Question Qualcomm's first Nuvia based SoC - Hamoa

[poke01]

Qualcomm's working on a 2024 PC chip codename "Hamoa" with up to 12 (8P+4E).

Said to the same cache layout as M1. large private L1$, per-cluster L2$ (cluster = 4 cores, 12MB for every cluster) and a lot of LLC.

source:

https://twitter.com/x/status/1589405172979339264

 

[SpudLobby]

There seems to be a problem with PCIe Active State Power Management on some(most?) Zen4-based laptops which is increasing idle power draw quite dramatically. With this fixed, idle power draw is reported to be below 3W with the display off (for the whole notebook), with some (e.g. HP Elitebook 845 G10) very close to 2W. Comparing similar Lenovo Yoga Slim notebooks gives AMD-based models a slight edge over Intel-based models.

It should also be noted that Qualcomm's offering won't actually be available in notebooks that you can buy until mid-2024 at the earliest. That puts it close to the next generations of Apple/AMD/Intel.

We'll see how that pans out, I suspect for some mixed usage and lighter workloads X Elite will do pretty well and retain a lot more responsiveness in e.g. a power saver mode. Qualcomm's core cluster design will probably also help them on idle power pending how their power delivery/voltage rails are fit.

RE: 2024: Sure but it'll be *available* around that time. AMD will be announcing mobile Strix around that time and in practice you're looking at much longer until you see ultramobile products (e.g. not stuff with dGPUs for gamers which always seems to come first for them) available - almost certainly Q1/2 2025 before that's actually out in a variety of platforms when we compare to Phoenix this year.

Intel is a mixed bag. Meteor Lake is lower volume and I doubt Lunar Lake will be here until very late 2024, it's also on N3B and Arrow Lake is pretty cost heavy as well.


I want to point out the die size for the X Elite is around 170mm^2 on N4P, even slightly smaller than AMD's 8c Phoenix on the same (well, plus a P that offered no density gain) process. While I think QC will still charge a bit more for it, the idea this is going to have a cost structure that runs it's prices into what Lunar Lake on N3B or Arrow Lake and it's tiles and N3/i3 cost is wrong. That bodes well for keeping it pretty competitive especially seeing as it already beats Phoenix by a significant margin on about everything (and likely more efficient iso-perf).


Strix Point though is on N4P, so not as bad there, albeit they'll probably take a further area hit.

Anyways, I don't think the situation is quite as bad for Qualcomm as it seems. When you keep AMD and Intel timelines from announcement into play along with availability from the former and then more importantly, the cost structures of those next-generation products from both companies vs the X Elite it doesn't look as if QC ends up isolated and without a competitor.

 

[Henry swagger]

We'll see how that pans out, I suspect for some mixed usage and lighter workloads X Elite will do pretty well and retain a lot more responsiveness in e.g. a power saver mode. Qualcomm's core cluster design will probably also help them on idle power pending how their power delivery/voltage rails are fit.

RE: 2024: Sure but it'll be *available* around that time. AMD will be announcing mobile Strix around that time and in practice you're looking at much longer until you see ultramobile products (e.g. not stuff with dGPUs for gamers which always seems to come first for them) available - almost certainly Q1/2 2025 before that's actually out in a variety of platforms when we compare to Phoenix this year.

Intel is a mixed bag. Meteor Lake is lower volume and I doubt Lunar Lake will be here until very late 2024, it's also on N3B and Arrow Lake is pretty cost heavy as well.


I want to point out the die size for the X Elite is around 170mm^2 on N4P, even slightly smaller than AMD's 8c Phoenix on the same (well, plus a P that offered no density gain) process. While I think QC will still charge a bit more for it, the idea this is going to have a cost structure that runs it's prices into what Lunar Lake on N3B or Arrow Lake and it's tiles and N3/i3 cost is wrong. That bodes well for keeping it pretty competitive especially seeing as it already beats Phoenix by a significant margin on about everything (and likely more efficient iso-perf).


Strix Point though is on N4P, so not as bad there, albeit they'll probably take a further area hit.

