Lol, are you for real? Listen, it's obvious you're a fan of the fruit, but I use both for my job and your statement just isn't true. The M series lose their luster when you throw anything that isn't apple at it. Citrix and MS Teams? How does 7 hrs of battery life sound. The Yoga 7i and Galaxy Book3 Pro both last longer than that with an identical workload.
Apple has its niche and they make some of the nicest casual computing / final cut pro machines... but they aren't the best work machine, let alone windows machine for enterprise applications.
His statement clearly showed that he didn't run any app inside windows for arm on parallels:-)
However, your statement is absolutely true. I have an almost max specced 15inch M2 max, but for general business there are no apps available.
You can't even do presentations, because of the missing Miracast Support. Miracast won't even work inside an windows VM, because it uses wifi-direct to the TV.
Otherwise it's a really well build laptop and i love it in every aspect. Just the usecase is limited. But im an it administrator, that needs anyway almost only the terminal:-) So im Okay at least:-)
The best Windows laptop is probably dependent on your needs, but it ain't that, chief.
I mean, look at the OpenData tests for Blender. An RTX 3050 laptop is taking out the $4k variant of the Mac Studio in rendering. Different machines for different needs.
Based on the power envelope, it seems we're comparing the wrong things.
The M2 should be compared to the Base Z1 (8W), M2 Pro (10c/16g) against the Z1 Extreme (15W), and the full M2 Pro (12c/19g) against the 7840U (25W). It's trickier comparison with the 40W M2 Max (12c/30g) versus the AMD 7840H (35W). Or the 50W full M2 Max (12c/38g) versus the AMD 7940HS (55W).
So that would make it less interesting, since I expect macOS and ARM to have a sizeable advantage. But just because Windows and x86 are 1-2 generations (or more) behind is NOT the entire reason to switch platforms. It really boils down to the user and their needs, specifically.
A more interesting direction is to consider why this claim might be true (in terms of HW, if you are not utterly locked to x86). In terms of cores, Apple has 4 P at the low-end, 6 midrange (Pro) 8 at the high end (Pro and Max). E cores basically catch up the extra performance from hyperthreading. Apple frequencies are lower, but IPC is higher, so single core is essentially a wash; but Apple can sustain that IPC across all the cores whereas AMD will have to dial back the frequency when all cores are working.
In terms of GPU, more or less nV, AMD, and Apple all have the same sort of performance in what they call a core, or SM, or CU. At the low end M2 Apple gives you 8 or 10 of these (M2) at the midrange 16 or 19 of these (Pro). So at comparable points Apple is ahead in GPU. Likewise at comparable points Apple gives you more memory bandwidth, both for CPU (nice) and GPU (essential).
Finally Apple has a very nice (probably unmatched) media engine including really good H.265 encoding and a low power inference engine (which sounds good but remains mostly unused still, on both the Apple side and even more so the x86 side). And other stuff like the SSD engine and the Secure Enclave (which doesn't seem to hit frequent problems the way these things all seem to on the x86 side – maybe that's unfair to AMD, I don't know the details of which TMP has which problems).
It's not crazy, if you can live in a Windows ARM world, to want to do so on a Mac, in the same way that plenty of people used an x86 Mac as their default Windows or Linux machine.
Bottom line is that Windows lost its dominance years ago, unless you insist on changing the market from "computing" to something a lot thinner like "desktop computing on x86". It's fine to still love Windows x86 if you want; it's less fine to remain deluded as to its true place in the full story of 2023 computing.
Yeah, from your very own graph, it says Android (phones) dominate browser share. Followed by Windows. And in a distant 3rd place you have iOS, while OSX is 4th.
So the dominant market share for handheld device is Android And the dominant market share for PC (desktop/laptop) is Windows
Yeah, no one cares so long as the cooling and battery life is good. These are not difficult problems to solve. Use a vapor chamber, and big 99Whr battery.
Not sure where you've been bud. M1 caught everyone with their pants down when it launched yes, but it's been surpassed on single core, multicore, and even IGP.
These parts will definitely beat M1 on performance, though its remaining magic is being able to perform well fanless. Someone could probably make a viable competitor out of these AMD parts tuned down a bit.
