will be Arm chips, and massive GPUs.
@@WinterCharmVT i knew that, ARM architecture is quite powerful and very power efficient Fujitsu made supercomputer with arm-based custom chip I just said, I'm skeptical about Nuvia
Probably first time I actually couldn't keep up with what they were talking about. I'd rather see a normal video than an interview but Coreteks is always amazing.
Your questions were so good that he really didn’t answer a couple of them, instead preferring to pontificate. Well done though and thanks for putting this company on my radar. I look forward to seeing what they produce.
I would really like to know how they will deliver FP16 and FP32 performance using ARM designs because they are only good at Int8 and Int16 performance, what interconnect might they be using?
@ 21:00 Biggest, loudest laugh i had in over a year! Mr. Corteks you Rock! So much class and valuable informations, this time you excelled in the entertainment excellence. Cheers...
That Royal Navy/ Pirates analogy rang home, 19:44. He didn't even get to sing out a tune to it with that guitar behind him. He left NUVIA and rejoined RedHat not long after this. It surely didn't work out well enough for Jon Masters.
He's one clever guy. I cut my programming teeth coding ARM. Some good memories & some very clever people back in the 1990s :) . Sophie Wilson said in interviews that the instruction set was designed to be compact and highly desirable.
As much as I wanted to enjoy this video, it's basically just a very long and drawn out interview. (you could have mention that at the start of the video or the description) After 40 minutes (with some naps during some part - I was already in bed ^^), I had to stop. I'm gonna wait for a super condensed version. But thanks for the video. :)
I didn't even know there was new content - UA-cam didn't show anything to me - but I was watching some tech videos and thought, it's been a while since Coreteks made a video, so I came here and surprise, a new video posted a few hours ago.
He is, by FAR, the nicest person i ever heard from the industry. Looking forward for their disruption. Thank you for that interview, very well done. Thumbs Up!
A few points: 1) The longest Intel X86 instruction appears to be 17 Bytes and not just 16. 2) Many of the 8086 instruction were quite worthless back then but are still there today 3) More complex instruction decodes carry a pipeline depth cost 4) RISC-V, I think made some unwise choices 5) Many folks with CS degrees came from glorified diploma mills 6) Many of today's trends are blind alleys and we don't know which
a.d. 5 - and in case they didn't, they don't know anything about building operating systems, compilers ... at best, they are algorithm savvy because they aim for Google interview.
Agent PorPor it’s not. This is a big boys game. There are already 64-128 core arm cpus for the server market. Look up Thunder. This company has zero chance of success.
Thanks for a really great in-depth conversation with an exciting company that we'll hopefully see much more of in the future. You have quite the nack for "predicting" the future.
We need a paradigm shift. The traditional CPU is what doesn’t scale well by just adding cores, it’s obviously approaching a limit. (In the calculus sense of the term.) The thing that Moore’s Law says is “we get 2x the number of transistors every X months.” Performance is supposed to come from using them well. Just adding more cores doesn’t do that. There are diminishing returns to just adding more cores or even more execution units to a core. GPUs were a paradigm shift, but we need something more general purpose that scales well. And that is not a Von Neumann architecture.
8:55 Having a nimble team, just the right size, no overhead...shows video of several very heavy people :) Impeccable timing....and no, I don't apologize...obvious is obvious!
Moore’s law is about transistor counts, not clock speeds. Clock speeds were a result of using more transistors to increase pipelining and it was a good use of transistors. But now we need to learn how to go beyond just increasing clock speeds and find more innovative ways to use transistors.
... or much better transistors. A CMOS NOR gate needs 4 transistors. If only 2 were needed then the chip gets smaller. The fastest silicon transistors work up to something like 300GHz but making a CPU out of those is hard. The thermal conductivity of the substrate matters.
Ken Smith The C in CMOS stands for complementary. Prior to CMOS they used 1/2 the transistors, complementary metal oxide use a transistor to switch in both directions rather than letting the voltage float when the transistors weren't active. This was a "Moore's law advancement", because when you don't enough transistors you can't do it. If you barely had enough transistors to make an 8 bit CPU, CMOS is not possible in the same applications.