Anyways, I don't think the situation is quite as bad for Qualcomm as it seems. When you keep AMD and Intel timelines from announcement into play along with availability from the former and then more importantly, the cost structures of those next-generation products from both companies vs the X Elite it doesn't look as if QC ends up isolated and without a competitor.

Apple is struggling with sales. Qualcomm wont be any better plus amd and intel will have superior products

 

[SpudLobby]

Apple is struggling with sales. Qualcomm wont be any better plus amd and intel will have superior products

Apple had multiple years of similar or superior growth for the Mac and also has managed to price the M2 products too high relative to the RAM and SSD options.

Qualcomm is in a totally different position on that front and has various Windows OEM's to work with along with a broad appeal in a part that offers great max perf, great perf/W, ample heterogeneous processing, good battery life and without any of Apple's marginal pricing shenanigans.

They are much more similar to AMD and Intel here, albeit with a new product. More importantly, cyclical market winds in industries like this [which would affect AMD and Intel as well for PC's in 2024 and 2025] are a separate question to the long term, secular changes and/or the appeal of a product.

As for superior products, maybe. But also higher cost and arriving later. Strix Point won't genuinely be in ultraportables until 2025 and MTL is meme volume. Arrow Lake on N3 and I3/20A is a mixed bag but not that much better on performance per leaks, and will have a much worse cost structure than Qualcomm on N4P, same with Lunar Lake on N3, though that one is a bigger threat due to the battery life potential.

So we'll see. The die size leak on the X Elite should give people pause relative to how much performance they've eked out of this thing because it's clearly not as expensive to build as anticipated.

 

[FlameTail]

I am curious what the X Elite's succesor will look like.

X Elite will be competitive with Strix Point but not Strix Halo, though to be honest- Strix Halo is like an entirely different class of processor.

Will Qualcomm have an answer to Strix Halo when it arrives?

 

[SpudLobby]

I am curious what the X Elite's succesor will look like.

X Elite will be competitive with Strix Point but not Strix Halo, though to be honest- Strix Halo is like an entirely different class of processor.

Yes agreed. I think the successor will be multiple SKU's though. Probably a 4/6/8+4 part and then a larger one, and maybe a few branches between them depending on clocks. They won't do anything as variant as AMD and Intel though IMO. E.g. they might have two different dice and two SKUs branching out from those, or one larger die with 3 SKUs branching from it w/ some disabled cores and lower clocks on some.

But basically irrespective of the SKU stuff I expect a new node + some architectural changes to the GPU/NPU/P Cores, and then some new little cores. Maybe a bit more system cache.

Will Qualcomm have an answer to Strix Halo when it arrives?

No. Not a chance. That's expensive, niche, and AMD will pull it off only because A) They have a gaming market to target with laptops and compatibility B) they can make it economical with some advanced packaging. Qualcomm has no such market captive or OEM relationship, it's not that huge of a market to begin with, and they're far more likely to simply branch out by having multiple SKU's for different wattages and price tiers IMHO.

And if Qualcomm gets into chiplets for laptops they won't do it until there's a pretty dang good low power overhead + reasonably economical option - plus it'd still have to make sense and at their current die sizes with TSMC yields, meh.

If you want an Arm competitor to Strix Halo, there is another company that might do that eventually.

 

[roger_k]

Did you measure that 5.5W in PowerMetrics within the terminal? Two things assuming so.

Apples private performance monitoring APIs, but I suspect this is the same that powermetrics uses internally. Which doesn’t make your points any less valid of course.

II) You want the whole package power + the DRAM + power delivery stuff, and this doesn't measure that. When Andrei measured the X Elite vs the M2 Max and did all these power measurements, he measured the entire platform (not the screen but like, the package incl SoC and + DRAM + power delivery) minus the idle draw.

I agree with this, however, on current versions of macOS it is not possible to measure the DRAM or power delivery stuff, at least from what I am aware. Older versions of powermetrics did offer information but this feature was removed a while ago. So t is still unclear to me how this comparison is done. I am also confused about the minus the idle draw part. Surely the baseline power consumption of the SoC is part of the system usage too?

This is what you actually want to know about an SoC. Not just core power, or even just package power (sans DRAM), and you want at least a decent measurement tool and methodology.

