AMD - The (Evolving) Master Plan 

That one actually made me laugh loudly. I love this guy's humour.

The humour comes how casually he says it lol

My god it's the legendary triple entendre!

orbit doesn't exist, the NASA Universe is a Lie

@enemy boat and nvidia is still using maxwell

Time well spent!!

I'm gonna have to watch this in a few sittings

Too true! Not complaining tho xD

I need to pee

A nerds dream become true! ;)

I am okay with this. :)

+1

Agreed!!

yup

You haven't even finished watching it, don't you?

ye, some even fap on dat material. No, not me.

consider it as an early xmas present ;)

He had an epiphany; that's what happened, and he is rewarded in watchtime by me for it.

Saaaammmeeeee

You made me laugh with that image of MJ excitedly munching his popcorn! Cheers for that! 🤣

I just went and made a cup of coffee and peeled some oranges for this video.

What did your girlfriend say?

She came.

Home alone with your girlfriend?

@@arlobubble3748 his right hand

You forgot bicycle pump & puncture repair kit...ref number 3

If I wasn't broke AF, I would. I would support quite a few RU-vidrs.

I do! Best investment on RU-vid if you ask me. This was well worth the wait! Like Thanksgiving diner completely full! Now I'm going to get something to eat and watch it again!

I do! It was investment well spent 👍

I understood everything and I am the wife. Lol

@@tenshi7angel , if it's not the scottish accent which would knock my wife out of the room (no, it's not you), it would be the tech savy things discussed here :)

@@eEnzo0 Yeah sadly his accent is either something to love or hate. :|

@@tenshi7angel Even though english isn't my native language I can understand Jim without any problems. I actually like his accent, kind of reminds me of Braveheart or How to train your Dragon :)

Bro that was funny

What did he figure out?

@@Disco_Potato watch the video again!!

When you pretend to know what's going but you really don't so instead of explaining what you understood, you just tell the person to watch the hour long video again. lol

@@snark567 Have a little faith why dont u ? Basically the scalability of the io controller which would give amd to go hbm2 for double precision while maintaining gddr6 for gaming cards is a genius move if u ask me and the reason i didnt spell it out in the comments as i want my fav youtuber to get watch time .

Intel is the slowest to learn its lesson here. You can be sure nVidia saw what Intel did wrong when Zen came out, and made note of it. 2019 and 2020 looks to be a heated war involving Team Red, Team Green and Team Blue.

Flew by! Really easy to watch and very well done!

Indeed! ;)

I got to listen to this at work. Was great. :)

He pretty much hit the nail on the head. Though he missed some context for "HBM 2.5" and future on die memory configurations.

it's still all speculations.

And most certainly that it will all come true.

I came for the future hehehehehe Surprise boys :3

I will give you a very much Reformed / Calvinist answer: absolutely nothing, it is all free grace ;-)

LOL

It's like dragon balls cell fight. Where Goku dies and Gohan defeats cell.

@@falconeagle3655 great analogy.

Not sure why but your comment makes me really emotional... ... But at the same time, greatly optimistic of AMD post-Navi.

Who needs silicon wafers when you can just use diapers? ¯\_(ツ)_/¯

Lol

Haha i laughed out loud

Stop stealing Reggie from Nintendo quote when the switch is full Nvidia ! xD

​@@ethan-fel Yeah, I wonder how and why that happened considering when you trace the history of Nintendo's graphics hardware, they've been With Ati longer than literally anyone else in the console space. Granted a lot of that is down to a technicality and corporate failures/buyouts. Silicon Graphics which collapsed but formed ArtX, which was bought out by Ati, which in turn was bought out by AMD... My guess is given the state of AMD GPU's when the switch launched, there simply wasn't anything in AMD's range that met the requirements. - while Ati led the mobile market between about 2006-2011, it lost almost all it's market share in the mobile space to Nvidia. And, well, the switch isn't even remotely close to laptop form factor and power envelope, it's in line with tablets and mobile phones... What are your options there if not Nvidia? Tegra is a mobile phone chipset but with familiar PC-like tools available. Your other options really aren't very good (Mali graphics? Eh.), and AMD literally doesn't provide any such option whatsoever. They have nothing in that market segment, good OR bad... Just non-existent. Meanwhile when you look at the only Nintendo systems so far that have had a portable form factor comparable in size to the Switch, they only gained any 3d hardware at all around 2004, and the entire family consists of PowerVR tech. PowerVR. A company that had SOME desktop cards in the late 90's but quickly gave up on that entirely, and is now almost a forgotten name in most people's minds... Their chips powered the Sega Dreamcast... But the name all but disappeared from mainstream awareness, because they bowed out of the PC market very early. And yet, they persisted all this time specialising in low power, mobile GPU's... So given what the switch is, and what Nintendo designed, they needed a chip powerful enough to give acceptable results in the home console space, but sufficiently low power to fit into a form factor that, realistically is that of a 7 inch tablet... - AMD? Not an option - they haven't got a single GPU that's even in the ballpark of the power demands. Even laptop grade low end AMD chips are 30+ watts (which is about 3 times total power draw of the system even in docked mode) - PowerVR? - well, they've got the low power stuff down, hence why they were used in the DS range. And a pre-existing and fairly long-lasting working relationship with Nintendo already. But... Well, I don't know if you've ever seen the best PowerVR chip currently around, but... Let's just say the results are not that impressive. - The other main contender is the various Mali graphics cores that ARM provides; They vary enormously in performance, and since ARM doesn't manufacture anything you'd have to ask a company to build you a chip containing such cores. Performance of even the best of these isn't anything to write home about though, and similar to the issues with a PowerVR chip. What does that leave? - Nvidia. In principle they're no better than AMD for this purpose given the vast majority of their product range, yet for some odd reason, even though on the whole the chip is not that popular in the mobile space, Nvidia chose to create a derivative of their desktop part that is suitable for use in mobile phones and similar small form factor devices. While not nessesarily powerful by mainstream standards, it's pretty beefy by mobile phone chip standards, and equally as importantly, fully supports the feature set that game developers are used to from PC hardware and other console graphics chips... Not the most powerful thing ever, and it almost seems like a random whim that Nvidia even designed such a chip in the first place... But under the circumstances... It's the best option they had given their goals. And one that AMD was outright incapable of providing. They had some alternatives to Nvidia, but AMD wasn't one of them, not to build what they had in mind... Given Nvidia's track record in the console space I'd wonder how long this deal will last before Nintendo (like Microsoft and Sony before them) gets pissed off with it. But given the lack of a viable alternative, that may not matter. I wouldn't be at all surprised though, that if AMD ever decides to make a small GPU in a similar category to the Tegra, that Nintendo would go right back to using AMD just like every other console manufacturer (and nintendo themselves prior to the launch of the switch). But, you can't ask a company to provide you with something they don't have...

