[SPON: Use "brokensilicon“ at CDKeyOffer to get Win 11 Pro for $23: www.cdkeyoffer.com/cko/Moore11 Win 11 Pro ($23): www.cdkeyoffer.com/cko/Moore11 Win 10 Pro ($17): www.cdkeyoffer.com/cko/Moore10 Win10 Home ($15): www.cdkeyoffer.com/cko/Moore10h Office 2016 ($29): www.cdkeyoffer.com/cko/Moore16 Office 2019 ($48): www.cdkeyoffer.com/cko/Moore19 25% software discount code: brokensilicon 3% OFF EVERYTHING discount code: dieshrink ]
Hi Tom, if China were to have an unexpected breakthrough in photonic chips could they potentially leapfrog 🐸 past Intel, AMD & NVIDIA¿?¡! ... ...... ......... 369 ruclips.net/video/fk38BdJlujk/видео.htmlfeature=shared
Medusa turns people to stone. Stone like a _granite ridge_ or the sculptures of _Raphael._ It's actually a really cool way to combine the mythological greek avians, artists, and rock formations, of previous codenames. If only the same amount of effort to maintain consistency was applied to the name of the final products.
Honestly I'd much rather tell people I'd just bought an AMD Strix Halo or Medusa than an AMD AI 300. Why are the engineers so much better at making cool names than the marketing teams?
@@Gen0cidePTB Engineers want you to remember their babies. Marketing teams just want to rip you off as much as they can, with and by any mean they can think off. like the AI trend thing.
It really is a shame that investors gutted Intel to chase short-term profits in the 2010s. A smaller company is running circles around them in terms of the pace of improving their technology, developing new products, and delivering them to customers. AMD is competing so hard that I’m worried about pricing over the next couple of years as they don’t seem to have much incentive to keep prices low.
They didn't just gut it, they took all prospects of getting the new node done on time for ANY THING out the windows as they pre padded their bonuses in lieu of big fat federal monies. The way Intel share holders see it, the feds won't allow the company to fail because they want fabs more than they want Intel as a engineering firm.
Intel does still exist And AMD has the very real possibility of competing with itself No one will buy a R9 11950x if it costs 1000$ if the R9 9950x is 600$
@@bornonthebattlefront4883it starts with 9950X at $599. Then you get the 10950X at $649. Then the 11950X at $749 and so on. Just look at GPU pricing. The 980 was $549 10 years ago which is $724 today. The 4080 Super is selling well at $999 today. Without proper competition performance stagnates or prices rise which is not good for the consumer.
Yeah, AMD's increasing of prices is more aggressive than Intel's. I think Arrow Lake will make AMD's sales department cry again like during Bulldozer times 🙂
Zen 5c is on 3nm with 16 core chiplets with unified L3. AMD wants to deliver an "Epyc Dense" that absolutely demolishes Intel in performance per watt, performance per socket and performance per dollar. They already have two of the three, which gives them the lowest TCO, but they want the lowest up-front costs as well as the lowest TCO. Whether we'll see Zen 5c on Desktop is anyone's guess, but I doubt we will see Zen 5 "standard" ported to 3nm at this point. I think it will make more sense for them to shoot straight for Zen 6 on 3nm on desktop for a more impressive uplift.
AMD and Intel have been doing it for years, different codename for architecture, platforms, CPU cores, and chipsets. There's enough commonality where if you uncover a codename, you know the implications.
The worst energy drink. 🤢 And annoying marketing. They pay decent looking girls to drive around and hand out free cans. I short circuited their brains when I told them I didn't want that trash and ignored them.
11:43 Should have been colored red, but wow was that hot. Its been years and years since I've seen designs like that. Hardware companies used to use designs like that to advertise in PC hardware magazines. Oh and good leak. I'm really glad when I upgraded from LGA 1150 in 2022, I went to AM5. Zen6 looks like its going to be killer.
Hey, Tom. I've been a huge fan for a while and my request for you is this: Please make 5-10 tutorial videos over the next year, and make a "if you're new to this" playlist, or similar where people your viewers share with can understand your content better. I ask this because I have a few really technical friends, that are just lacking the right amount of technicality to care about what you're saying.
I know it's the flavour of the year, but I still haven't found any compelling AI applications. When they work, it's cool, but the hallucination factor makes it untrustworthy to me. Until that changes I'd love to see a product without the NPU and using that die space for something I'd actualy use.
In science AI based tools are a really big thing, not hyped, but actually already used. There AI is just integrated where it makes sense, combined with traditional approaches etc. That said, those applications are highly specialized, not of that universal generative AI that people are talking about non-stop. Funnily enough, those people using those tools, have little use for those weak NPUs on thoe APUs. They need workstation graphic cards with lots of RAM and also lots of SSD capacity. It will be interesting to see if actual use cases for those NPUs within off the mill APUs will materialize and solidify.
I have a Laptop with an AMD TPU, but i can not use it because AMD did not release any driver, framework or tool support to use the TPU on linux - at the moment AMD has to UP there software, if there don't it do not matter how many TOPS the TPU has if i can't use it.
@@Martin1991zab Just curious but what is the actual use case for a TPU? I get the point of having such a limited low power device for Tensor and Matrix maths, but what can actually be done with that application wise, and when is it better than just spinning up the GPU to 200-400mhz and doing it there? Even then the GPU has multiples more performance.
As another person with a 5800X3D, I won't need to upgrade my CPU until we get to _at least_ Zen 6, probably Zen 7 or 8. This will be dependent on GPUs getting powerful enough that they would be significantly bottlenecked by the 5800X3D, and software becoming CPU-intensive enough and/or local AI-acceleration becoming pervasive enough that the 5800X3D is no longer sufficient.
