Zen 5 Laptops are HERE - Ryzen Ai 300 Series Explained
Вставка
- Опубліковано 25 чер 2024
- The Zen 5 based Ryzen AI 9 HX 370 and Ryzen Ai 9 365 are coming to gaming laptops and thin and light laptops a bit sooner than expected. The end result is the Ryzen 300 series or what was called Strix Point competing directly against Intel Lunar Lake and Qualcomm Snapdragon X Elite and X Plus. But will it be enough?
Check out the Montech Sky Two GX (Sponsor) - geni.us/SKYTWO
Timestamps:
0:00 - Ai, Ai, Ai, Ai...and more Ai.
1:19 - More Naming Confusion & an Opportunity
2:49 - Product Name Decoding...
3:53 - Sponsor Spot
4:23 - Zen 5 is Here
5:13 - Zen 5 vs Zen 5c Explained
6:50 - Processor Scaling to 24 Threads
7:34 - The RDNA 3.5 Mystery
8:20 - XDNA 2 for Ai
8:50 - Ryzen Ai 300 Series Lineup (For now)
10:17 - Performance vs Apple M3 & Lunar Lake
11:58 - Still Many Questions
GET MERCH ►hwcstore.com
SUBSCRIBE ►bit.ly/SubHWC
This video is sponsored by Montech. As an Amazon Associate we earn from qualifying purchases.
Follow us for more updates!
Twitter bit.ly/HWC_Twitter
FaceBook bit.ly/HWC_Facebook
Instagram bit.ly/HWC_Instagram
--------------------------------------------------------------
#ryzenai - Наука та технологія
Going with 300 series just to be ahead of Intel's 200 is pretty funny. MORE NUMBER MORE BETTER!
They are actually scared of lunar lake XD.
They did the same with the motherboard chipsets. >
Don't for get AI 😂
Just copying Intel's confusing marketing. i5 is bigger than i3 etc. I find it hilarious that Intel made fun of AMD and then copied them with the MOAR CORES meme.
@@johnt.848but Intel just didn't gave us only more core unlike AMD
they gave us variety of cores
Saw adverts for a vacuum cleaner the other day, powered by AI!!... thats when you know the marketing bs has gone too far
AI toilet paper in the works.
Slapping the letters "AI" on your product is a great way to make me immediately not care about it
Still you clicked and commented
@@fratista You know AMD is a chip maker and not a youtube channel, right?
But at least for these APUs it´s true, they ahve an accelerator explicitly made to run neural networks (NPU)
The naming was what was disappointing about 7000 series, not the CPUs themselves.
The 7000 series was all over the place. Most of the lineup was rebranded Zen 2 & 3 chips.
@@HardwareCanucks To be fair, the Zen 2 were new silicon with RDNA2 instead of Vega, the Zen 3 were literally 5000 and 6000 series rebranded
@@HardwareCanucks hence why they should have given them 4000 naming, 5000 and 6000 naming according to their architecture, so all buyers would know these are just old CPUs essentially.
@@TabalugaDragonthey were old CPU architectures but to AMDs credit they did have a die shrink to 6nm.
@@puran1313 they deserve no credit(the marketing team and higher-ups I mean) if they make old stuff look like new, much better stuff.
Excellent breakdown. Thank you.
Love From Bangladesh most awaited video sin's last November
New naming is okay but only having “HX” or nothing with no wattage rating means battery life(the most important part of a laptop with any decent snappy cpu) will be hard to guess.
Also the regular and C cores, do they have to go over infinity fabric? In which case that a loss of power right there, and you might be better served with 8 monolithic cores where power isn’t lost on chip to chip communication. I’m really not clear on this last part, and am unaware of how it went with the zen4 mixed parts.
Strix point is monolithic
When the GeForce 200 series came out, they made up a 100 series using the old G90 line of chips. Nothing has ever been named and renamed more than the G92b, the chip where it all started ...
Well there goes that decoder ring
The Default TDP is less for these processors as compared to Ryzen 8000 series. As of now the ones they have released and can be seen on their site is not really meant for gaming I think, just for thin and light laptops.
Hope this Apu’s cones to desktop asap. I will build SFF with it.
Keep an eye out with minisforum. They have a last gen itx mb.