@@pweddy1 I know what CMOS stands for. An NMOS chip would commonly have used more transistors than a CMOS part. In an NMOS, only N-channel MOSFETS were used. This means that the pull-up was an NMOS working a lot like a constant current source. In a TTL gate, if you wanted it to go fast, you needed to add the "Miller killer" circuit. This would generally mean that the two input TTL gate would have more than 3 active parts. TTL was too power hungry for making big chips. In IIL, logic like was used in wrist watches and other places where power needed to be low, the transistor count would be less than that of a CMOS gate but those chips are much too slow to make a processor. Practically everything is CMOS these days.
Coreteks forgot to put these links: Onur Mutlu Lectures - ua-cam.com/channels/IwQ8uOeRFgOEvBLYc3kc3g.htmlvideos Chris Mack Lectures- ua-cam.com/users/lithoguruvideos
I think they are building a barrel-type processor: a lot of cores, few threads. Hence why they claim better single core IPC than other. It's simple yet effective.
X86 is just the ISA, the vulnerabilities come from the microarchitecture. If Intel was using ARMv8 on their chips with the same branch predictor, Meltdown would still be a thing (as seen in some high performance ARM chips, which are vulnerable to Meltdown too). AMD was largely unaffected, because the vulnerability was already theorised (Intel knew this too) and chose to take it seriously. I'm looking forward to some RISC-V action if NUVIA is coming with that. Would be nice if they can help standardise a chipset for bootable RV computers across all chip makers (and avoid ARM's clusterfuck of a chipset arms race between Qualcomm, Apple, Samsung, Broadcomm, NXP etc. so you don't need to rebuild the OS just for a chipset change). RV has some serious advantages over normal RISC or CISC ISAs, which makes me a fan. Vector extensions, compressed instructions etc. (kinda weird that you can't have multf without divf though)
21:28 Well done. I wish this sort of research and questioning was applied in other fields.
😊
greeds law is rising and the future ....greed will increase 100x every year.
Jon Masters Thanks to this video, I recently picked up your “Building Embedded Linux Systems”. Good read thus far. I’m looking forward to applying the concepts to PPC, ARM and X86-64 architectures.
... tesla car fanboy
A car that makes higher/nastier pollution then a normal gas car
Great video man. I have new videos up!
Ohhhh, 1 hour of Coreteks, let me prepare my tissue boxes.
If you don't mind I'd like you to give me the name of the artist who drew the picture you have as your profile picture.
@@cristianr9168 Fns_kv
@@attq3980 thank you
And now that art talk is over let's be real
what are you doing with tissue box
@@abdAlmajedSaleh wiping his ass. Toilet Paper is still too precious.
There he is. I was wondering if you'd post anything soon.
finnally
now talk about nvidia and amd
@@masternobody1896 CoreTeks talks about what he wants to talk about, he covers interesting topics of the industry, if you want more Nvidia vs AMD talk, you should visit Moore's Law Is Dead or AdoredTV
intel still stuck in 14 nm....dang it raged...now amd is winning
1 hour video, now the delay makes sense. Tom on his Broken Silicon podcast recently mentioned some stuff you were working on which put my mind to ease knowing something big was coming.
The last time I was this early Intel was still competitive in the desktop market
AMD IS THE KING now
@@Speedjcp ok speed calm down
@@Speedjcp you should be excited about technology but you shouldn't be blindly loyal to any company or brand.
@¡Carajo! What you've just said is so utterly false it's not even funny. This statement right here is enough for anyone to realize you know very little about Intel and AMD.
@¡Carajo! Yea, 2% of customers do more than 1 thing on their PC/laptop/mobile device at a time... 🤦♂️ Even browsers are multi-threaded with each tab being a separate process. What planet are you on?
The only thing which bothers me with all these new companies coming with "innovative ideas" is that most of their executives are actually coming from most of the companies they are actually trying to compete with, which does not translate at first sight with market disruption but filling and exploiting some few spaces/market-shares still vacant in the industry...
There are no vacancies in the compute space that this company is filling, what could you possibly be driveling about? They're directly competing with tech giants, if they don't make something disruptive then they'll be out of business in no time.