So this isn't really helpful, you're not dealing with especially accurate measurements and on top of that you're not including other core contributions to power draw that are dependent variables of an SoC design + final platform.

That is all true, and yet the power consumption of 12-core Oryon in CB 2024 is surprisingly high given the level of performance displayed n combination with the efficiency claims.

 

[beginner99]

Anyways, I don't think the situation is quite as bad for Qualcomm as it seems

Not sure. They need to offer a very pronounced benefit over x86 offerings due to guaranteed software compatibility issues. Either they need something like apples rosetta and since the didn't mention it i doubt it exists or then a extreme battery life advantage or extreme cost advantage. Most people don't really need 15h of battery life. and the few times they do a battery back might be a suitable option.
At same price and similar performance I don't really see a reason to pick an ARM offering due to potential issues.

 

[SpudLobby]

Apples private performance monitoring APIs, but I suspect this is the same that powermetrics uses internally. Which doesn’t make your points any less valid of course.



I agree with this, however, on current versions of macOS it is not possible to measure the DRAM or power delivery stuff, at least from what I am aware. Older versions of powermetrics did offer information but this feature was removed a while ago. So t is still unclear to me how this comparison is done. I am also confused about the minus the idle draw part. Surely the baseline power consumption of the SoC is part of the system usage too?

Well, when we measure active (dynamic) power consumption you don’t want the static and idle power draw wherever that floats at. It’s not that we don’t care about the baseline or idle — far from it — it’s just not relevant to the actual under load differential vs 0. I just see it like tare on a scale. Fwiw this is how most measurement has been done from Andrei or formerly early Anandtech, to Ian Cutress with a few Snapdragon chips, to Geekerwan’s videos, albeit with varying access points be it either USB-C/lightning ports or at the Wall like with the M1 Mini review here, or the PMIC stuff (most accurate way). Subtracting idle in the case of like a phone if you’re measuring by USB C or at the wall would also help rid a good bit of the display draw too. Obviously if you measure at the package and DRAM it’s different and you can skip that.

That is all true, and yet the power consumption of 12-core Oryon in CB 2024 is surprisingly high given the level of performance displayed n combination with the efficiency claims.

Yeah, so I think for one thing the Mx/Mx Pro/Max stuff draws more power in total than like what Apple suggest where they just measure the CPU - same goes for AMD/Intel’s “TDP” even at PL1 steady state where they really are operating at x wattage — because in practice that’s only the SoC or maybe package but doesn’t include the overall platform with the DRAM (which is critical especially for chips like this minimizing access) or other losses in power delivery. And based on Andrei’s measurement for the chip here it looks like QC is more honest about it across that curve than Apple who seem to just measure *only* CPU wattage with the M2/3 and AMD/Intel do similar. Though in the MT comparisons to the 10/12 core 1355U or 1360P from Intel for instance — those are fair game and measured by Andrei aka full platform.


The second issue I think is: we look at the MT graphs realistically the X Elite has the option of being pushed over a limit. It still scales very well for a part that has an ideal area that looks to be in the 20-35W range (full platform again) for 12 cores.

In one case where it matches the 1360P’s Geekbench 6 MT for example at 68% less power, or 2x the performance at that same power. Both are being measured by Andrei’s methodology so full platform and it’s quite fair.

The 1360P’s peak in a Galaxy Book from GB6 listings looks like it’s about 9000-10500, which would be our peak value at around that 47-50W point for the 1360P.

Qualcomm’s X Elite can match that at 68% less total power, which puts it around the 16-18W mark and that’s indeed what the graph shows. And mind you that’s whole platform with the DRAM. (https://b2c-contenthub.com/wp-content/uploads/2023/10/Snapdragon-Elite-name-perf-1.png?w=1200)

So they’re (QC) at about 10K GB6 MT, 16-18W total power. That’s very, very godd*mn good. Even Apple’s M2 under load for MT workloads (before/if not throttled) will consume a good bit more than that when measured from the wall and won’t be quite as performant.

The M3 probably brings them a lot closer and maybe ahead here, but depends on how they measure and I know the CPU figures they gave in their presentation don’t include DRAM/total power because they don’t at all match most measurement from the Wall from the M2 (so the 17W peak performance for the M3 will be higher in practice).