The naming basically is about hardware. AI means NPU. HX means the max possible configuration (CPU cores, cache, GPU cores, etc.). All have the same default TDP (28W IIRC), and their cTDP range is from 15W to 45W. The actual TDP will depends on vendor and it's up them to list it (or allow user to choose it).
You have a few things wrong there.
They don't all have the same default TDP. There is no default TDP.
Max wattage is 54W
@@HardwareCanucks I have seen a few news articles (and maybe a video too, don't recall). Quote windows central: "The Ryzen AI 9 HX 370 has a base TDP of 28W, but the configurable TDP (cTDP) ranges from 15W to 54W." Same with techpowerup, base TDP is 28W. It feels that these are designed targeting 28W TDP, and cTDP (what you can actually configure) is 15W to 54W depending on your product capabilities/market.
There is also an image from AMD (see, for example, notebookcheck). That shows 28W TDP, 15-54W cTDP for both chips.
Love to see the ThinkPad in shot
Are these Apu released for PC or Laptop??
The 9700X sounds good but I'll wait for the X3D version which should be released later, meanwhile I might upgrade my AM4 system.
The 25% CU bump over the Radeon 780 probably means Radeon 890 = RX 6500XT (or RX 590 if clocked really high)
It's a 33% jump bro, you counter backwards.
@@10001000101 even better then if real life performance surprises me with being on-par with a GTX 1660.
until something flashier comes along, ai is the new nft
NFTs are just JPEG. Kinda stupid to compare a JPEG to an AI. That flashy thing is gonna be built on AI.
@@CasualG-merwhat a dumb statement after so many people lost their money "investing" in the "blockchain technology" that's was about to change the world, same like the "AI" or the "fusion energy is just around the corner". We see how the humane ai and rabbit got their investors lose their money already. plug your memory cells before trying to call something stupid
What they dont show you is that the 5c cores have much less L3 cache compared to 5 cores.
That Blender performance compared with the M3 graph though, if it's true, I might as well get a notebook with these chips.
battery life?
Does the ai cpu chips improve overall performance in terms of processing speed and gaming? 🤔🤔🤔🤔
Nope, AI is rubish…
Nice upgrade, overall, though personally think that the "AI" shouldn't have not have been part in the naming scheme, looks really stupid, ex: Ryzen 9 HX 370. Will I upgrade my Elite book 835 7840U/32GB, let's see in 2025, most likely not in 2024.
What is the architecture of the chips? x86?
Yes of course it still x86, when they eventually do make it arm you will hear them say it as much as they mention ai now...
I just wish they stick with the same system so folks like me who left the tech world for a bit can still catch up.
I like the 7840u naming 8945hs also makes sense when you get used to it. Just that it’s difficult to know what performance they pack when a weird combination comes around like eg 7530u.
Why do everybody like to compare with the m3. These are HX parts right, they should go straight for m3 max. Even m4 is faster than some base m3 pro in some CPU test. It’s like 185H trying to compare with M3, ridiculous no?
But I do like 7840u, even 8840u still very decent. Would get that if I can find it.
If you run the 300 at 54w then yes compare against m3 max. If you run it at 28w then compare against m3 pro. 15w to 20w? Compare against m3.
I think that AMD is being very forward/freaking a bit (with their branding included lol) on the mobile market because I believe they are aware they misread Intel(while at the same time perfectly reading the market)
If you look at how things stand, AMD has a great product, with a greatly simplified production, basically a single SKU for the mobile market with yields and binning providing everything else. A win from a BoM point of view yet...
On the thin and light market they are going against Lunar Lake (I think AMD thought it would be delayed a bit) which:
a) Will have faster ST performance (Lion Cove is faster than Zen 5) for everyday tasks as you have 4P vs 4-higher clocking cores.
b) On the MT side of things while AMD is sure to win as it's x4-8 Zen 5c with SMT (depending on the SKU) which are faster than the 4E+2LP Skymont cores on Lunar Lake, Skymont* should be more efficient/draw less power overall.
c) Lunar Lake will, probably, be more efficient than Ryzen AI 300 at low TDPs, as it's much more targeted, has a better node (N4 vs N3B) and uses On-Package Memory, however this leaves ample room for AMD in the 30-45W segment.
d) Both NPUs are comparable and that's what matters, as whoever is developing AI workloads is probably targeting Microsoft's 40 TOPS target not an specific hardware. So it's irrelevant. Power efficiency is what matters then, here we have no idea but I'm inclined to believe has the better architecture for the NPU.