If you plan to ride your current system into the sunset, probably a good 4-8 more years, but the real answer is the day you buy a new GPU. 5800x3d is already a limiting factor with a xtx 4080 or 4090 even at 1440p, any gpu coming out in the next gen or two would be a killshot for it and would most certainly need a faster cpu.
Another thoughtful video from you, Tom. I don't know what people from tech sites are posting or if it is already just AI summaries, but thank you for bringing some human thought into it.
A couple of weeks ago I asked about the STRIX HALO release date after you updated the STRIX Point release date. Now you released this video spot on after and I appreciate the effort! Exciting news as I plan to upgrade to a laptop to replace the curent desktop but need one that is just a strong. Medusa will be that Laptop I believe.
This answer was generated by ChatGPT: _Using AI on your laptop can enhance productivity, creativity, learning, and entertainment. Here are some specific ways you can leverage AI: Productivity: Automation: Use AI tools to automate repetitive tasks, such as email sorting, data entry, and scheduling. Writing Assistance: Tools like Grammarly or ChatGPT can help with grammar, style suggestions, and generating content ideas. Data Analysis: Use AI-driven software for data analysis, visualization, and generating insights from large datasets. Creativity: Content Creation: AI can help generate text, music, or art. For example, tools like DALL-E or MidJourney for art, or ChatGPT for writing. Brainstorming: Use AI to generate ideas for projects, stories, or problem-solving. Learning and Research: Personalized Learning: AI can tailor educational content to your learning pace and style, using platforms like Khan Academy or Coursera that adapt based on your performance. Research Assistance: Use AI to summarize articles, extract key points, and find relevant research papers. Entertainment: Gaming: AI enhances gaming experiences through smarter non-playable characters (NPCs) and more realistic environments. Media Recommendations: Platforms like Netflix and Spotify use AI to recommend movies, shows, and music based on your preferences. Communication: Language Translation: Use AI-powered translation tools for communicating in different languages. Voice Assistants: Use assistants like Siri, Google Assistant, or Alexa for hands-free operation of your device and smart home controls. Security: Threat Detection: AI can help identify and prevent cybersecurity threats by analyzing patterns and detecting anomalies. Password Management: Use AI to manage and generate strong passwords. Integrating AI into your daily activities can significantly enhance efficiency, creativity, and overall experience on your laptop._ If you have AI built into your laptop, you will be able to get uselessly generic responses to questions such as this, SLIGHTLY FASTER, and WITHOUT AN INTERNET CONNECTION (except when the AI needs to be updated or when it needs to access data that isn't stored locally, which will be basically always). Please spend $500 more on your laptop for an AI accelerator, starving salespeople need to make their commissions somehow!
In interview few years ago Mike Clark said that Zen 5 is what he is specially excited for and it is arch that brings exciting things to the table. I assume he would be one to know.
That first bit of the video I was like, duh isn't all this obvious? I remembered sadly how incredibly dense people tend to be. Any-who. Great job clarifying the internal names. I can see all the names confusing some, and hurts the optics of your perceived credibility too. For example if someone latches onto Morpheus, then sees now it's called Medusa could be viewed as a change in the story. Thus, your your perceived credibility is harmed in their eyes because of their confusion. That's why I think it's important you cleared that up for those willing to listen. For those than desire you to burn and be wrong, that name change could be a can of BB's lol
Such a cool name. May have to go team red for that one. Also, good call on knowing Gorgons have wings. Now we just need Euryale's Halo and Stheno Point as follow up products.
very VERY interesting theory... I like it! it makes a ton of sense, and the fact that you taught me Medusa had wings just seals the deal! lol Thanks Tom!!!
There’s some argument nowadays that moving the I/O die to 3nm before the CPU would be beneficial. TSMC N3 has no scaling for SRAM (i.e. cache) but logic scales decently..
You could see a zen 5+ on 3nm, but I have my doubts. I think zen 6 will be on 3nm. 3nm will be a mature process for desktop chips. That would be my bet
I hope that won't delay a RDNA4 successor to Strix Point if Strix Halo gets introduced later. I hoped by end of 1H 2026 I could get a RDNA4 STrix Point successor Desktop APU :(
Yeah the force was strongest with rdna2 it withered with rdna3 and rdna4 even if they barely beat 7900xt depending on price can bring the force back to life for me
At this rate will be struggling whether to name my next laptop Euryale (far) or Stheno (forceful) as there are just too many reasons to delay a purchase if you can.
Like Zen 5, I'm sure the target demographic for these products will be the AI / big data sector. On the bright side, Zen 5 will be more accessible to the average person when Zen 6 is released, so at least there's that.
I was a bit disappointed in Strix but I was confused. I thought this Stix was Halo so yeah now I am again exited for Halo. Seems like a huge upgrade for APU users.
@@djchristian82 Yeah I know it hasnt been released yet but from the specs and relatively similar architecture you can deduce it's not that big of a difference at least Strix Point.
I still think AMD is sandbagging on Zen 5 lol. I love your videos! Always makes me feel at home when I watch one. Been watching since 2018! You had 10,000 subscribers. Congratulations, Tom, you are doing awesome!
I doubt that Zen 5 will be significantly faster than AMD has indicated so far, but I _do_ think it could be cheaper than most people are expecting, and I honestly think that's what the majority of people need, rather than extra performance for high-end components. For example, if the 9900X isn't cheaper than the current price of the 7950X3D (£450, $475), it's basically pointless, because it's unlikely to be faster at either gaming or workstation tasks, and would use about the same amount of power. Even if it is faster than expected, it would still be a hard sell for most people looking for a high-end CPU if it was the same MSRP as the 7900X ($550) when the 7950X3D is $75 cheaper. It's manufactured on an N4 process that's significantly cheaper now than N5 was when the 7000-series was new, and leaks indicate that the CCDs are also slightly smaller, so AMD can definitely afford to sell it for less. Either way, I suppose we don't have long to wait. We'll see which of us is right (or possibly that neither of us is right and that AMD has stupid launch pricing that no-one sane will pay for, yet again) in a couple of weeks.