M1 has already reached its performance potential of design. So no further signify architectural gain can be expected other than adding more cores and shrinking the die. Most of its advantage comes from its integrated ram.but also one of its biggest weakness in terms of pushing clock speeds. In my opinion it's still a better trade off for laptops. And most non scientific workloads. By skipping a select tiny engineering and scientific consumer they managed to give the rest a far superior architecture
M1. yes, but only if you work, not play on it. m2 sucks and overheats, which is not worth extra performance. m3, we don't know anything, might be another 50'C palm rest disaster. What I can tell you, open a game, like sims4 or minecraft. Good luck Phoenix is perfect laptop for a kid, for school and light games. mac's are pure work/web laptops, that literally cannot game at all.
You have to understand one thing arm can never beat a x86 I both performance and energy if architected correctly. It is the fundamental difference between the simple instruction vs complex instruction cpu designs. Apples m1 just took a temporary leap taking advantage of process nodes and using the simplicity of arm cpu design to get to ram on chip, that's it. Also Intel being run by non technical people refusing to move forward with their architectural progression. Eventually both Intel and amd will get to the architecture which arm simply cannot compete because of the fundamental difference, yes it's much more complex for x86 so it's slower to get tger
Is that 9Tflops on the 780M IGP then? Wow, that's well beyond the Meteor Lake Tile GPUs even before they launch then on paper, I would love to see Anandtech in depth tests of both.
Yeah, I think we will end up seeing around the same performance from 128Xe and 12 RDNA3. They are both going to be memory bandwidth starved, since they will be using DDR5 4800~5600 at 128bit bus width. Meanwhile, GPUs use the same 128bit memory bus width, but with GDDR6x, which is roughly 4x faster than DDR5.
Quote : " Interestingly in the gaming performance figures, which it put up its Ryzen 7 7840U with Radeon 780M integrated graphics again Intel's Core i7-1360P with Iris Xe graphics, it's not too surprising to see that the Radeon 780M performed between 130% and 239% better than Intel in specific titles."
Meh - 1360P has 96EU, max theortical TFLOPS is ~2 .. the 780M has peak theoretical f32 of >8 TFLOPS .. so it isn;t performing well , relatively.. especially when consider that the 1360P and nominal 2TFLOP APU from AMD go head to head in performance on a range of games..
Possibly it's memory bandwidth starved.. Will wait for actual 3rdparty tests though..
...
Also they didn't even try on TDP - SoC with 2x CPU cores and 3x GPU cores is given exact same TDP... come one AMD
The GPU on both are most likely memory bandwidth starved. DDR5 of any speed will simply not perform as well as GDDR6 in terms of bandwidth. And I don't think AMD or Intel are willing to upgrade from a 128bit memory bus width to 192bit or 256bit, unless it's a workstation chip.
it's also since been pointed out to me that rdna3 probably isn't taking advantage of the new dual issue instructions (or can't) .. so the effective tflopd on the 780m is more like 4-5 tflops not 8+ .... this matches the benches listed in the article so...
also judging from some other reporting on the and z1 / z1 extreme (z1 extreme appears to be same chip) it looks like the benches amd supplied might be at a limited total tdp, nerfing the z1 extreme even more even if it had bandwidth..
it kind of looks like the 12cu part might be wasted in a 30w total power scenario (inferred from performance metrics given by Asus ROG for there new handheld)
It's probably time to stop paying attention to TFLOP numbers altogether. The high TFLOPS here are a quirk of how RDNA3 works.
About TDP, I think AMD just serves up what a predictable laptop market wants. Most laptops are intended to hit around a ~15W target, or ~28W. I'm sure if the laptop maker wants to get to fanless sub-10W with the quad-core, they could do it.
On the slide with "More Graphics power than ever before", I see that they have "Ryzen 7 and 9" listed. Is that a mistake? or are we expecting a Ryzen 7940U?
The practical state of Stable Diffusion is sad. It could be compiled with TVM, MLIR, AITemplate, Triton, any number of frameworks or bsckends. It could be widely cross device...
But no, the entire community is stuck on the A1111 repo running an old, abandoned Stability AI implementation in CUDA PyTorch eager mode with a complicated install.