@hot100easypack the irony that you don't even realize my response was to the OP, and not yourself, is all too funny here. Dunning-Kruger effect in it's fullest, you're perhaps barely average based on your inability to discern who's comment I was replying to.
@hot100easypack I'm happy to reply to yourself after reading your comment now, let's unpack:
1. Your English sucks
2. "It makes as much or more sense that they can see some opportunity but by which the prior company will not take action on for the reason that it would be cannibalizing their existing market share. So in that sense it is not fill a vacancy" - This is complete gobbledygook, he said they're competing for massive single threaded gains, the stated goal is one that has massive competition and has no applicable "self-cannibalizing" that would prevent the major competitors from being in direct competition with Nuvia for market share. Your "it is not fill a vacancy" isn't even using a real example, just some vague nonsense hypothetical you pulled out your ass to play devil's advocate Mr. "I'm triple digit IQ"
3. So for economical reasons, the companies fighting for data center market-share are unwilling to increase their IPC/single threaded performance, despite that provably being untrue, because that would somehow hinder them in gaining more market-share?
Yeah, you're a genius all right.
@@aiGeis What surprises me the most is the entire business models of all these innovative companies is held up competence but is built on business models and product ideas which are very fluid like and can be disagreed and agreed upon easily just like opinions and still they thrive relatively easily. I think everyone can come up with same class of ideas and philosophies and even better to run the world. It's like you can see maze map beforehand but you can make it less hard only if you could navigate and comeback at same spot without your eyes, so you could finally see the image of the maze. LOL well that's my thought.
Oh boo hoo...
Just want you to know that when i see the new coreteks video notification its a better day to me =)
i love the part when you quote himself it was fun
Haha. Yes, that was good. The ol’ right there, Fred.
I loved it! He got me good 😊
1. Super startup! - red flag.
2.
- -"We will offer real step up on single thread."
- "what about NVIDIA?"
- - "Market is so huge there is place for everyone =3"
I loled.
Worked in enough startups and yeah, this is a red flag. One thing is wishing for great improvement, another thing altogether is delivering a working product.
:3
@@RoonyKingXL The typical vibe in any startup is over excitement about whatever product being developed. It's also known as believing your own hype fallacy. It often reaches cult like levels until few years later reality comes knocking when the funding dries up.
So it's difficult yo get excited about promises when I know how these startup companies work internally. They require the hype for funding and building the team, but most startups end up failing.
@@RoonyKingXL
Startup - is basically only starting firm, like investment project.
Superstartap is like calling your gas station - SUPER gas station. You know, we a not just bank, we SUPER bank.
It is sounds really arrogant in regards on startups nature(it is something that only started). Startup - is attempt to start something, calling it super is, eah...
What's never talked about is the dozens of startups in the silicon valley that for years have had millions dumped into them, and have yet to produce anything useful/profitable. Seems the tech industry is well on it's way towards yet another collapse.
As an academic-minded professional, it is great to see names like John Hennessy, David Patterson and Onur Mutlu being mentioned. As far as the SoC goes, while some of the aspirations sounded quite ambitious, it is great to someone openly addressing the real targets, benefits and challenges of architectural improvements rather than hiding behind buzzword-oriented marketing. Also, all of your questions were spot on. This is definitely my favourite video from this channel now. Thanks a lot for making this. Looking forward to more content like this!
7 minutes in and i'm dying from the constant scrapping sound.
It's terrible. Think I'll have to turn it off once I reach my limit.
damn you! you couldn't keep it to yourself!? I was blissfully ignorant until you decided to ruin it for me
@Marc Jackson Instructions unclear, scrapping sound MUCH worse on my Vandersteens. Now my neighbors hate you, too. Give better advice you ponce
I put it through Nvidia Voice. It works :)
@@veryInteresting_ dude. thats a pretty smart idea
oho, a one hour upload from Coreteks!
Couldn't wish for anything else at 3am in the morning :)
You know, I just wanted to go to bed. Looks like sleep's gonna have to wait for one more hour :)
True!! Coreteks is the best
ah, more tech ASMR
This guy always delivers.