And AMD? I think a 7840U with LPDDR5 probably does okay relative to Intel but I doubt it does as well as QC here. And it probably helps QC have more cores, but that’s a skill issue on AMD’s part given they used more silicon for a product with an iGPU and CPU less efficient than Qualcomm and with much less MT at peak should someone choose.

 

[FlameTail]

There is something odd in these charts presented at Snapdragon Summit:


Notice how the X Elite's curve ends at about 30 FPS.

But that's not it's peak iGPU performance. In the reference device benchmark testing, it went all the way upto 44 FPS.

So Qualcomm intentionally cut-off the curve to make the efficiency look better

 

[FlameTail]

Oops sorry.
My bad.

I thought the X-axis showed FPS. It's not. It's showing power consumption.

 

[SpudLobby]

Oops sorry.
My bad.

I thought the X-axis showed FPS. It's not. It's showing power consumption.

Lmao yeah they didn’t cut anything off. Their perf/W just blows AMD out at least on this one benchmark. To be fair with GPUs given software has an effect on performance so tangibly beyond the binary of “compiled for x/y ISA” with CPUs, your “performance per watt” can be altered pretty quickly and I’d prefer us use a benchmark that isn’t from the mobile world like this.

Like so a DirectX12 benchmark such as TimeSpy for the X Elite, if that benchmark is ported to Arm, would be great to see.

I suspect QC’s Adreno really is more efficient than a 12CU RDNA3 setup like in that 7940/7840 above albeit not quite to that extent from the graph where the 780M just never comes close. Probably though the advantage in the 5-15W range is still big looking at Adreno 740 and then RDNA3’s disaster just generally.

What we also know from the Control Arm64 port they demoed on the “23W” TDP device vs the 7840HS or whatever is that the FPS was similar across two settings, like 30 FPS on one (for both) then 60 +-3 (again, for both) on the other setting.

We don’t know power consumption on both of those devices when it was running it, but I’d wager at worst Qualcomm were similar at worst tbh.

 

[Doug S]

So what - X Elite systems won't cost as much and they aren't weighed down by the bus and GPU area Apple have, it's not going for that market. It's more like a vastly better version of AMD's Phoenix tbh.

Why do you believe it won't cost as much? Have they announced ANY information about pricing? What about Qualcomm's overall margins and general greediness makes you believe they will price the systems attractively? If they will have better power/performance ratios than Intel alternatives why won't they be priced at a premium to the competition?

Maybe they won't be priced as high as Apple, we'll see, but they are competing much more with Intel/AMD than with Apple so that price comparison is more relevant.

 

[FlameTail]

Why do you believe it won't cost as much? Have they announced ANY information about pricing? What about Qualcomm's overall margins and general greediness makes you believe they will price the systems attractively? If they will have better power/performance ratios than Intel alternatives why won't they be priced at a premium to the competition?

Maybe they won't be priced as high as Apple, we'll see, but they are competing much more with Intel/AMD than with Apple so that price comparison is more relevant.

Personally, If I were Qualcomm- I would make sure to price it aggressively lower so as to make the biggest bang/splash in the PC industry.

Remember the Apple M1? It landed with such a splash that it sent shockwaves throughout the entire tech community, such that even the most casual tech enthusiasts today know the Greatness of Apple Silicon.

Qualcomm has to make an impression. Yes- they have been making WoA SoCs for the last few years, but now their is a fresh opportunity with the Snapdragon X Elite supercharged by the Oryon CPU.

It would be foolish to become greedy too soon. Like eating the cake before it is fully baked. Remember, the marketshare of WoA chips is miniscule compared to x86.

 

[Doug S]

Not sure. They need to offer a very pronounced benefit over x86 offerings due to guaranteed software compatibility issues. Either they need something like apples rosetta and since the didn't mention it i doubt it exists or then a extreme battery life advantage or extreme cost advantage. Most people don't really need 15h of battery life. and the few times they do a battery back might be a suitable option.
At same price and similar performance I don't really see a reason to pick an ARM offering due to potential issues.

Seems pretty clear they will be relying on Windows' built in ability to run x86 software on ARM, if they something better they would have mentioned it as you say.