No one is going to get SD running on XDNA beyond a simplistic demo, because apparently no one cares :/
would be nice to get a 4C/8T part with a 780M or higher CU iGPU for some gaming power ... but AMD will not do it for some reason. I'd kill for a 7640 with an 780M or something with 16CUs for a minipc/console/steamOS gaming machine ... but no, can't have that ....
With proper segmentation it would not. Generally the users are biased between CPU or GPU with balanced being actually the least wanted option. If you are buying a CPU powerhouse, you WILL pair it with a dedicated GPU or you only need a light GPU. On the other hand cheap does only need a low end CPU and GPU just for web/office. However good GPU and medium range CPU is nowhere to be seen and that drives a lot of buyers away. They either need to overpay on the CPU or look for a standalone GPU.
It would only eat into console sales but those are driven by other factors (game exclusives etc.).
Intel offered them Broadwell eDRAM (which they made for Apple), and PC OEMs passed.
AMD offered them Van Gogh, and PC OEMs passed... Until Valve took it up.
AMD *had* the very thing you are describing on their roadmap, and they seemingly canceled it. Why wouldn't they, if no one is going to sell laptops with it?
Disabling half the CPU cores doesn't make sense in the first place. The yields are too high. You should wish for cheap, plentiful 6-cores. Which at least got an improvement from last gen, 6 to 8 CUs.
Strix Point should have 16 CUs, and then Strix Halo could be a real winner, if you can pay the price.
This would make a great comp with Apple chips. After all both of them are made with TSMC N5/4 process. Biggest problem with AMD laptop chips is they are niche. you only see older 5 series laptops at local stores. I never saw a single Rembrandt laptop last year in any store. But you see tons of Alder Lake laptops in store.
AMD's physical hardware is good, their firmware and drivers not so much. I repair electronics for a living and AMD powered laptops represent, in my small world, an outsized percentage of glitchy notebooks brought in for repair. The majority of the time its driver incompatibility issues.
I do have 2 AMD laptop. Both gaming laptop with Nvidia dGPU. 1 with mux switch having no issue. 1 that not have mux switch having issue that windows keep update to old driver that some game will stop working because driver to old, or get micro stutter. No DDU will not solving this problem. Need rollback driver update from device manager, twice, after windows update.
Weird, given the length of this article and the lack of hardware or benchmarks, that this isn't a pipeline story. It very much fits what one would expect from a single page announcement blurb with a bit of hands-off observation.
A large secondary mobile market upgrade wave is coming in from Kaby Refresh and Whiskey Lake seen on trade-in trend returning to secondary market. AMD, Intel, Nvidia are all positioned for it which means Apple must be positioned. AMD and Intel channel still have back through Cezanne and Tiger mobile surplus to clear and it does stretch back into Comet not all mobile upgrade will go for the entire generational leap but purchases of primary production appear promising for those who can justify the price performance of a fashion statement and for utility buyers surplus has to be cleared on price drops.
Specific Apple mobile here is the WW channel by quarter volume wave all-inclusive M1, Max, M2;
I'll add on the laptop mobile market situation into 2023, its apparent from the data, x86 both AMD and Intel price performance graded surplus stacks consisting of Rembrandt, Cezanne and back to Renoir and Intel Alder and Tiger, are positioned to be dumped on Apple M_ refresh into to the year. The tactic is one, to mire Apple in a flood of clearance sale x86 laptops and two for AMD to flush under Intel riding within that flow of Intel clearance sale. This is a tired and true x86 duopoly tactic to beach Apple in a wash of x86 laptops and the only way to avoid being dumped on is to ride the x86 wave and to do that requires superior price performance and Apple also has its installed base to count on 'niche segment wise' as fairly protected sales channel. mb
I was waiting on the Lenovo Z16 with a 7040HS chip--I think they prematurely showed it on their website back in February and then it disappeared. Ended up grabbing an Alder Lake LG Gram 16 instead but it's missing a usable GPU.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
71 Comments
Back to Article
lemurbutton - Wednesday, May 3, 2023 - link
Won't hold a candle to M1, let alone M2 and the upcoming M3.supdawgwtfd - Wednesday, May 3, 2023 - link
Get back to me when you can run Windows on that hardware well.Until then, Mac will continue to be niche and the world will run in x86
lemurbutton - Wednesday, May 3, 2023 - link
The best Windows laptop is literally an M2 Max Macbook Pro running Parallels.JustAnotherITGuy - Wednesday, May 3, 2023 - link
Lol, are you for real? Listen, it's obvious you're a fan of the fruit, but I use both for my job and your statement just isn't true. The M series lose their luster when you throw anything that isn't apple at it. Citrix and MS Teams? How does 7 hrs of battery life sound. The Yoga 7i and Galaxy Book3 Pro both last longer than that with an identical workload.Apple has its niche and they make some of the nicest casual computing / final cut pro machines... but they aren't the best work machine, let alone windows machine for enterprise applications.