Thanks Coreteks
yes cute guy😂
FINALLY!!! Been waiting for so long for a vid from you man! I havn't even seen it yet, I just had to express my happiness wuuhuhuuu!!!! :D
You would have thought they would have avoided clashing names with the top of UK search results for Nuvia which is a 'International nuclear engineering, project management and service provider'.
Why would that matter? Are you saying the same about Puma? :)
@@LoveScreamTrue
As far as my searches tell me there is only one brand or commercial entity simply called Puma.
What you find in search results for a brand name is important. For Puma I expect to find results for the sportswear brand and the whole first page results on Google and Bing are for that brand, except for one on Bing for the animal.
What a fabulous interview. There is much to take away from it. I really appreciate the time you put into this one, Celso. I would love to see more interview videos in the future, for the plebs that are tech hobbyists, such as I, love the insights gained from talking to others within the silicon/coding space. Thank John for us as well. I'll be keeping my eye out for Nuvia news.
I think this is easily one of my favorite videos that you have done. Your channel deserves high praises and you deserve all the success that comes your way (if not more). Interview or otherwise, I will always look forward to your videos.
This left me with a large question. If most performance gains are to be found in software optimization, what incentives will compel companies to optimize their software? I am thinking of companies like Adobe, which practically has a monopoly in the art world with Photoshop. There are only a few large software packages for a given field, and slapping on poorly optimized extras seem like the favored approach to keeping consumers happy. But, we need a lot of this software to practically be rewritten from scratch.
Linux servers are a special case there, in that 99% of the software is shared, you can make the other 1%, and the end user never sees any of it - just an API. All you're writing is there 1%, and you tie that into the other 99%.
RoonyKingXL
GIMP is ages behind Photoshop, but that is irrelevant. He focus is on features over performance.
Blender 3D is an amazing software package, but it still emphasizes features and lags behind commercial products in performance.
Laurel Sporter
I am not sure what you mean. Just because something is open does not mean that it is optimized. On the topic of Linux, we could use a new OS built from scratch on modern technology.
@@Austin1990 Most optimization goes into a tiny fraction of a percent of any given program's code, and your own code is also pretty tiny.
So, you have service X. You build it on top of FOSS software. You only need to identify and optimize what really needs it in X, and then find what's bottlenecking it from there outside - which is mostly code you can analyze, too, if need be. You get to produce a small code base that does complicated work, by using massive amounts of other code to build on top of. In many cases, a lot of that code, if it needs to be, such as for computation, will typically be well optimized from the start, and just needs config tweaking. For what isn't, or can't be, you can write a caching layer in top of it, swap to different software used for certain tasks (without customers ever even seeing anything change), and more.
A client workstation app takes far more manhours to produce than a server-based service. A team of a handful of good devs can make the next big cloud app/service, or any niche business software, and have the available time to optimize it, whatever exactly that entails. There are a plethora of frameworks for all manner of things, different types of DBMSes, and so on, relieving a lot of the tedious work, while allowing a lot of flexibility, and the ability to just use them, so you can focus on your marketable solution to a problem some bureaucracy probably created. A decent GUI *alone* can take many people a long time to develop, and GUIs are tedious AF to code up.
On top of that, MS has priced themselves out of small business and startup server/service development customers, over the past several years. Windows server software is increasingly becoming a nonstarter, no matter how compelling some features may be, and without running that software, the expensive VS licenses aren't worth it, either - it's not free to spin up hundreds of complete containers or VMs on demand, like with FOSS systems.
Linux is the serious option, today, excepting a few FreeBSD weirdos. If another FOSS OS ecosystem took over, it would apply for that, too. It won't apply to Windows or OS X, though, nor any desktop program development.
IoW, nothing these guys are doing, whether they succeed or not, really will matter at all for desktop apps. Totally different market. For their potential customers, it's about TCO in datacenters, for which more efficiency means more profit, and there's plenty of room for improvement. For client side commercial apps, you need competition, and there's a high barrier to entry. FI, Da Vinci has kicked Adobe into making huge performance improvements - good old competition.
Austin P autodesk and google also good names to include
You can tell he was really impressed with the questions.