- Come decemeber AMD will be competing on the performance segment with Arrow Lake-H (if Intel doesn't keep delaying/missing targets lol...) which is 6P+8E cores vs 4(Z5)+8(Z5c).
Arrow Lake H's CPU tile is significantly more performant than Lunar Lake's CPU tile as it uses Intel A20 (PowerVia + RibbonFET) rather than TSMC N3B. So here AMD needs Strix Halo to be competitive on MT workloads while losing in ST/Low NT.
• But Arrow Lake's NPU performance is abysmal though and I don't know how important it will be for the target audience I could very well see, for example, Adobe apps leveraging NPUs to such a degree as to make x1.25 CPU/GPU performance (either way) irrelevant in a couple of years.
IMHO AMD is well aware that while their product stack is great right now (with much better margins BTW), Intel's cadence and strategy could very well get AMD in a hard place in 1-2 generations more(on the mobile market). There should be people at AMD who remember what happened with Core.
Still Intel's current execution is still a mess nonetheless. So it could very well be that Intel won't deliver long term.
Also Intel is throwing everything and the kitchen sink manufacturing and packaging-wise to get efficiency improvements, which AMD could also eventually take advantage of.
*Particular notice should be given to Intel's Atom effort as the team has been making impressive strides in the last 7 years. It would be ironic, even poetic, if in the end Intel's savoir turns out to be the humble Atom core.
So... What's the performance per watt here? Have they solved the x86 power hungry hair dryer issue or not? Do we get more than 2-3 hours of battery life on a high performance CPU?
The power problem actually has more to do with the OS and how it handles things like standby/sleep modes. ARM only looked so efficient because Intel got lazy and stagneted.. With both AMD & Intel innovating ARM is far weaker per watt. Now we need Windows to be redesigned for mobile or pick a more mobile friendly OS to maximize battery life.
@@tj_2701Windows has been redesigned for ARM.
@@tj_2701 I have Windows 11 ARM running in a VM on my MBP 16" M1 Max. Better thermals and power usage than any x86 CPU by miles even with an unoptimized OS.
Meanwhile, my Dell Precision 5560 laptop that I use for work is constantly revving its fans with negligible CPU usage.
How quickly we have forgotten that MBPs used to be Intel and had thermal issues. It'll be a while before x86 gets its act together and by then it may be too late.
Looks like my next laptop will be an AMD APU for a linux build. If I can leverage that NPU locally with ollama, I'll be in heaven.
All i want is the Asus G16 but with one these new cpus
This is the most detailed video on Zen 5, thank you.
best wishes from India.
It's funny to me I had a professor who used to work for Intel. He told us when they were designing chips they used "AI" to run extensive testing and simulation that they couldn't. This was easily 15 years ago when he was doing this but now magically we are using AI in processor design for the first time. Yeah okay
Wow, fucking up another naming sceme and making another Nvidia laptop GPU fiasco? I hate this industry sometimes.
What is fuck up about it 370 or 365?
Did I miss something? What GeForce fiasco?
@@HardwareCanucks Everything around laptop GPU power limits. AMD is repeating history.
@@bottomtext5872 That comes down to the laptop manufacturers, not Intel, Nvidia or AMD. If the CPU and GPU manufacturers were serious they wouldn't make the TDP configurable and the laptops would have to be designed to suit the limits.
@@bottomtext5872 you mean laptop GPU specs? because desktop 4070 for example has 28% more cores and 50% more vram compared to the laptop version. Power limits matter, but raw specs matter more. 4080M for example is 68% faster than the laptop version and is on par with the desktop 4070 in performance and vram amount.
If I never have to hear the word AI again, it'll be too soon.
The new naming scheme is better than the previous one, and the average consumer brains do work on the logic of bigger number better.
Average consumer would just get macbook not amd or intel
When is Strix Point Halo 40CU coming out?
I believe that this is bs.
40 CU is more than the Rx 7600, wich have a 160w TDP.
It more than even the PS5 GPU.
And all that power consumption and die area, on computers with tiny memory bandwidth.