Ramalama - Saturday, July 22, 2023 - link
His statement clearly showed that he didn't run any app inside windows for arm on parallels:-)However, your statement is absolutely true. I have an almost max specced 15inch M2 max, but for general business there are no apps available.
You can't even do presentations, because of the missing Miracast Support.
Miracast won't even work inside an windows VM, because it uses wifi-direct to the TV.
Otherwise it's a really well build laptop and i love it in every aspect. Just the usecase is limited.
But im an it administrator, that needs anyway almost only the terminal:-)
So im Okay at least:-)
mukiex - Wednesday, May 3, 2023 - link
The best Windows ARM laptop is an M2 Macbook Pro.The best Windows laptop is probably dependent on your needs, but it ain't that, chief.
I mean, look at the OpenData tests for Blender. An RTX 3050 laptop is taking out the $4k variant of the Mac Studio in rendering. Different machines for different needs.
evolucion8 - Thursday, May 4, 2023 - link
Sorry to burst your bubble, but even the 6900HX gave the M2 a run for its money in several benchmarks and the power consumption gap wasn't that wide.Kangal - Saturday, May 6, 2023 - link
Based on the power envelope, it seems we're comparing the wrong things.The M2 should be compared to the Base Z1 (8W), M2 Pro (10c/16g) against the Z1 Extreme (15W), and the full M2 Pro (12c/19g) against the 7840U (25W). It's trickier comparison with the 40W M2 Max (12c/30g) versus the AMD 7840H (35W). Or the 50W full M2 Max (12c/38g) versus the AMD 7940HS (55W).
So that would make it less interesting, since I expect macOS and ARM to have a sizeable advantage. But just because Windows and x86 are 1-2 generations (or more) behind is NOT the entire reason to switch platforms. It really boils down to the user and their needs, specifically.
usiname - Wednesday, May 10, 2023 - link
To much bulls.. on one place, even M2 non pro has power consumption of 21w let alone m2 pro (12c)timecop1818 - Thursday, May 4, 2023 - link
For the price of that you can buy an x86 laptop that will absolutely DESTROY anything from crapplename99 - Wednesday, May 3, 2023 - link
A more interesting direction is to consider why this claim might be true (in terms of HW, if you are not utterly locked to x86).In terms of cores, Apple has 4 P at the low-end, 6 midrange (Pro) 8 at the high end (Pro and Max). E cores basically catch up the extra performance from hyperthreading. Apple frequencies are lower, but IPC is higher, so single core is essentially a wash; but Apple can sustain that IPC across all the cores whereas AMD will have to dial back the frequency when all cores are working.
In terms of GPU, more or less nV, AMD, and Apple all have the same sort of performance in what they call a core, or SM, or CU. At the low end M2 Apple gives you 8 or 10 of these (M2) at the midrange 16 or 19 of these (Pro). So at comparable points Apple is ahead in GPU. Likewise at comparable points Apple gives you more memory bandwidth, both for CPU (nice) and GPU (essential).
Finally Apple has a very nice (probably unmatched) media engine including really good H.265 encoding and a low power inference engine (which sounds good but remains mostly unused still, on both the Apple side and even more so the x86 side).
And other stuff like the SSD engine and the Secure Enclave (which doesn't seem to hit frequent problems the way these things all seem to on the x86 side – maybe that's unfair to AMD, I don't know the details of which TMP has which problems).