@@Gorgula Not all that unlikely, CPU generational IPC is like a bag of mixed nuts, some stuff will be great while others will be about the same. I expect we'll see some nice increases in some areas, similar performance in others. One potential point for a sandbagging claim would be if zen 5 holds higher boost clocks under load, if zen 4 boosts to around 5ghz under full load, zen 5 holding 5.5 and only dipping a couple hundred mhz would be an even bigger improvement than suggested. I have no idea how the zen 3 benchmark ever made it to release, pretty absurdly wrong for anyone who has any idea on the performance of zen 3.
With all the delays it's important to keep in mind this is because of TSMC, and Nvidia and Intel using TSMC now and Intel buying up a lot of N3 capacity, so with this whole "they don't think they'll be able to launch on N3, that's because of the lack of capacity from TSMC. Zen 4 was supposed to be on N3 for the first year of N3 wasn't great and it was unknown when TSMC would have their issues worked out. They of course now have their 2nd gen N3 which is a bit better and they have better yields and quality with it. So this pushed AMD Zen 5 to N4 and Zen 6 to N3.
We shouldn't be too disappointed as N4P performs very close to N3(B) while likely being a lot cheaper to produce. N3E starts to widen the gap but for now, only Apple gets it...
Most of techsphere are not very accurate with actual pattern recognition when it involves obscure references lol. Intel really needs to up their game, the objective side of me is genuinely worried about Intel. Whereas the subjective side, is endlessly laughing away like Nelson from Simpsons, "Karma's got Intel Good" for stagnating progression for so damn long.
is it only me or LNL looks way more interesting than strix point ? we finally get XMX up scaling, and overall it looks like a very strong "GPU first" die, with half decent drivers that will be amazing for handhelds ( unlike MTL) I do wonder what will be the latency of P to E cores since its 2 different clusters and that may be a problem (and u cant play on 1 since 4 cors with no HT is just not enough) the handheld market really need one of 3 things: 1) LNL but with 6P cores (or to hope 4+4 will work well ) and good GPU drivers ( fingers crossed) 2) AMD to finally have a good ML based upscaling (RDNA 5) ? 3) nvidia to have like licensed X86 cores from intel or AMD and glue an ada lovelace/blakwall GPU to that ( ya 1 in a billion but we can still dream)
The only thing I can say is that I called the complex routing for the initial Rome SKUs... Everyone should be confident that FPx and AM5 chips will be have the new arch for Zen6... Even Apple had to use 4nm for chips last year because of delays in 3nm... Since 4nm is optically shrunk 5nm, 3nm will have a more profound update - even though SRAM isn't shrinking - so there will be advantages... But TSMC delays may also explain why Turin isn't 256 cores... They need a real shrink to get to a 32C CCX... Client and graphics are emerging more now because they have outpaced intel in server\HPC and eclipsed Nvidia in AI compute... 300A is a great platform for inference and is high margin to support client and graphics... Since OEMs dismissed 7900M XT, Strix Halo will FINALLY beat Nvidia enough to kill everything under 4070 mobile...
I don't see how a 4 P core/4 LP E core part from Intel was supposed to compete with a 16C/32T part from AMD. I think these are completely different market segments, form factors, and use cases these devices are targeting, so comparing them seems disingenuous. If I wind up being wrong on that and they occupy the same segments and are priced similarly sure, but right now I don't really see that.
What? I think you have this completely wrong. Persephone is Zen 4, Nirvana is Zen 5, Morpheus is Zen 6 for servers. Medusa is Zen 6 for clients. Divergent architecture. It's the codename of the core, not a design. Your contacts are speaking out their asses.
Zen 3 launched like 16 months after Zen 2, and in fact the 5950X only launched 12 months after the 3950X. That's all I am suggesting is possible. It would still come over a year after Zen 5. But it sounds unlikely to be a full 2 years later.
Maybe I'm just buying too much into the AI hype, but it would be nice to have a CPU that has a decent NPU on AM5 Desktop. Especially with Zen 6, that would be great.
They could put an NPU in the IO die, but I kinda doubt they do that, as efficiency isn't as important on desktop and (i)GPUs can be used to. But there will very likely be the usual -G APUs eventually, i.e. Strix Point on desktop, then Medusa Point (?) with zen6.
Successor of 8700g is something i need. 🗿 The raw processing power could be 30% more than 8700g And i gpu could be strong. Medusa has wings? Isn't that a snake hair women.
N4 compute die + N3 I/O die would be interesting. Or perhaps Strix Hal0 will share N3 Zen5 compute die with Turin classic instead for N3 + N3? Maybe that's what they mean by the "Ultimate" premium. Too bad, Turin, no iGPU means N6 I/O die for you! If so, then Zen5, Zen5c, and RDNA3.5 will all have both N3 and N4 versions. I'm starting to feel bad for my wallet and the engineers.
AMD PLEASE GIVE ME STRIX HALO FOR THE LOVE OF GOD AND GIVE IT VCACHE FFS just don't make it cost more then 3k for a maxed out laptop with 128GB of ram or 192?
I have bad news for you... AMD has only been peddling value offerings in laptops because they had to. If Strix Halo is the killer of all things it WILL be priced accordingly.
Key point being missed at intel and AMD: Nobody (except microsoft) cares about NPUs. They're useless and a waste of die space. And this is true of people working on AI. NPUs are pointless. They're completely unsupported for any important AI workloads. I care about the Strix Halo, because quad channel DDR5 is finally enough memory bandwidth to matter. But NPU? toss that in the trash, and just add more cache or something!