280 GB/s on rx 7600, compared to 60-80 GB/s of system memory.
@@rj7250a They have already shown it's 40cu
its not they have it on public roadmaps lol@@rj7250a
Asus tuf 14?
why not just call them like AI 7 Mobile or AI 9 Station. IDK why confusing people is a must
nobody
absolutely nobody
AMD copying Intel's horrible naming scheme
what about linux support?
Amd usually does pretty good on linux support idk if you can do anything with the ai shit on linux tho
The number 300 does not appear to have anything with the Intel naming. It might be 3rd processor in a row with AI implementation or something. It would be so lame to put bigger number then a competition, and they know it.
xDNA1, xDNA1+, xDNA2 so 3rd "gen" so marketing can call it 300 and be one "better" than Intel. Hawk Point was 60% more TOPs so effectively a "plus" version of xDNA1 even if they never officially called it that. Oh and marketing departments suck.
I think I'll buy a Zen 5 laptop this year, then if Zen 6 has 32 core CCX's, I'll do a new desktop build.
Which is a better buy, the 8040 or the cheaper 7040?
Those are literally the same CPU
I bet Microsoft pushed also pushed them to include AI for their Copilot brand
4 performance cores, worthy of being put to use in the next steam deck
There are 12 "Performance" Cores.....
Nice Alpinist on the wrist
Bigger Number Better 😅
Yeah it's gonna cost a lot for a thin laptop this time
My kids diapers are now Ai built, no idea what that means. 😂
So we are talking x86?!
One needs to tell us in todays news ...
So, now we have Ryzen "AI" name scheme for laptops, probably Ryzen AI Z2 for gaming handhelds.
Seriously not the best name, at all...
Let's see what RDNA 3.5 can pull. If there isn't an uplift of 30% then I probably bury the idea of an AMD APU for my desktop PC for the next two years...
It will be around 40% improvement, but we have to see if it’s the same TDP.
@@KenanGwallter At same count of CUs or with more CUs?
@@Quast more CUs and faster RAM, at least that’s what most tech sites tell us. We’ll have to wait and see, I’ve been looking into buying an ultrabook for work and I am waiting for benchmarks on these APUs and the Qualcomm Elite stuff. Whoever gets me more battery life wins.
@@KenanGwallter Feels like we need a far more substantial RAM speeds jump in DDR5 vs DDR4 to take any noticeable effect
@@Quast there are multiple things to take into consideration apart from RAM speed when it comes to APUs
Finally it has AI. That way I can.... Oh right.
Am I the first!?
Zen 5 let's go!
C cores are AMD's answer to arm cores
Ai Ryzen what's up?
XDNA first appeared in ryzen 9/7/5/3 7xxxh/u.
So that's probably gonna be rebranded AI 100.
Then 8000 series would be AI 200.
And new one would be AI 300.
Yea. You should definitely expect these rebrands 😂
Snapdragon released X elite with 40 Tops
AMD released Ryzen AI chips with 50 top.
Intel: Imma end this man's whole career!
Released lunar lake with a total of 120 tops.
11:07
this chart is worth nothing for people like me who are using resolve. As Intels hardware acceleration on arc in resolve is great.
I would assume that AMD is cherry picking.... we´ll see.
I dont get why AMD keeps pressing with mono=designs of CPUs, Intel, Apple, even all cellphones are Big.Little design, and AMD just throws a gimicky AI to a tired platform, and we all applaud...
The newest computer I have is my iPad Pro m1. Every other computer I own is from 2013 and I don’t care.😂
The naming is stupid!
Well I mean, it makes sense because 300 is greater than 8000
AMD giving refreshes a new name is the doing of OEMs. OEMs demand this from the CPU makers because they say that the customer buying their products demand a shiny new number even if it's the same product or they wont buy it. Marketing departments suck. 😋
all the x86 laptops have the same problem, heat, and battery life.
AMDRYZENAI9HX370
congrats AMD you now have serial numbers as a product naming convention. This is absurdly horrendous, copying the homework of another D tier student in product naming (intel).
Hearing it was the OEMs that pressured AMD to put AI in the naming protocol. Hence the last minute change of name.
Lunar Lake is definitely a puzzle. The leaks suggest that it’s all 8C/16T architecture and 17-30W TDPs which would give Team Blue nothing to compete with in the high performance sector - formerly H/HS. Unless they have Apple Silicon rivalling Performance per Watt…🤔
Interesting times for sure. And some great product coming our way. 🤩
Imho I think "AI" branding here in this case is fine because it literally has AI chip in it.