I agree the market is problematic as most people don't need more battery life than they are already getting from their laptop - those who do are power users who are probably less likely to consider a non x86 alternative. But there is another niche where they might succeed in that's in silence. Having a laptop that doesn't send its fans into jet engine mode when you doing something that puts a load on it (or doesn't have fans at all) is desirable by enough people that they can make a living on that.

I'm just skeptical of the people who claim ARM is going be a quarter of the Windows market in a few years or whatever. That kind of prediction assumes Intel and AMD are just going to sit still and not address competitive threats they see on the horizon. If there's sufficient interest in Qualcomm based PCs I could see Intel & AMD designing specialized mobile cores - i.e. specifically target the more efficient frequency range Apple/QC CPUs do rather than the 6 GHz their designs do. I think they could compete pretty well on power/performance if they did. Other than maybe an extra cycle or two in decode pipestages, there's nothing inherent in x86 that makes it less efficient than ARM.

 

[SpudLobby]

Why do you believe it won't cost as much? Have they announced ANY information about pricing? What about Qualcomm's overall margins and general greediness makes you believe they will price the systems attractively? If they will have better power/performance ratios than Intel alternatives why won't they be priced at a premium to the competition?

Maybe they won't be priced as high as Apple, we'll see, but they are competing much more with Intel/AMD than with Apple so that price comparison is more relevant.

Well, a 170mm^2 die and a Qualcomm premium over a Phoenix 8c SoC (which of course will happen, before you blow your stack acting like I’m saying they’ll undercut Phoenix) will absolutely still blow Apple out on the overall value proposition because OEM’s sell this thing at the end of the day, and they set the baseline and marginal RAM, SSD, and display/chassis/etc pricing. So RE: Apple, Throwing in an extra ~ $100-150 passed to the consumer over the cost of Zen 4 for a similar class of laptop wouldn’t change a dang thing if we’re comparing to Apple because of how insane they are.

For ex: 14” M3 MacBook Pro that can actually be used actively cooled
$1599
8GB of LPDDR5 6400
and a PCIE3-speed 512GB SSD
(And That display is nice but there are similar 2.5-2.8K 120HZ LCD options for Windows by now)

Aaand here’s where it gets nuts. You want 24GB of RAM and a 1TB SSD in it?

$2,199. You can’t even go further than 24GB on the M3 stuff. Marginal pricing literally at + $400 for 16GB of LPDDR5 6400, and + $200 for 512GB of extra SSD storage that really isn’t that fast. Ofc, more than one 4K display out also isn’t a thing which is killer for a lot of people.

I will bet you that especially as you inch up the RAM and SSD upgrades beyond a baseline X Elite system, generally you’re going to get a better bang/$ than Apple. But again I think this is just low hanging fruit.

Anyway, the fact that even in Charlie’s negative hit piece Qualcomm were willing to give OEM’s $$$ to make up for using their own PMICs and make the die affordable (at least for premium constraints) — and then we later find out that was BS and they can use their own standard stuff anyways from Qualcomm themselves — I don’t think Qualcomm’s pricing will be too insane.

The whole reason Charlie says “this is a chip to watch” based off that die size investigation he did is obvious. He just means QC can afford to sell it for a more reasonable premium (still keeping a decent cut) than almost anyone anticipated relative to the performance of this chip.

Again: will it still cost more than Phoenix or ADL/RPL and a 2+8 MTL system? Yes of course due to PMIC stuff like Apple do and then the fact that it is a competitive part from QC.

But do we think Arrow Lake, Lunar Lake, or Strix Point will be price winners? I doubt it, I think the first two will be a mess on that front because of N3 or 20A/I3 (depends) and the last one will ofc be a larger die than Phoenix and AMD will absolutely charge a premium over Zen 4.

So I don’t think it looks as bad as people think keeping in mind where AMD and Intel are going tbh.

 

[poke01]

$2,199. You can’t even go further than 24GB on the M3 stuff. Marginal pricing literally at + $400 for 16GB of LPDDR5 6400, and + $200 for 512GB of extra SSD storage that really isn’t that fast. Ofc, more than one 4K display out also isn’t a thing which is killer for a lot of people.