I wouldn't hold your breath. all rumors point to mainstream GPU under $600 at most. Performance between 7800 XT and 7900 XT. I'm 99% sure it will still only have 1 GCD.
Chiplet RDNA4 is supposedly cancelled entirely. RNDA3 didn't pan out the way they wanted it, so they're only doing the monolithic chips until they're 100% sure they can pull it off.
Medusa name points to a higher performance APU. If AMD could get the GPUs in the APU to scale up without big jumps in Power ie same performance to watts then they could finally start to beat Nvidia. Recent APUs have had impressive graphics with small wattage. Maybe new interconnect tech will allow AMD to bring better CPU & high performance Chiplet GPUs. I predict AMD becoming the goto for developers because they can mix high performance cpu(X86&Arm), gpu, AI, fpga, modules. It would be hard for Nvidia to compete as lack CPU(they have arm design but not X86) Intel have CPU X86 but not arm and are dabbling with GPU but need to state they are committed to staying in GPU long term there short term dables with Mcafee obtron&SSDs, I believe they made fpgas but not updated for a bit this all makes gpu people nervous. So AMD only one most hand held device, console small system trust . I read AMD have a good relationship with Tesla who dropped nvidia for being too controlling.
Sure, and let's talk about the Cypress Cove and Strix Halo leaks too. Would love to! :) (Actually I have talked about the Zen 3 leaks a dozen times btw)
[SPON: Use "brokensilicon“ at CDKeyOffer to get Win 11 Pro for $23: www.cdkeyoffer.com/cko/Moore11
Win 11 Pro ($23): www.cdkeyoffer.com/cko/Moore11
Win 10 Pro ($17): www.cdkeyoffer.com/cko/Moore10
Win10 Home ($15): www.cdkeyoffer.com/cko/Moore10h
Office 2016 ($29): www.cdkeyoffer.com/cko/Moore16
Office 2019 ($48): www.cdkeyoffer.com/cko/Moore19
25% software discount code: brokensilicon
3% OFF EVERYTHING discount code: dieshrink ]
A14aàà
Hi Tom, if China were to have an unexpected breakthrough in photonic chips could they potentially leapfrog 🐸 past Intel, AMD & NVIDIA¿?¡! ... ...... ......... 369
ruclips.net/video/fk38BdJlujk/видео.htmlfeature=shared
I can stay up for another 15 minutes I guess
Literally
Everytime
Got called out for work, then saw this pop up
Guess I’m staying up a bit more!
its 10 in the morning here :D
4:48am here :(
Medusa turns people to stone.
Stone like a _granite ridge_ or the sculptures of _Raphael._
It's actually a really cool way to combine the mythological greek avians, artists, and rock formations, of previous codenames.
If only the same amount of effort to maintain consistency was applied to the name of the final products.
Bad idea from my point of view. Medusa will turn your system in a useless block of stone. It's a scary code name.
Honestly I'd much rather tell people I'd just bought an AMD Strix Halo or Medusa than an AMD AI 300. Why are the engineers so much better at making cool names than the marketing teams?
@@Gen0cidePTB Engineers want you to remember their babies. Marketing teams just want to rip you off as much as they can, with and by any mean they can think off. like the AI trend thing.
3DFX still had the best code names, but I'm down with Medusa. RIP 3DFX Voodoo, Banshee, Rampage...
Engineering code names meet marketing product names, that is where the naming consistency comes from
Came for the leaks, stayed for the Greek mythology. I had no idea they had wings. WTF.
Came for the leaks, stayed because of the Greeks
@@shadowlemon69 Came for leaks, stayed for Greeks.
@@franzlyonheart4362 I came.
Quality comment.
Yeah those Greeks drank so much Red Bull :)
Put a polarizer filter on your camera lens to reduce the reflection on your glasses.
It really is a shame that investors gutted Intel to chase short-term profits in the 2010s. A smaller company is running circles around them in terms of the pace of improving their technology, developing new products, and delivering them to customers. AMD is competing so hard that I’m worried about pricing over the next couple of years as they don’t seem to have much incentive to keep prices low.
They didn't just gut it, they took all prospects of getting the new node done on time for ANY THING out the windows as they pre padded their bonuses in lieu of big fat federal monies. The way Intel share holders see it, the feds won't allow the company to fail because they want fabs more than they want Intel as a engineering firm.
Intel does still exist
And AMD has the very real possibility of competing with itself
No one will buy a R9 11950x if it costs 1000$ if the R9 9950x is 600$
@@bornonthebattlefront4883it starts with 9950X at $599. Then you get the 10950X at $649. Then the 11950X at $749 and so on. Just look at GPU pricing. The 980 was $549 10 years ago which is $724 today. The 4080 Super is selling well at $999 today. Without proper competition performance stagnates or prices rise which is not good for the consumer.
Especially when they have previous gen x3d chips too @@bornonthebattlefront4883
Yeah, AMD's increasing of prices is more aggressive than Intel's.
I think Arrow Lake will make AMD's sales department cry again like during Bulldozer times 🙂
I WAS ABOUT TO SLEEP TOM
ITS 6 IN THE MORNING TOM
It's 6 in the morning somewhere around the world at any one time.
*shrugs* ;)
Moore's Law is downing coffee
Always worth staying up for
3nm IOD sounds insane no matter how you would frame it.
4nm IOD and 3nm cores sounds much more probable.
Now I understand why that executive sitting next to Lisa Su promised that Zen 5 was coming to 3nm as well.
I would bet on 3nm refresh, too. They already completed the CPU on 3nm. Another nut kick for lunar turd.
Zen 5c is on 3nm with 16 core chiplets with unified L3. AMD wants to deliver an "Epyc Dense" that absolutely demolishes Intel in performance per watt, performance per socket and performance per dollar. They already have two of the three, which gives them the lowest TCO, but they want the lowest up-front costs as well as the lowest TCO.