Apart from the static this is the best video on your channel. You've got bright days ahead of you if you keep this up
“Human malware” i love it
First coined by Gamers Nexus.
doesn't exist
What do you think he said?
ARMv8, SPARC and PowerPC could be brought together to one Compatible Plattform (Mottherboard aso) performing at an Advanced level...
Nuvia video
imo, Nuvia is a bit overhyped for silicon startup that gonna shake the market with arm-based chip
It remind me of WeWork
🙁
WeWork was just flawed
Yeah, still the hype is a little over the head for startup without actual product or schematic architecture yet,
Just promises
@@ZAR556 indeed...I'm slightly skeptic ...haha...
Underestimating ARM chips is one of the biggest mistakes you can make right now, in the silicon industry.
Some of the most powerful chips that will come out in the next 2 years will be Arm chips, and massive GPUs.
@@WinterCharmVT i knew that, ARM architecture is quite powerful and very power efficient
Fujitsu made supercomputer with arm-based custom chip
I just said, I'm skeptical about Nuvia
I was in need of a new video of yours. FeelsGoodMan
Everyone else: Regular upload Schedules. Coreteks: 1 hour documentary about a revolution in the silicon industry...
10/10 for this.
He kept everyone longing for the 'specifics'. I didn't watch it past 34:00, but too early to tell of they are gonna become the Tesla of chips?
Made it to 29:35, what i gathered was, we have something, cant say what, but maybe one day. Nebulus is the word.
Probably first time I actually couldn't keep up with what they were talking about. I'd rather see a normal video than an interview but Coreteks is always amazing.
8:55 The dude on the left has Win98 installation on his screen O.o
Desktop background pic I guess. Win98 didn't do 16:10
Coreteks another great skin care product video . Awesome was looking for a good tech video
Thanks great work
Your questions were so good that he really didn’t answer a couple of them, instead preferring to pontificate. Well done though and thanks for putting this company on my radar. I look forward to seeing what they produce.
I can't get over how well these videos are done
I would really like to know how they will deliver FP16 and FP32 performance using ARM designs because they are only good at Int8 and Int16 performance, what interconnect might they be using?
I like that you upload longer but better videos
damn waiting for you coreteks haha!
Perfect timing
Thank you for the interview! Cheers!
Great so when do they go public?
@ 21:00 Biggest, loudest laugh i had in over a year!
Mr. Corteks you Rock!
So much class and valuable informations, this time you excelled in the entertainment excellence.
Cheers...
That Royal Navy/ Pirates analogy rang home, 19:44. He didn't even get to sing out a tune to it with that guitar behind him. He left NUVIA and rejoined RedHat not long after this. It surely didn't work out well enough for Jon Masters.
What the heck happened to the Audio around 43:00 in? Suddenly there's a massive amount of white noise. It's extremely tiring to listen to.
He's one clever guy. I cut my programming teeth coding ARM. Some good memories & some very clever people back in the 1990s :) . Sophie Wilson said in interviews that the instruction set was designed to be compact and highly desirable.
hahahaha the last vid was 40 mins. now we have a breakthrough.. happy 1hour video coreteks.
love the informative videos man
So, not only are your speculations pretty much bang on, you are actually inspiring innovation throughout the industry. Awesome content.
Thumbs up first, then one hour of cool tech talk voice immersion!!!
Good luck for Nuvia and all those who seek knowledge on new challenges!
Grand! love the ambition! looking forward to see them progress! Excellent questions Coreteks! 🥰
Great last question and overall interview. Thanks for the info.
did you do a voiceover for the interview to get the voice right? 🤨
I came to the comments to ask the same thing... 🤔
Thanks for the new voice profile, much less fatiguing!
"Hehehe..."
-Coreteks, 2020
22:00
22:05 *breathing intensifies*
Aaaaand it has been sold to Qualcomm :P
When a corteks video show up and you are breathless....
As much as I wanted to enjoy this video, it's basically just a very long and drawn out interview. (you could have mention that at the start of the video or the description)
After 40 minutes (with some naps during some part - I was already in bed ^^), I had to stop.