It's not crazy, if you can live in a Windows ARM world, to want to do so on a Mac, in the same way that plenty of people used an x86 Mac as their default Windows or Linux machine.
name99 - Wednesday, May 3, 2023 - link
You can see the latest OS share trends here:https://daringfireball.net/2023/05/browser_and_os_...
Bottom line is that Windows lost its dominance years ago, unless you insist on changing the market from "computing" to something a lot thinner like "desktop computing on x86".
It's fine to still love Windows x86 if you want; it's less fine to remain deluded as to its true place in the full story of 2023 computing.
meacupla - Thursday, May 4, 2023 - link
Yeah, from your very own graph, it says Android (phones) dominate browser share.Followed by Windows.
And in a distant 3rd place you have iOS, while OSX is 4th.
So the dominant market share for handheld device is Android
And the dominant market share for PC (desktop/laptop) is Windows
And then, oh, what's this? Apple halted production of M2 chips? https://www.macrumors.com/2023/04/03/apple-stopped...
Gee, I thought they had a solid market share for PCs, when Windows was shrinking.
Blastdoor - Saturday, May 6, 2023 - link
Apple’s niche is the top (in terms of income) 20% of the market.ET - Wednesday, May 3, 2023 - link
A well thought out response, which I assume takes into account the AMD slide that shows the 7840U beating the M2.lemurbutton - Wednesday, May 3, 2023 - link
There is no way it can be an M1, let alone M2. Don't fall for cherry-picked benchmarks.JustAnotherITGuy - Wednesday, May 3, 2023 - link
This post paid for by Apple. Inc.tipoo - Wednesday, May 3, 2023 - link
Like Apple sort of forgetting the 3080 performance per watt chart just kept going, instead of ending it where their M1 Max was lolThey make impressive parts for the low power they use but it's not a knockout like M1 was at launch
bejito81 - Wednesday, May 3, 2023 - link
you're mistaking the M2 for the M2 pro or M2 max, the M2 itself isn't that powerful and as passmark the 7840u will be way more powerfulas for the iGPU part, M2 is crap and will get easily destroyed in any gaming scenario by the 7840u
we all saw you're an apple fanboy, time to open your eyes
Qasar - Wednesday, May 3, 2023 - link
lemurbutton,here is a thought... instead of posting your i love apple bs. post links to show your claims to be true. until then. all you post is bs
nandnandnand - Thursday, May 4, 2023 - link
Translation: Only fall for Apple's cherry-picked benchmarks, ignore any process node advantage Apple might have over competitors.ikjadoon - Wednesday, May 3, 2023 - link
At much higher TDP. All the 7040U CPUs are base 28W TDP.meacupla - Wednesday, May 3, 2023 - link
Yeah, no one cares so long as the cooling and battery life is good.These are not difficult problems to solve. Use a vapor chamber, and big 99Whr battery.
tipoo - Wednesday, May 3, 2023 - link
Not sure where you've been bud. M1 caught everyone with their pants down when it launched yes, but it's been surpassed on single core, multicore, and even IGP.These parts will definitely beat M1 on performance, though its remaining magic is being able to perform well fanless. Someone could probably make a viable competitor out of these AMD parts tuned down a bit.
mukiex - Wednesday, May 3, 2023 - link
This will likely make for an interesting benchmark cycle on Youtube in the next few months, though.Samus - Thursday, May 4, 2023 - link
lemur, if the M1 is so great, why has Apple not gained any market share since it launched almost 3 years ago?https://www.cpubenchmark.net/compare/4104vs5266/Ap...
sharath.naik - Thursday, May 4, 2023 - link
M1 has already reached its performance potential of design. So no further signify architectural gain can be expected other than adding more cores and shrinking the die. Most of its advantage comes from its integrated ram.but also one of its biggest weakness in terms of pushing clock speeds. In my opinion it's still a better trade off for laptops. And most non scientific workloads. By skipping a select tiny engineering and scientific consumer they managed to give the rest a far superior architecturedeil - Thursday, May 4, 2023 - link
M1. yes, but only if you work, not play on it. m2 sucks and overheats, which is not worth extra performance. m3, we don't know anything, might be another 50'C palm rest disaster.What I can tell you, open a game, like sims4 or minecraft.