Whether we'll see Zen 5c on Desktop is anyone's guess, but I doubt we will see Zen 5 "standard" ported to 3nm at this point. I think it will make more sense for them to shoot straight for Zen 6 on 3nm on desktop for a more impressive uplift.
Medusa for a cpu arch name is metal AF🤘
i have never thought about the codenames much. But the explanation you delivered at the start was eye opening. Great work!
AMD and Intel have been doing it for years, different codename for architecture, platforms, CPU cores, and chipsets.
There's enough commonality where if you uncover a codename, you know the implications.
You know what also have wings?...Redbull
Maybe it will get an AI too? Who knows, noone knows
Nah that only gives you wings, until you get sued then it gives you wiiiiiiiiiiiiiings
@@fedjapetrovich2408 I wonder what headlines we could get...
_Max Verstappen retires. Red Bull F1 Team signs new AI driver._
The worst energy drink. 🤢
And annoying marketing. They pay decent looking girls to drive around and hand out free cans.
I short circuited their brains when I told them I didn't want that trash and ignored them.
And you know who else has wings?
MY MOM! OOOOOH
11:43 Should have been colored red, but wow was that hot. Its been years and years since I've seen designs like that. Hardware companies used to use designs like that to advertise in PC hardware magazines.
Oh and good leak. I'm really glad when I upgraded from LGA 1150 in 2022, I went to AM5. Zen6 looks like its going to be killer.
Hey, Tom. I've been a huge fan for a while and my request for you is this: Please make 5-10 tutorial videos over the next year, and make a "if you're new to this" playlist, or similar where people your viewers share with can understand your content better. I ask this because I have a few really technical friends, that are just lacking the right amount of technicality to care about what you're saying.
That might be helpful for both viewers and this channel. Hopefully Tom or another one from the team sees this.
I know it's the flavour of the year, but I still haven't found any compelling AI applications. When they work, it's cool, but the hallucination factor makes it untrustworthy to me. Until that changes I'd love to see a product without the NPU and using that die space for something I'd actualy use.
In science AI based tools are a really big thing, not hyped, but actually already used. There AI is just integrated where it makes sense, combined with traditional approaches etc.
That said, those applications are highly specialized, not of that universal generative AI that people are talking about non-stop. Funnily enough, those people using those tools, have little use for those weak NPUs on thoe APUs. They need workstation graphic cards with lots of RAM and also lots of SSD capacity.
It will be interesting to see if actual use cases for those NPUs within off the mill APUs will materialize and solidify.
Use RAG to reduce hallucinations - add specific content to have it work with. You can do that via prompts or you can use a data lake to help.
Considering the promised longevity of AM5 platform. I have high hopes for Zen 6 and maybe a couple more generations after it. Zen 6+ or something
What about zen 6 x3d without SMT just for gaming with second CCD with Dense cores.
AM5 platform longevity till Zen 7 🤞
Will probably be left over worse yields converted to lower core + vcache models. Hopefully they don’t do any xt models, pointless products.
@@APU-iGPU I doubt you'd want to be restricted to am5's limits by then.
@@102728 Limits as in what?
I have a Laptop with an AMD TPU, but i can not use it because AMD did not release any driver, framework or tool support to use the TPU on linux - at the moment AMD has to UP there software, if there don't it do not matter how many TOPS the TPU has if i can't use it.
Aye mature software support is something AMD absolutely has to deliver
@@Martin1991zab Just curious but what is the actual use case for a TPU? I get the point of having such a limited low power device for Tensor and Matrix maths, but what can actually be done with that application wise, and when is it better than just spinning up the GPU to 200-400mhz and doing it there? Even then the GPU has multiples more performance.
The question is, which Zen will we get to before I definitely need to upgrade from my 5800x3d system
I'm betting Zen 7. 4 gens will be plenty, and you get to jump on AM6
5800x3D it's the RTX 1080ti of CPUs, if you only game you are Gucci for a long time.
As another person with a 5800X3D, I won't need to upgrade my CPU until we get to _at least_ Zen 6, probably Zen 7 or 8.
This will be dependent on GPUs getting powerful enough that they would be significantly bottlenecked by the 5800X3D, and software becoming CPU-intensive enough and/or local AI-acceleration becoming pervasive enough that the 5800X3D is no longer sufficient.
@@nathangamble125zen 8 is 10 years ahead bro
If you plan to ride your current system into the sunset, probably a good 4-8 more years, but the real answer is the day you buy a new GPU. 5800x3d is already a limiting factor with a xtx 4080 or 4090 even at 1440p, any gpu coming out in the next gen or two would be a killshot for it and would most certainly need a faster cpu.
Another thoughtful video from you, Tom. I don't know what people from tech sites are posting or if it is already just AI summaries, but thank you for bringing some human thought into it.
bro, Friday, and I'm alone. Chilling have no plans, and even though I gave up, I'm still here. 🍻
A couple of weeks ago I asked about the STRIX HALO release date after you updated the STRIX Point release date. Now you released this video spot on after and I appreciate the effort! Exciting news as I plan to upgrade to a laptop to replace the curent desktop but need one that is just a strong. Medusa will be that Laptop I believe.
Now I have to wait for zen 6
Time travelled back to Feb! TSMC 2nm process, the internal codename for AMD's Zen 6 cores is Morpheus.
What exactly am I supposed to use AI on my laptop for?
This answer was generated by ChatGPT:
_Using AI on your laptop can enhance productivity, creativity, learning, and entertainment. Here are some specific ways you can leverage AI:
Productivity:
Automation: Use AI tools to automate repetitive tasks, such as email sorting, data entry, and scheduling.