You go beyond $2199 for the base M3 14" then that's where you stop and rethink.

The M3 Pro 14" for $1999 is a much better value. It fixes all the issues the base M3 has. 18GB RAM, 2 display out and faster SSD gen 4

 

[SpudLobby]

You go beyond $2199 for the base M3 14" then that's where you stop and rethink.

The M3 Pro 14" for $1999 is a much better value. It fixes all the issues the base M3 has. 18GB RAM, 2 display out and faster SSD gen 4

True.


But you’re still in other compromises thanks to Apple. Like so if you wanted 32+ GB of RAM (in this case 36GB) you can do that, but it’s $2399 or another $400.

And then if you wanted the 1TB SSD, another $200.

That’s $2599 for a 36GB/1TB PCIE G4 combo and the LPDDR5 is just good enough, 192-bit bus (downsized) and 6400. Bandwidth is only ahead of e.g. the X Elite & it’s faster by just 14GB/sec now, negligible. Probably the same for Arrow Lake & Strix Point when they arrive.

Apple’s segmentation that includes display outputs, core counts (which aren’t as reliant on memory bandwidth at all between say the M3 -> M3 Pro 6+6 MT performance, not like an iGPU, they could do an M3 with 6+6.), and caps the RAM from an M3 at 24GB I think forces more people into laptop choices they don’t actually fully want at an obscene level of cruft. See for instance that 14c GPU on the M3 Pro you end up with. It’s just an awkward choice I think with what they’ve gone with.

This is by design naturally, but Qualcomm aren’t doing this kind of segmentation because you simply can’t get away with that in the Windows market (besides the fact that they don’t have and likely won’t any time soon a bigger bus SKU).

Related:

https://x.com/never_released/status/1720806689253859437?s=46

 

[FlameTail]

I agree the market is problematic as most people don't need more battery life than they are already getting from their laptop -

I disagree.

I am sure most users will absolutely desire the improved battery life.

Battery life is one thing you can never really have too much of.

 

[Geddagod]

Battery life is one thing you can never really gave too much of.

Ye, I second this. With stuff like extra perf, people might not notice it much if they are just doing simple tasks. But battery life is something that is immediately and very noticeable.

 

[SpudLobby]

I disagree.

I am sure most users will absolutely desire the improved battery life.

Battery life is one thing you can never really gave too much of.

That’s correct. In practice most laptops don’t really achieve the kind of battery life that like Apple can give (and with high resolution displays and good responsiveness under those constraints). So it’s not really like today the battery life is great for actual mixed loads and decent performance. Honestly people are just out of touch if they think battery life is an afterthought and we’ve hit some saturation line as of today. Not even close for Windows market.

 

[SpudLobby]

Ye, I second this. With stuff like extra perf, people might not notice it much if they are just doing simple tasks. But battery life is something that is immediately and very noticeable.

Yeah. I’m really looking forward to the X on this note.

 

[SpudLobby]

Seems pretty clear they will be relying on Windows' built in ability to run x86 software on ARM, if they something better they would have mentioned it as you say.

I agree the market is problematic as most people don't need more battery life than they are already getting from their laptop - those who do are power users

Dude no. Complaints about battery life ring from FAANG developers to affluent creatives to college students in totally non-STEM majors who use their laptop for Netflix, G Docs or Office and for their pictures.

Probably the only demographic that doesn’t care beyond an absurdly modest and performance crippled satiation point are people here or worse yet, gamers that functionally use their 65-120W laptops as Place A -> Place B desktops to plug back in. You have this wrong on the idea it’s specific to “power users” which is vague anyways and more performance specific or arguably external monitor, loads of RAM and peripheral specific.

I'm just skeptical of the people who claim ARM is going be a quarter of the Windows market in a few years or whatever. That kind of prediction assumes Intel and AMD are just going to sit still and not address competitive threats they see on the horizon.

I think we risk going towards a kind of absurdly strong efficient market hypothesis for consumer competition here. “That kind of prediction assumes Intel are just going to sit still and not address competitive threats they see on the horizon”.