MY WALLET IS NOT

KuraIthys quite frankly I doubt it since they signed a partnership for several years with Nvidia. Nvidia helped a lot in developing the switch Nintendo also moved to ARM, not x86, something that AMD doesn't do at all.

@@ethan-fel AMD DOES do ARM - currently only in select server space parts, but it's ARM all the same: www.amd.com/Documents/A-Hierofalcon-Product-Brief.pdf AMD has had their ARM license since an acquisition made even before their ATI acquisition. I'm sure that if someone approached them asking about an ARM plus a current, upcoming, or even made-in-partnership GPU (Like Sony's help on Navi) they'd be willing to talk. I'm sure that if Nintendo had actually tried talking to AMD's semi-custom division they might have been interested in building something. The set of factors that might have forced Nintendo to talk to Nvidia instead, if they had even tried AMD first, was AMD's already existing projects with Sony and Apple and Koduri's mutiny attempt. Dealing with all that, the talks may have had a timetable caveat that Nintendo couldn't accept.

I was thinking about exactly this too. Arcturus is like a fancier version of a bear, could imply that Arcturus has a bit more oomph to it than simply "bear".

It seems to me that AMD should stop and contact Jim to plan for the next 5 years.

Imma still hop in the HYPEtrain

You must have travelled nearly 4KM.

Haha, same here! Prefect video length for a train ride

Grats!

Save some of that for AMD's CES keynote!

Good ole priapism.

They make a blue pill to fix that...

@@ammarossdanan Blue pill? ayy lmao that's a funny way to spell shintel

You have 3 more hours before you need to seek medical attention.

GPUs are iterative thingies, "Zen" just doesn't happen there.

@@whoruslupercal1891 CPUs are also iterative, except if you drop out of the competition for five years and need to come back with a jump. Sounds familiar?

@@DeeSnow97 AMD never dropped out of GPU competition. Also Zen just does things Intel did.

Come on, A-MD Team!

AMD needs to Hire this guy. Jim truly had a great vision and knowledge to do iy

The last GPU I have absolutely no complaints about is the 7970. I loved that card its the only one I've bought two of. I still have them both and they still work perfectly, running OC'd as far as I could get them without flashing the bios. It was such a great card that should've been sold WAY more than it did. I'm hoping they'll compete in the high end again soon.

halsoy Just saying, all three of my 7970’s failed.

@@6Twisted I have 1 7970 that i have bought at the release , it's stil fonctionning today after all these year of 24/7

Yes I think you're right and that it's all about the I/O die. AMD already has the core chips, all that's needed is different I/O dies and that's what the companies can pay them for. I'll maybe updated this video in a few months. ;)

Once AMD moves to 7nm with volume, they will have a shit-ton of 14nm capacity at GF so they can afford to waste a little mm2 to have a single small I/O die that's flexible. Multiple different chips adds to cost for the masks and QA. It's not a small amount either.

AMD will achieve 10-15% sever market share with Zen 2 Rome,that means GF needs to produce 3.2-4.8 million 420mm2 14nm I/O DIEs per year for Epyc 2 alone.

@@henryfan48 Well within their capacity because their foundries are optimizing for 14nm rather than dedicating machines to 7nm.