Writing Assistance: Tools like Grammarly or ChatGPT can help with grammar, style suggestions, and generating content ideas.
Data Analysis: Use AI-driven software for data analysis, visualization, and generating insights from large datasets.
Creativity:
Content Creation: AI can help generate text, music, or art. For example, tools like DALL-E or MidJourney for art, or ChatGPT for writing.
Brainstorming: Use AI to generate ideas for projects, stories, or problem-solving.
Learning and Research:
Personalized Learning: AI can tailor educational content to your learning pace and style, using platforms like Khan Academy or Coursera that adapt based on your performance.
Research Assistance: Use AI to summarize articles, extract key points, and find relevant research papers.
Entertainment:
Gaming: AI enhances gaming experiences through smarter non-playable characters (NPCs) and more realistic environments.
Media Recommendations: Platforms like Netflix and Spotify use AI to recommend movies, shows, and music based on your preferences.
Communication:
Language Translation: Use AI-powered translation tools for communicating in different languages.
Voice Assistants: Use assistants like Siri, Google Assistant, or Alexa for hands-free operation of your device and smart home controls.
Security:
Threat Detection: AI can help identify and prevent cybersecurity threats by analyzing patterns and detecting anomalies.
Password Management: Use AI to manage and generate strong passwords.
Integrating AI into your daily activities can significantly enhance efficiency, creativity, and overall experience on your laptop._
If you have AI built into your laptop, you will be able to get uselessly generic responses to questions such as this, SLIGHTLY FASTER, and WITHOUT AN INTERNET CONNECTION (except when the AI needs to be updated or when it needs to access data that isn't stored locally, which will be basically always). Please spend $500 more on your laptop for an AI accelerator, starving salespeople need to make their commissions somehow!
The blockchain synergy in the cloud.
To boost tech stock prices. There are zero applications.
The forbidden question !
Run ollama
Hmm if tape out is Q2 2025, production 6 months later is extremely aggressive, though that definitely implies no retail availability until 2026
Can't wait for the Zen 6 desktop leaks. I'm not pulling the trigger on upgrading with zen 5 until I have an idea of how zen 6 will look like.
In interview few years ago Mike Clark said that Zen 5 is what he is specially excited for and it is arch that brings exciting things to the table. I assume he would be one to know.
That first bit of the video I was like, duh isn't all this obvious? I remembered sadly how incredibly dense people tend to be. Any-who. Great job clarifying the internal names. I can see all the names confusing some, and hurts the optics of your perceived credibility too. For example if someone latches onto Morpheus, then sees now it's called Medusa could be viewed as a change in the story. Thus, your your perceived credibility is harmed in their eyes because of their confusion. That's why I think it's important you cleared that up for those willing to listen. For those than desire you to burn and be wrong, that name change could be a can of BB's lol
Medusa Halo is a cool CPU name
I don't share out a lot of stuff, but there has to be some computer nerds somewhere on my feed the need some juicy details. Appreciate the video 🤘
Thanks for sharing!
Such a cool name. May have to go team red for that one. Also, good call on knowing Gorgons have wings. Now we just need Euryale's Halo and Stheno Point as follow up products.
Oh snap look at me go
Oh snap look at you go
I read that in Ricky Gervais's voice
@@Westlee_ Oh snap look at you go go.
we need news on the X3D. Im trusting on you bro!
He has already covered it extensively. It will not take as long as Zen 4 X3D, but it will still take a little while.
Any news about a Z2 Extreme for handhelds maybe based on strix?
I would be very surprised if there wasn't a special bin of both Strix and Kraken for handhelds released by early next year. 😊
very VERY interesting theory... I like it! it makes a ton of sense, and the fact that you taught me Medusa had wings just seals the deal! lol
Thanks Tom!!!
There’s some argument nowadays that moving the I/O die to 3nm before the CPU would be beneficial. TSMC N3 has no scaling for SRAM (i.e. cache) but logic scales decently..
Nice vudeo Tom. Do you see AMD making a shrunk down Zen5 on N3 before zen6. Does it even make sense?
You could see a zen 5+ on 3nm, but I have my doubts. I think zen 6 will be on 3nm. 3nm will be a mature process for desktop chips. That would be my bet
I hope that won't delay a RDNA4 successor to Strix Point if Strix Halo gets introduced later. I hoped by end of 1H 2026 I could get a RDNA4 STrix Point successor Desktop APU :(
I have a feeling that Medusa is going to be an extremely interesting chip.
I was really hoping for a Q4 Strix Halo, even though I know that was wishful thinking.
I wish for RDNA 4 and 5 to bring balance to the Force.
Yeah the force was strongest with rdna2 it withered with rdna3 and rdna4 even if they barely beat 7900xt depending on price can bring the force back to life for me
They kill all the younglings!
😂
At this rate will be struggling whether to name my next laptop Euryale (far) or Stheno (forceful) as there are just too many reasons to delay a purchase if you can.
Like Zen 5, I'm sure the target demographic for these products will be the AI / big data sector. On the bright side, Zen 5 will be more accessible to the average person when Zen 6 is released, so at least there's that.
50% IPC increase. It's going to kill Intel. 😅
I was a bit disappointed in Strix but I was confused. I thought this Stix was Halo so yeah now I am again exited for Halo. Seems like a huge upgrade for APU users.
@@djchristian82 Yeah I know it hasnt been released yet but from the specs and relatively similar architecture you can deduce it's not that big of a difference at least Strix Point.
I'd like an insight on zen 6 chiplet. Mostly are we getting more than 8corrs per ccx? Why not make them 10/12 :p. O 16 no smt. Ooh imagine
I'm really hoping to see an NPU on desktop Ryzen soon-ish (no, those weird 8700 APUs don't count).
I come for the cdk offers
Something slightly odd about the audio in this one - like a tiny echo maybe? I can't quite tell but it just sounds a little worse than usual.