I'm gonna wait for a super condensed version. But thanks for the video. :)
Duration is on the right bottom corner of every YT video ;-)
What are playback speed multipliers
Just one question: when is IPO time?
perhaps when they have an actual product? If not then surely it's a scam.
I didn't even know there was new content - UA-cam didn't show anything to me - but I was watching some tech videos and thought, it's been a while since Coreteks made a video, so I came here and surprise, a new video posted a few hours ago.
I’m nicely set and relaxed, getting my feel of Coreteks
Well, if Nuvia doesn't turn out to be "new VIA", I'd love to see what they can accomplish.
More like this please! Brilliant insights as always.
I loved the last question that really helped. I learned a lot from this. THX
Awesome as always. Really interesting interview, you've obviously done your homework.
Coreteks, you gotta talk about their Qualcomm acquisition now.
So... There was an exit strategy after all? Qualcomm just acquired them
What, Coreteks talks same voice tone everyday? (interview). Though he go into super-narrator mode for vids :D
What are you even saying?
@@nowonmetube what u even asking?
@@MrValgard your first sentence doesn't make any grammatical sense imo.
@@nowonmetube idk in my language it have sense, sorry for bad translation. 7ppl understood me tho
@@MrValgard well it didn't for me, so I was asking 😂
GOOD STUFF MAN!
I've been waiting and the wait has been so worth it >:D
An hour long coreteks hurray!
He is, by FAR, the nicest person i ever heard from the industry. Looking forward for their disruption. Thank you for that interview, very well done. Thumbs Up!
Love the subtle background music, not sure why though... Maybe bc it helps create continuity throughout the entire video...idk
Coreteks is the JerryRigEverything voice for tech interviews.
A few points:
1) The longest Intel X86 instruction appears to be 17 Bytes and not just 16.
2) Many of the 8086 instruction were quite worthless back then but are still there today
3) More complex instruction decodes carry a pipeline depth cost
4) RISC-V, I think made some unwise choices
5) Many folks with CS degrees came from glorified diploma mills
6) Many of today's trends are blind alleys and we don't know which
a.d. 5 - and in case they didn't, they don't know anything about building operating systems, compilers ... at best, they are algorithm savvy because they aim for Google interview.
Already missed you hun :)! super video again from you!
Qualcomm liked this video
Focus on both single core performance and scalability... what?
Lisa Su has entered the chat.
Great interview and I’m really looking forward to checking out the people that he mentioned towards the end of the video: lecturers etc
One hour of Coreteks, oh yeah baby!
Grande trabalho!
Continua assim, até um mês atrás nem sabia que eras português.
Os teus documentários são excelentes.
@53:45
Is that the Node-RED flow editor open on the notebook there?
I have more faith in Apple creating faster ARM cpus to be honest. They have the resources and need to build custom fast ARM cpus. 🐯🐯🐯🐯🐯🐯🐯🐯
Yes I’m waiting for apple to do fast ARM CPUs for their macs but now it’s just up to the opposing side (windows and servers) to be using ARM.
Agent PorPor it’s not. This is a big boys game. There are already 64-128 core arm cpus for the server market. Look up Thunder. This company has zero chance of success.
Agent PorPor emojis are the greatest Japanese invention ever you spastic! 🖕🖕🖕🖕🖕🖕😎😎😎😎😎😎😎
Agent PorPor pay attention you spastic ua-cam.com/video/FMviiJaPHIE/v-deo.html
Agent PorPor anime is the greatest form of entertainment on planet earth.
Thanks for a really great in-depth conversation with an exciting company that we'll hopefully see much more of in the future. You have quite the nack for "predicting" the future.
We need a paradigm shift. The traditional CPU is what doesn’t scale well by just adding cores, it’s obviously approaching a limit. (In the calculus sense of the term.)
The thing that Moore’s Law says is “we get 2x the number of transistors every X months.” Performance is supposed to come from using them well. Just adding more cores doesn’t do that. There are diminishing returns to just adding more cores or even more execution units to a core. GPUs were a paradigm shift, but we need something more general purpose that scales well. And that is not a Von Neumann architecture.