Good luck
Phoenix is perfect laptop for a kid, for school and light games.
mac's are pure work/web laptops, that literally cannot game at all.
sharath.naik - Wednesday, May 24, 2023 - link
You have to understand one thing arm can never beat a x86 I both performance and energy if architected correctly. It is the fundamental difference between the simple instruction vs complex instruction cpu designs. Apples m1 just took a temporary leap taking advantage of process nodes and using the simplicity of arm cpu design to get to ram on chip, that's it.Also Intel being run by non technical people refusing to move forward with their architectural progression. Eventually both Intel and amd will get to the architecture which arm simply cannot compete because of the fundamental difference, yes it's much more complex for x86 so it's slower to get tger
isthisavailable - Wednesday, May 3, 2023 - link
I wish we could get a proper Anandtech review of Ryzen 7 7840U vs M2 vs i7 1360PA5 - Wednesday, May 3, 2023 - link
I think there's only like 2 people that work here now. Takes all their time to repost press releases.tipoo - Wednesday, May 3, 2023 - link
Is that 9Tflops on the 780M IGP then? Wow, that's well beyond the Meteor Lake Tile GPUs even before they launch then on paper, I would love to see Anandtech in depth tests of both.techjunkie123 - Wednesday, May 3, 2023 - link
sadly that doubling of theoretical TFLOPS from RDNA2 to RDNA3 barely gives a performance increase.....it's mostly on paper.tipoo - Wednesday, May 3, 2023 - link
Ah I see. I used to be up with things...So these should be around the ~4Tflops the 128 Xe core Tile GPU for Meteor Lake is providing then?
meacupla - Wednesday, May 3, 2023 - link
Yeah, I think we will end up seeing around the same performance from 128Xe and 12 RDNA3.They are both going to be memory bandwidth starved, since they will be using DDR5 4800~5600 at 128bit bus width.
Meanwhile, GPUs use the same 128bit memory bus width, but with GDDR6x, which is roughly 4x faster than DDR5.
tipoo - Friday, May 5, 2023 - link
That's where the Adamantine L4 cache on MTL will be interesting, if they keep it better fed they have a chance at jumping past the 780M I thinkJasonMZW20 - Wednesday, May 3, 2023 - link
“FP16” which is at double rate.2.7GHz is good for 4.147 TFLOPs FP32, or if dual-issue FP32, 6.220 TFLOPs if at least 50% FP32 can be dual-issued.
The numbers listed in the slides are for 2.8GHz in the HS parts.
tipoo - Wednesday, May 3, 2023 - link
Ah ok. So it should be in the ballpark of the 128 Xe core Tile GPU on Meteor Lake, drivers depending.Be very interested to see a deep dive comparison on Anandtech, if pray we still get those
CyrIng - Wednesday, May 3, 2023 - link
Ryzen 3 7440U is not mentioned yet on AMD, worldwide.zamroni - Wednesday, May 3, 2023 - link
You can buy 7040hs laptop now.I see asus g14 7940hs are available in best buy
jamesindevon - Wednesday, May 3, 2023 - link
Who thought Phoenix was a good name for an AMD laptop processor? (Now the 7000X3D series do seem to be dying "in a show of flames and combustion"...)xol - Wednesday, May 3, 2023 - link
Quote : " Interestingly in the gaming performance figures, which it put up its Ryzen 7 7840U with Radeon 780M integrated graphics again Intel's Core i7-1360P with Iris Xe graphics, it's not too surprising to see that the Radeon 780M performed between 130% and 239% better than Intel in specific titles."Meh - 1360P has 96EU, max theortical TFLOPS is ~2 .. the 780M has peak theoretical f32 of >8 TFLOPS .. so it isn;t performing well , relatively.. especially when consider that the 1360P and nominal 2TFLOP APU from AMD go head to head in performance on a range of games..
Possibly it's memory bandwidth starved.. Will wait for actual 3rdparty tests though..
...
Also they didn't even try on TDP - SoC with 2x CPU cores and 3x GPU cores is given exact same TDP... come one AMD
meacupla - Wednesday, May 3, 2023 - link
The GPU on both are most likely memory bandwidth starved.DDR5 of any speed will simply not perform as well as GDDR6 in terms of bandwidth.