Let's go keep on Ballin
I have a source leaking that says Zen 6 would be faster than Zen 5.
And it will release after Zen 5, before Zen 6 with improvements.
Any thoughts on CPU and GPU benchmark leaks on LunarLake?
Listen to the next Broken Silicon
I still think AMD is sandbagging on Zen 5 lol. I love your videos! Always makes me feel at home when I watch one. Been watching since 2018! You had 10,000 subscribers. Congratulations, Tom, you are doing awesome!
Sandbagging about zen 5 but straight up lying about am4 re-releases simultaneously? Seems unlikely.
@@Gorgula It doesn't bode well for the claimed 16% IPC uplift either then...
I doubt that Zen 5 will be significantly faster than AMD has indicated so far, but I _do_ think it could be cheaper than most people are expecting, and I honestly think that's what the majority of people need, rather than extra performance for high-end components. For example, if the 9900X isn't cheaper than the current price of the 7950X3D (£450, $475), it's basically pointless, because it's unlikely to be faster at either gaming or workstation tasks, and would use about the same amount of power. Even if it is faster than expected, it would still be a hard sell for most people looking for a high-end CPU if it was the same MSRP as the 7900X ($550) when the 7950X3D is $75 cheaper. It's manufactured on an N4 process that's significantly cheaper now than N5 was when the 7000-series was new, and leaks indicate that the CCDs are also slightly smaller, so AMD can definitely afford to sell it for less.
Either way, I suppose we don't have long to wait. We'll see which of us is right (or possibly that neither of us is right and that AMD has stupid launch pricing that no-one sane will pay for, yet again) in a couple of weeks.
Thanks for watching!
@@Gorgula Not all that unlikely, CPU generational IPC is like a bag of mixed nuts, some stuff will be great while others will be about the same. I expect we'll see some nice increases in some areas, similar performance in others. One potential point for a sandbagging claim would be if zen 5 holds higher boost clocks under load, if zen 4 boosts to around 5ghz under full load, zen 5 holding 5.5 and only dipping a couple hundred mhz would be an even bigger improvement than suggested. I have no idea how the zen 3 benchmark ever made it to release, pretty absurdly wrong for anyone who has any idea on the performance of zen 3.
I just want Zen6 '___800X3D' to come out on AM5.
Lol, I dislike the name of Medusa. But if it's 32 core, I will be looking forward it.
Nice way to spend the last 15 mins of my lunch break
Tom keeps teasing us with Zen 6/7 info.
I won't be upgrading from phoenix but this is torture.
Zen 10, coming soon. 1 million cores, 100 ghz, 10 T of cache. Uses 1 watt. RUclipsrs suggest waiting a little longer for the next model. 😅
If I install a Medusa cpu will I have to use a mirror to turn computer off?
just get a laptop with a glossy display, problem solved
Yes
Do you talk about the tech specs of CPUs, or just the gossip of it?
man the new look with glasses is pretty surprising ngl
With all the delays it's important to keep in mind this is because of TSMC, and Nvidia and Intel using TSMC now and Intel buying up a lot of N3 capacity, so with this whole "they don't think they'll be able to launch on N3, that's because of the lack of capacity from TSMC. Zen 4 was supposed to be on N3 for the first year of N3 wasn't great and it was unknown when TSMC would have their issues worked out. They of course now have their 2nd gen N3 which is a bit better and they have better yields and quality with it.
So this pushed AMD Zen 5 to N4 and Zen 6 to N3.
We shouldn't be too disappointed as N4P performs very close to N3(B) while likely being a lot cheaper to produce. N3E starts to widen the gap but for now, only Apple gets it...
Most of techsphere are not very accurate with actual pattern recognition when it involves obscure references lol. Intel really needs to up their game, the objective side of me is genuinely worried about Intel. Whereas the subjective side, is endlessly laughing away like Nelson from Simpsons, "Karma's got Intel Good" for stagnating progression for so damn long.
I just wonder in the next 5 or so years what will be worth upgrading my r9-7950x to? It's a really nice cpu.
Looking Diferent, Looking Good 👏👏👏.
is it only me or LNL looks way more interesting than strix point ?
we finally get XMX up scaling, and overall it looks like a very strong "GPU first" die, with half decent drivers that will be amazing for handhelds ( unlike MTL)
I do wonder what will be the latency of P to E cores since its 2 different clusters and that may be a problem (and u cant play on 1 since 4 cors with no HT is just not enough)
the handheld market really need one of 3 things:
1) LNL but with 6P cores (or to hope 4+4 will work well ) and good GPU drivers ( fingers crossed)
2) AMD to finally have a good ML based upscaling (RDNA 5) ?
3) nvidia to have like licensed X86 cores from intel or AMD and glue an ada lovelace/blakwall GPU to that ( ya 1 in a billion but we can still dream)
MI450 packaging presumably means CoWoS right? Is an OSAT about to start doing CoWoS?
These intros feel like Tourniquet intros. Lengthy. 😂🤣 (high praise)
Edit: also, Gorgons had wings!? I do need to brush up.
I miss when AMD had their best APUs for affordable prices.
The only thing I can say is that I called the complex routing for the initial Rome SKUs... Everyone should be confident that FPx and AM5 chips will be have the new arch for Zen6... Even Apple had to use 4nm for chips last year because of delays in 3nm... Since 4nm is optically shrunk 5nm, 3nm will have a more profound update - even though SRAM isn't shrinking - so there will be advantages...
But TSMC delays may also explain why Turin isn't 256 cores... They need a real shrink to get to a 32C CCX...
Client and graphics are emerging more now because they have outpaced intel in server\HPC and eclipsed Nvidia in AI compute... 300A is a great platform for inference and is high margin to support client and graphics... Since OEMs dismissed 7900M XT, Strix Halo will FINALLY beat Nvidia enough to kill everything under 4070 mobile...