This is all very exciting regardless of who is closest to the truth

People like first and then watch. Just like I did

Best year for the PC in a long, long time. ;)

Same - it would be great but I don't see it happening. ;)

@@adoredtv But what I do see happening is the new Uarch might just catch Nvidia off guard after Navi. With Nvidia putting all bets on RTX. I think AMD will have a more well redefined arch that will be highly more optimized for games now. The more I test out games on my RX 590 the more I see it coming close to performance of my friends 1070. Which is pretty funny.

​@@adoredtv ​ I would not think of multi GPU setup like Zen1 multi-cpu setup but splitting the gpu in smaller parts like zen2 setup. Like your speculation with the memory controller. But with more parts like fabbing each of the 4 compute engine separately. The bottleneck should the global datashare cache latency. With further improvement in infinity fabric it may be feasible. It's strange that gpu could not be split like cpu a la Zen2 with gpu having less interprocess synchronisation being highly data parallel. One would think separating each part and connecting them with a fabric or an interposer should be easier that doing the same for cpus.

Maybe not another week...I'm gonna be aiming for a change of pace in December and a change of focus too, so some "lighter" stuff hopefully. ;)

Yeah it could be just a little too early but I'm getting a lot of good vibes over Navi. ;)

Raja fits right in at Intel. Hopefully, it won't come to this: marketrealist.com/2017/11/intel-poach-amds-raja-koduri/ "In response to Raja Koduri’s departure, AMD stated that it has a strong graphics team and advanced graphics IP (intellectual property). The company warned that it would defend its IP if the need arises, hinting that it might resort to lawsuits to protect its IP."

@@adoredtv Same here, I think Navi will be the beginning of what people were waiting for. I think we will find in 2019 that there were several good reasons why Lisa spurned Raja, by favoring development of Navi over Vega.

nrencoret - I have been saying for over one year that the slimy serpent called raja koduri intentionally fucked over RTG. I also referenced the first article you have listed. I also mentioned a year ago that the rift between raja and Lisa really deepened after his hyped up claims with Polaris. He made Lisa look like a fool at the AMD capsaicin and spice event. I saw the look on her face and she was not happy. I believe she made a very calculated move by castrating raja and taking away 66% of his R&D team and placing them on the Navi project. I bet there was a lot of drama going on at AMD (between Lisa and raja) behind closed doors. She reallocated all the loyal AMD gpu engineers to the Sony/Navi project and left the slimy snake raja with his band of incompetent 'engineers' and mutineers to work on Vega. She knew Vega would fail and would cripple raja's reputation. It serves him bloody well right. She also knew Vega could be salvaged with the right team of trustworthy and competent engineers. Enter David Wang. His very first meeting was in February 2018 to discuss the future of Vega, shortly after Lisa released an official statement saying 'we believe Vega has legs'. I highly suspect that raja invested most of his time and effort into sabotaging Polaris and Vega. He designed two gpu architectures full of subtle flaws. He then very conveniently defected to intel, where another slimy Indian guy (Dr. Murthy whats his name) embraced raja with open arms. It sounds like pre-meditated industrial espionage to me. I honestly believe Lisa Su will never hire another Indian engineer again thanks to the snake raja. He has dishonoured all of us Indians, but what else can you expect from his kind?. They are very good at back stabbing. IMHO, raja is a grossly over-rated gpu engineer. He was hoping to steal and plagiarise AMD gpu architectural designs but Lisa slapped him with a bullet proof NDA. I think he is struggling to design an original gpu arch, which is why he subsequently hired the intel guy who designed the Larrabee gpu to be his lieutenant. When I joined the dots over a year ago, I received nothing but ridicule. You are fortunate as you were not subject to that.

@@justinlegrande1630 I don't think it will that badly to be honest. Intels iGPU hardware isn't actually that bad for its size, the main problem it has is that Intel never has really put that many resources into driver development. Heck, the entire reason their Linux drivers were open source before AMDs is because that meant they could have good Linux drivers with only a few devs working on it. Their dGPU line shouldn't actually need insider knowledge from nVidia or AMD so much as more money pumped into specific areas of driver development and working on making sure their GPU architectures scale up without hitting diminishing returns too early. Not that it's really that bad, they know that their iGPUs mainly need to be good at rendering Windows/Linux/OS X and maybe video playback which is what they excel at. Given the amount of time until they actually launch dGPUs, launching something based on their Gen11 IGPs but scaled up on the 10nm node with their main effort being good drivers and having a featureset that works well with what people likely already own around 2020-2021 would at least give them a solid product to build on for a more competitive second generation of GPUs. (eg. Something along the lines of Relive/Shadowplay, proper fixes/optimisations for specific games, a decent amount of driver settings, nicer UI, maybe even Freesync support.)

Cheers bud!

Same, looks great! Was a patron who created it and it just looked too good not to use.

@Shazbot81 Ah but 8.4 Tflops is 2x the ps4 pro and can be sold cheaper than a 10Tflop machine. Last thing PlayStation needs is another PS3 that nobody buys f9r years because it's too expensive.

@Shazbot81 At this point PlayStation are so far ahead they can almost ignore Xbox. As long as their hardware doesn't suck or c9st way too much the games libary will carry them through. PC may well be heading to 15 Tflops but console gamers as a majority don't care about or even notice 60fps or "true 4K".