And I don't think AMD or Intel are willing to upgrade from a 128bit memory bus width to 192bit or 256bit, unless it's a workstation chip.
xol - Wednesday, May 3, 2023 - link
it's also since been pointed out to me that rdna3 probably isn't taking advantage of the new dual issue instructions (or can't) .. so the effective tflopd on the 780m is more like 4-5 tflops not 8+ .... this matches the benches listed in the article so...tipoo - Sunday, May 7, 2023 - link
Is this a software or hardware limit?FineWine?
lmcd - Tuesday, May 9, 2023 - link
Software, but it's a permanent software limit barring console scenarios (where designing for exact architecture is a thing).nandnandnand - Thursday, May 4, 2023 - link
Strix Halo rumored to go to 256-bit. I bet it will be expensive.meacupla - Thursday, May 4, 2023 - link
now that looks impressive.xol - Thursday, May 4, 2023 - link
also judging from some other reporting on the and z1 / z1 extreme (z1 extreme appears to be same chip) it looks like the benches amd supplied might be at a limited total tdp, nerfing the z1 extreme even more even if it had bandwidth..it kind of looks like the 12cu part might be wasted in a 30w total power scenario (inferred from performance metrics given by Asus ROG for there new handheld)
xol - Friday, May 5, 2023 - link
[edit] getting different things confused ..ignore
nandnandnand - Thursday, May 4, 2023 - link
It's probably time to stop paying attention to TFLOP numbers altogether. The high TFLOPS here are a quirk of how RDNA3 works.About TDP, I think AMD just serves up what a predictable laptop market wants. Most laptops are intended to hit around a ~15W target, or ~28W. I'm sure if the laptop maker wants to get to fanless sub-10W with the quad-core, they could do it.
meacupla - Wednesday, May 3, 2023 - link
On the slide with "More Graphics power than ever before", I see that they have "Ryzen 7 and 9" listed.Is that a mistake? or are we expecting a Ryzen 7940U?
nandnandnand - Thursday, May 4, 2023 - link
Let's see the XDNA engine running Stable Diffusion.brucethemoose - Friday, May 5, 2023 - link
The practical state of Stable Diffusion is sad. It could be compiled with TVM, MLIR, AITemplate, Triton, any number of frameworks or bsckends. It could be widely cross device...But no, the entire community is stuck on the A1111 repo running an old, abandoned Stability AI implementation in CUDA PyTorch eager mode with a complicated install.
No one is going to get SD running on XDNA beyond a simplistic demo, because apparently no one cares :/
Oxford Guy - Saturday, May 6, 2023 - link
It’s horrible anyway for doing people other than portraits. Tries to be too many things and suffers from low-quality training images.It’s absurd that it can’t even manage to create rabbits with tails and two ears.
haplo602 - Thursday, May 4, 2023 - link
would be nice to get a 4C/8T part with a 780M or higher CU iGPU for some gaming power ... but AMD will not do it for some reason. I'd kill for a 7640 with an 780M or something with 16CUs for a minipc/console/steamOS gaming machine ... but no, can't have that ....meacupla - Thursday, May 4, 2023 - link
That would be too sensible, and the overlap would eat into sales of more expensive SKUs.haplo602 - Thursday, May 4, 2023 - link
With proper segmentation it would not. Generally the users are biased between CPU or GPU with balanced being actually the least wanted option. If you are buying a CPU powerhouse, you WILL pair it with a dedicated GPU or you only need a light GPU. On the other hand cheap does only need a low end CPU and GPU just for web/office. However good GPU and medium range CPU is nowhere to be seen and that drives a lot of buyers away. They either need to overpay on the CPU or look for a standalone GPU.It would only eat into console sales but those are driven by other factors (game exclusives etc.).
brucethemoose - Friday, May 5, 2023 - link
No, I think OEMs just don't care for that.Intel offered them Broadwell eDRAM (which they made for Apple), and PC OEMs passed.
AMD offered them Van Gogh, and PC OEMs passed... Until Valve took it up.