I don't see how a 4 P core/4 LP E core part from Intel was supposed to compete with a 16C/32T part from AMD. I think these are completely different market segments, form factors, and use cases these devices are targeting, so comparing them seems disingenuous. If I wind up being wrong on that and they occupy the same segments and are priced similarly sure, but right now I don't really see that.
I agree, it is disingenuous. So tell Intel to stop comparing them.
Bro knows more than the employees 💀
Let's just assume everything has a wing, be it Granite ridges or lakes
Hey Clarke, have you seen Superman? ;-)
What? I think you have this completely wrong. Persephone is Zen 4, Nirvana is Zen 5, Morpheus is Zen 6 for servers. Medusa is Zen 6 for clients. Divergent architecture. It's the codename of the core, not a design. Your contacts are speaking out their asses.
Looks like intel tried to get ahead of amd with ai but amd changed their schedule and chip layout to catch up and barely outperform intel.
At this point I think AMD might be sandbagging Zen5/6, knowing the competition is floundering, aka pulling an Intel.
ROCKIN' some Sweet Glasses!
Hold on Zen 5 is about to come out in July 2024 and your saying Zen 6 in 2025 . That one year to sale out ALL zen 5 in one year. I don't think so.
Zen 3 launched like 16 months after Zen 2, and in fact the 5950X only launched 12 months after the 3950X. That's all I am suggesting is possible. It would still come over a year after Zen 5. But it sounds unlikely to be a full 2 years later.
Maybe I'm just buying too much into the AI hype, but it would be nice to have a CPU that has a decent NPU on AM5 Desktop. Especially with Zen 6, that would be great.
They could put an NPU in the IO die, but I kinda doubt they do that, as efficiency isn't as important on desktop and (i)GPUs can be used to. But there will very likely be the usual -G APUs eventually, i.e. Strix Point on desktop, then Medusa Point (?) with zen6.
Why would AMD name their CPUs after ancient monsters? Are they warning us about what AI underneath really is? *X-Files music plays*
doesn't a gorgon and Medusa turn people to stone much like granite? Iunno
So that means they want us to skip Zen 5?
Is zen 6 ryzen getting 32 cores using AM5 socket?
I figured that Medusa was the actual AI controller. I dunno, that makes more sense to me
great infos, thanks 😊
Successor of 8700g is something i need.
🗿 The raw processing power could be 30% more than 8700g
And i gpu could be strong.
Medusa has wings? Isn't that a snake hair women.
Dogs that look like their owners!
Always love seeing the leaks and the progress that AMD is making on Intel… but my lizard brain just constantly mumbles… “Is better than 7800x3D?” 😂
N4 compute die + N3 I/O die would be interesting. Or perhaps Strix Hal0 will share N3 Zen5 compute die with Turin classic instead for N3 + N3? Maybe that's what they mean by the "Ultimate" premium. Too bad, Turin, no iGPU means N6 I/O die for you! If so, then Zen5, Zen5c, and RDNA3.5 will all have both N3 and N4 versions. I'm starting to feel bad for my wallet and the engineers.
Is the strix point n4x or n4p
Why 3nm iod? Was that not the point
AMD PLEASE GIVE ME STRIX HALO FOR THE LOVE OF GOD AND GIVE IT VCACHE FFS
just don't make it cost more then 3k for a maxed out laptop with 128GB of ram or 192?
I have bad news for you... AMD has only been peddling value offerings in laptops because they had to. If Strix Halo is the killer of all things it WILL be priced accordingly.
I love your info and vids!
Thanks mate! Spread the word!
wow medusa is such a rock hard name makes me cool 😮
Key point being missed at intel and AMD: Nobody (except microsoft) cares about NPUs. They're useless and a waste of die space. And this is true of people working on AI. NPUs are pointless. They're completely unsupported for any important AI workloads. I care about the Strix Halo, because quad channel DDR5 is finally enough memory bandwidth to matter. But NPU? toss that in the trash, and just add more cache or something!
Chippets took the world by storm, imagine rdna4 how good it will be
RDNA4? Imagine what RDNA5 will be like!
I wouldn't hold your breath. all rumors point to mainstream GPU under $600 at most. Performance between 7800 XT and 7900 XT. I'm 99% sure it will still only have 1 GCD.
You must be very far out of the loop since its RDNA5
Chiplet RDNA4 is supposedly cancelled entirely. RNDA3 didn't pan out the way they wanted it, so they're only doing the monolithic chips until they're 100% sure they can pull it off.
Medusa name points to a higher performance APU. If AMD could get the GPUs in the APU to scale up without big jumps in Power ie same performance to watts then they could finally start to beat Nvidia. Recent APUs have had impressive graphics with small wattage. Maybe new interconnect tech will allow AMD to bring better CPU & high performance Chiplet GPUs. I predict AMD becoming the goto for developers because they can mix high performance cpu(X86&Arm), gpu, AI, fpga, modules. It would be hard for Nvidia to compete as lack CPU(they have arm design but not X86) Intel have CPU X86 but not arm and are dabbling with GPU but need to state they are committed to staying in GPU long term there short term dables with Mcafee obtron&SSDs, I believe they made fpgas but not updated for a bit this all makes gpu people nervous.
So AMD only one most hand held device, console small system trust .
I read AMD have a good relationship with Tesla who dropped nvidia for being too controlling.
I blame God of War franchise for my ignorance for not knowing about Medusa's having wings in mythology.
Wanna talk about the zen three leak? Ok lets talk about the zen three leak
Sure, and let's talk about the Cypress Cove and Strix Halo leaks too. Would love to! :)
(Actually I have talked about the Zen 3 leaks a dozen times btw)