Zen 5 looks like a leap comparable to Zen 2 to 3 to 5. I would like to see more laptop OEMs ship with AMD options.
One point in regards to Apple - Zen 5 is more of a Apple M4 family competitor. The M4 is already on the iPad.
Only $8000 dollars
What's worse than Canada? A tech guy bitching about product names.
I'll wait for Lunar Lake from Intel
Well,in my opinion, the benchmarks were the worrying part of the presentation.They claim that they're beating out the M3 in video editing,whatever this means for AMD ofcourse,it could be playback fps,exporting time,apply effects to the clip,stabilization ecc, with their top of the line 12 cores 24 threads cpu, so 4 cores and 16 threads more then Apple, while probably using 20W more if they are pushing it to the maximum TDP as you said, how can this be a good results to show? If Zen 5 is so much better then competition as they're saying, why not use the 10 cores version to compare with 8 core M3? Or why not compared to the M3 pro 12 cores? We'll see tomorrow what Intel can achieve with Arrow Lake and Lunar Lake before judge,but i'm not that confident in Zen 5 at the moment. I'm not saying that it's bad, just not what i expected (and no, i wasn't one of the people waiting for a 40% ipc improvement over zen 4). Also, speaking about AI since AMD is declaring itself at the front of the race when it comes down to laptops, are they awere of the fact that they're at the 3rd gen of processors with an NPU, but you still can't see it in the fucking task manager, when Intel not only have this option since the release of Meteor Lake,but they already optimized 500 models to run natively on the NPU apart from Copilot,even the just announced Phi 3 from Microsoft, while Lisa said that they're working to bring 150 models to Ryzen AI before the end of 2024...come on, when it comes down to software compatibility Intel is way ahead of them (we need to see where Qualcomm will be in a couple of weeks). Zen 5 on desktop will not have any dedicated NPU compared to Arrow Lake, so all of the light AI models, like studio effects in videocalls for example, will require you to use the CPU, consuming more energy. Now, i know that power efficiency on desktop is a bit less important then laptops and, until now, Amd has been much more efficient then Intel,it's undeniable, but now that they are also on a more competitive node compared to 10nm (on desktop), things could change. My point is, AMD didn't tell us anything about power efficiency for Strix,not a single comparison with Qualcomm or Apple (Intel is clearly behind with Meteor Lake when your pushing all cores) and in 2024 is ESSENTIAL because is what the market wants,even more then absolute performance. They said they are ahead of everybody else when it comes down to consumer AI because they have 5 more TOPs compared to Lunar Lake and Qualcomm, but in reality they have less then 1/3 of the model optimized to run on XDNA compared to Intel NPU, and their desktop cpus won't have any NPU compared to Arrow Lake. I want to wait and see real world performance from both companies, but from what they showed yesterday, they left a big door open for Intel to comeback in my opinion, in both laptops and desktop, also because i read a lot of comments from gamers who are waiting for a discounted 7800X3D or the X3D version of Zen 5 that might come in the same period of Arrow Lake, so the time advantage AMD has on desktop doesn't seems to have got all of the enthusiast gamers,price will be crucial for AMD to succeed. Let's wait and see, but this year the outcome of the battle between Intel and AMD might be different then the last few years.
But oh yes you where indeed one of the 40 % IPC MLID people 😂😂😂 that's all your camouflaged rant boils down to 😅😂
@@lllongreen IPC is not the problem, again, i don't feel confident in zen 5 on mobile at the moment because they showed benchmarks compared to competition, but didn't share the power needed to achieve such numbers, the only thing they said is that's more efficient then Zen 4. Look at Qualcomm for example,they said "we achieve this numbers on a 23W TDP platform and this on an 80W TDP platform", while AMD doesn't. And again,when compared to Apple M3, why against the base 8 cores and not the M3 pro 12 cores since Strix also has 12 cores? That's what i didn't like about the presentation, since we're talking about laptops and even AMD in their presentation has always talked about their great efficiency. Other things i pointed out are just facts, Granite Ridge won't have an NPU, and their software optimization is worse then Intel for AI model in the consumer space. For gamers also, nothing new and related to IPC directly, they knows that, even if the IPC would've been 40% for Zen5, the X3D version give you much more fps anyway and that's the cpu to go for maximum performance, there's a reason why 7800 X3D sold so well. But considering the average fps gains from Zen 4 to Zen 5 they don't feel the urge to upgrade to a non X3D version, even from the previous gen, so they will probably wait. I don't see a rant about IPC in my comment, i just look over the hype and the non sense marketing, that's why i want to wait and see reviews and deep analysis before judge.