AMD *had* the very thing you are describing on their roadmap, and they seemingly canceled it. Why wouldn't they, if no one is going to sell laptops with it?
nandnandnand - Saturday, May 6, 2023 - link
Disabling half the CPU cores doesn't make sense in the first place. The yields are too high. You should wish for cheap, plentiful 6-cores. Which at least got an improvement from last gen, 6 to 8 CUs.Strix Point should have 16 CUs, and then Strix Halo could be a real winner, if you can pay the price.
trivik12 - Thursday, May 4, 2023 - link
This would make a great comp with Apple chips. After all both of them are made with TSMC N5/4 process. Biggest problem with AMD laptop chips is they are niche. you only see older 5 series laptops at local stores. I never saw a single Rembrandt laptop last year in any store. But you see tons of Alder Lake laptops in store.WhatchaMean - Thursday, May 4, 2023 - link
AMD's physical hardware is good, their firmware and drivers not so much. I repair electronics for a living and AMD powered laptops represent, in my small world, an outsized percentage of glitchy notebooks brought in for repair. The majority of the time its driver incompatibility issues.m00bie - Sunday, May 14, 2023 - link
I do have 2 AMD laptop. Both gaming laptop with Nvidia dGPU. 1 with mux switch having no issue. 1 that not have mux switch having issue that windows keep update to old driver that some game will stop working because driver to old, or get micro stutter. No DDU will not solving this problem. Need rollback driver update from device manager, twice, after windows update.PeachNCream - Thursday, May 4, 2023 - link
Weird, given the length of this article and the lack of hardware or benchmarks, that this isn't a pipeline story. It very much fits what one would expect from a single page announcement blurb with a bit of hands-off observation.Bruzzone - Friday, May 5, 2023 - link
A large secondary mobile market upgrade wave is coming in from Kaby Refresh and Whiskey Lake seen on trade-in trend returning to secondary market. AMD, Intel, Nvidia are all positioned for it which means Apple must be positioned. AMD and Intel channel still have back through Cezanne and Tiger mobile surplus to clear and it does stretch back into Comet not all mobile upgrade will go for the entire generational leap but purchases of primary production appear promising for those who can justify the price performance of a fashion statement and for utility buyers surplus has to be cleared on price drops.Specific Apple mobile here is the WW channel by quarter volume wave all-inclusive M1, Max, M2;
q4 2020 = base
q1 2021 = 5.8x
q2 = 2.29
q3 = 0.98 trough
q4 = 1.61
q1 2022 = 1.41
q2 = 1.14
q3 = 1.13 flattish
q4 = 1.16
q1 23 = 0.87 trough Apple appears prepared for refresh
Apple nine quarter compound average volume = 1.65
Analyst suspects Nvidia mobile dGPU volume to surpass desktop in 2023.
Mike Bruzzone, Camp Marketing
Bruzzone - Saturday, May 6, 2023 - link
I'll add on the laptop mobile market situation into 2023, its apparent from the data, x86 both AMD and Intel price performance graded surplus stacks consisting of Rembrandt, Cezanne and back to Renoir and Intel Alder and Tiger, are positioned to be dumped on Apple M_ refresh into to the year. The tactic is one, to mire Apple in a flood of clearance sale x86 laptops and two for AMD to flush under Intel riding within that flow of Intel clearance sale. This is a tired and true x86 duopoly tactic to beach Apple in a wash of x86 laptops and the only way to avoid being dumped on is to ride the x86 wave and to do that requires superior price performance and Apple also has its installed base to count on 'niche segment wise' as fairly protected sales channel. mbOxford Guy - Saturday, May 6, 2023 - link
Single-channel RAM?Arnulf - Sunday, May 7, 2023 - link
"to to mix" - one "to" to many (near the top)Ryan Smith - Monday, May 8, 2023 - link
Thanks!StevenMo - Wednesday, May 10, 2023 - link
Michael Fischetti is a great Personal Injury Attorney. Call him: 833-645-3247UNCjigga - Wednesday, May 10, 2023 - link
I was waiting on the Lenovo Z16 with a 7040HS chip--I think they prematurely showed it on their website back in February and then it disappeared. Ended up grabbing an Alder Lake LG Gram 16 instead but it's missing a usable GPU.