8:55 Having a nimble team, just the right size, no overhead...shows video of several very heavy people :)
Impeccable timing....and no, I don't apologize...obvious is obvious!
AMAZING video! I love your work. 👍
Dude! the content here is golden. Where's all the subscribers at?
Do a video on Chinese homegrown chips !
He get's street cred for having an SGI O2 sitting there in the background.
Is that a huge load of coffee in John Master's cup at 4:52 or is it just black tea with milk? I'm concerned about his heart rate.
Hi, I love your videos. From past, present and future. Keep recording.
Moore’s law is about transistor counts, not clock speeds. Clock speeds were a result of using more transistors to increase pipelining and it was a good use of transistors. But now we need to learn how to go beyond just increasing clock speeds and find more innovative ways to use transistors.
... or much better transistors.
A CMOS NOR gate needs 4 transistors. If only 2 were needed then the chip gets smaller.
The fastest silicon transistors work up to something like 300GHz but making a CPU out of those is hard.
The thermal conductivity of the substrate matters.
Ken Smith
The C in CMOS stands for complementary. Prior to CMOS they used 1/2 the transistors, complementary metal oxide use a transistor to switch in both directions rather than letting the voltage float when the transistors weren't active. This was a "Moore's law advancement", because when you don't enough transistors you can't do it. If you barely had enough transistors to make an 8 bit CPU, CMOS is not possible in the same applications.
@@pweddy1
I know what CMOS stands for. An NMOS chip would commonly have used more transistors than a CMOS part. In an NMOS, only N-channel MOSFETS were used. This means that the pull-up was an NMOS working a lot like a constant current source.
In a TTL gate, if you wanted it to go fast, you needed to add the "Miller killer" circuit. This would generally mean that the two input TTL gate would have more than 3 active parts. TTL was too power hungry for making big chips.
In IIL, logic like was used in wrist watches and other places where power needed to be low, the transistor count would be less than that of a CMOS gate but those chips are much too slow to make a processor.
Practically everything is CMOS these days.
Hey Coretex. Keep up the good work!!! Its been a while.
Qualcomm bought it ?
Coreteks is ARM ISA fanboy confirmed
At least ARM works, has standards now, has no hindrances to performance (as a standard CPU ISA), etc.. RISC-V fan boys I just don't get...
ARM is the future I think
he is just a fan of poweerrrrrrrrrrrr
@@hyakinthos_0902 trueeeee
@@WillFaustCuber RISC-V is the future!
3:06 its like the voice of god asking questions lol
RISC V FPGA dev board can now easily be had, what a world we live in.
Coreteks forgot to put these links:
Onur Mutlu Lectures - ua-cam.com/channels/IwQ8uOeRFgOEvBLYc3kc3g.htmlvideos
Chris Mack Lectures- ua-cam.com/users/lithoguruvideos
always a pleasure to watch
WHATS THE TL:DW
How many Ghz? number of cores? nm process?
3ghz??? 128core??
I think they are building a barrel-type processor: a lot of cores, few threads.
Hence why they claim better single core IPC than other.
It's simple yet effective.
You are back!!!!!!!!!!!!!!!!! Finally!
X86 is just the ISA, the vulnerabilities come from the microarchitecture. If Intel was using ARMv8 on their chips with the same branch predictor, Meltdown would still be a thing (as seen in some high performance ARM chips, which are vulnerable to Meltdown too). AMD was largely unaffected, because the vulnerability was already theorised (Intel knew this too) and chose to take it seriously.
I'm looking forward to some RISC-V action if NUVIA is coming with that. Would be nice if they can help standardise a chipset for bootable RV computers across all chip makers (and avoid ARM's clusterfuck of a chipset arms race between Qualcomm, Apple, Samsung, Broadcomm, NXP etc. so you don't need to rebuild the OS just for a chipset change).
RV has some serious advantages over normal RISC or CISC ISAs, which makes me a fan. Vector extensions, compressed instructions etc. (kinda weird that you can't have multf without divf though)
When will Nuvia go public?
Why do I bother watching these BEFORE I like them? I know they'll be informative and insightful already...
It is being acquired by Qualcomm apparently