Just say ryzen A9 don't even bother saying AI in the name
I prefer the other two guys on Hardware Canucks. I have a hard time watching his videos. No offense to him.
Man I hate this naming so fucking much and I hate the whole Copliot bs aswell honestly I'm eventually just going to switch to Linux due to Microsoft pushing AI up our asses
Happily on Linux since 2004. Just do it!
People have been saying that for ages and yet the mass exodus hasn't happened. There's a reason that people don't use Linux.
26 with 1 line on ur forehead😂 hahah
Let Me say Two Things :
1. We haven't Asked fro Good Build quality Laptop to be Thin as the IPad M3
2. Please make your crappy Motherboard Baked Software available on Linux. 😡😡😡🤬
I hate it when I'm gaming and when someone throws a smoke and two people are on the screen my fps goes from 240~230 ish to like 145 and below making everything strttttueer.
i hate when youtube randomly messages me adblocker detected just like a fking nerd. where's my privacy
Intel Core Ultra 200 is more exciting product, can you make the comparison when these 2 laptops available.
would like to see the performance per watt comparison.
Won't be looking at anything AMD until they reveal details on RDNA 4 for desktop and laptop. RDNA 3.5? -________-
rdna 4 is literally update on rdna 3 so better wait for rdna 5
@@itsmilan4069 good to know! I was mistaken and though RDNA 4 was going to be the big update. Thanks!
Ai 300 is spartan 300, they wont give up talking about Ai (Appeasing investor) until 1 crippled dude betray em
.
Crazy how ahead apple is. Comparing base m3 to top of the line AI 9HX 370, makes no sense tbh.
*gets excited of hardware canucks new video
*this guy is the host: 🚶🚶
bro has a canal in his forehead
lol
12:42 They all can' t compete with Apple Silicon.
So whiny
the first person comment
If someone buys AMD GPUs for AI they are such a lost cause lmao. All it can do is run online image generators and zoom background stuff. That's it. My i7 9750H has been doing it for years now.
I understand you. But the point of NPU is to offload such workload to dedicated core. Right now AI is nothing more than marketing jargon.
Not sure from where gpu came from. Video is about laptop cpu/npu. 🥱
M4: hold my beer
.... the video is just talk about naming that is so stupid... thumb down
Interesting new hardware such a shame that microsoft is such a dumpster fire.
I can see people abandoning microsoft OS, and if they choose to run AI likely just locally on linux with ollama..
The ai push is well,.
I can't say I see the appeal.
Let’s be honest, nobody cares about x86 in laptops apart from mobile gamers with lots of cash.
Except all of corporate America. 🤣
Am i the only person who thinks AI is completely boring. I understand its usefulness. But I just don't care. At all.
Intel fanboy channel
can't accept anything good
It’s in new zephyrus. Sadly way weaker than intel ultras
MEH... Apple silicon is far better than this crap
yes better with 25% native app support 37% on translation and rest either crash on don't run to begin with 😂
@@itsmilan4069 hey atleast my os doesnt come with bloatware.. good luck with windows lmao
Yes, 54W Ryzen "AI" VS only M3 20W TDP💀 And also when it finally comes out, it will face against M4 already
@@itsmilan4069 Windows sheep still using Internet Explorer💀 Meanwhile 1% of Windows apps are native ARM
@@AzVfLpower consumption is determined by oem could be 15-54w huh
As I understand AI 3xx mean the 3rd generation of AI CPU. 😂
While Intel AI 2xx mean 2rd generations AI CPU of Intel.
I think this has a lot to do with Intel claiming they released the "World first AI gaming cpu" with Core Ultra.I think OEM partner want them to name that 😂
I understand that but the number why? Please just year-architecture-coreCount
The only thing I wonder all the time - why the f I need this STUPID AI BS in my life doing literally nothing to any meaningful workflow.
Because marketing