Do we really want to pull this card for…. AMD and Intel even in or especially in the last 10-15 years and over the idea they couldn’t let Nvidia, Qualcomm, and MediaTek grab 10-25% market share by say end of decade?


This argument is also IMO conflating a response with a total suppression of competition. Intel sure had time to see Apple go from 0 to 100 with their cores and SoCs and they just did basically nothing serious to the degree that it wasn’t even purely a cost move on Apple’s part, Intel just straight up sucked in 2020 and even now they still can’t beat an M1 where it counts. Apple doesn’t compete yada yada, sure except they lost an easy customer that constituted an outsized share of their premium chip sales and they had them banked for good otherwise. All of this was even *after* they lost mobile and tablets! They were still flying blind. And it’s not just the fabs. Their cores and uncore are a bloated mess.



Lunar Lake will come fwiw, so Intel clearly already feels the heat. It’s also an expensive part on N3B though, so we’ll see how it goes.

 

[Doug S]

Well, a 170mm^2 die and a Qualcomm premium over a Phoenix 8c SoC (which of course will happen, before you blow your stack acting like I’m saying they’ll undercut Phoenix) will absolutely still blow Apple out on the overall value proposition because OEM’s sell this thing at the end of the day, and they set the baseline and marginal RAM, SSD, and display/chassis/etc pricing. So RE: Apple, Throwing in an extra ~ $100-150 passed to the consumer over the cost of Zen 4 for a similar class of laptop wouldn’t change a dang thing if we’re comparing to Apple because of how insane they are.

Depends on whether the LPDDR5X is going to be on a motherboard or on a carrier for tighter integration like Apple did. If the latter, then Qualcomm will be able to mark it up however they want.

So sure the SSD and display will be supplied by OEMs, and there will be some substandard displays for people who want to save money. Those same savings are available to OEMs making x86 laptops. Do you really believe Qualcomm is going to sell its stuff for LESS than Intel and AMD do? I would bet it is the other way around.

Again I don't think it is matters too much how they compare to Apple because Apple is not their real competition. Apple sells to a different market - people decide questions like "Android or iPhone" and "Windows or Mac" before they start shopping. There just aren't a lot of customers cross shopping Macbook and LG Gram or whatever because the application availability/software compatibility is different and has to be taken into account.

The only exception might be people who are just using their laptop for browsing and watching videos as that is a use case that's truly OS agnostic, but those people aren't buy M3 Max and won't be buying Qualcomm's high end thing either. They would be perfectly happy with what Qualcomm is selling now, or whatever x86 laptop, or a Chromebook for that matter.

 

[Doug S]

I think we risk going towards a kind of absurdly strong efficient market hypothesis for consumer competition here. “That kind of prediction assumes Intel are just going to sit still and not address competitive threats they see on the horizon”.

Do we really want to pull this card for…. AMD and Intel even in or especially in the last 10-15 years and over the idea they couldn’t let Nvidia, Qualcomm, and MediaTek grab 10-25% market share by say end of decade?


This argument is also IMO conflating a response with a total suppression of competition. Intel sure had time to see Apple go from 0 to 100 with their cores and SoCs and they just did basically nothing serious to the degree that it wasn’t even purely a cost move on Apple’s part, Intel just straight up sucked in 2020 and even now they still can’t beat an M1 where it counts. Apple doesn’t compete yada yada, sure except they lost an easy customer that constituted an outsized share of their premium chip sales and they had them banked for good otherwise. All of this was even *after* they lost mobile and tablets! They were still flying blind. And it’s not just the fabs. Their cores and uncore are a bloated mess.



Lunar Lake will come fwiw, so Intel clearly already feels the heat. It’s also an expensive part on N3B though, so we’ll see how it goes.

Intel knew that there was nothing in the world they could do to prevent Apple from dropping x86 Macs. Apple would have switched even if Intel was coming out with better CPUs than what Apple had, though Apple having something better or at least equivalent obviously smoothed the transition for them.

Intel knew that Apple's ability to steal x86 Windows market share was extremely limited, because they run a different OS, and they only compete on the high end not in the mass market. There would be little to gain by 'responding' to Apple because they cannot ever get Apple to make x86 Macs again and will be just as limited in their ability to steal Mac share as Apple is in its ability to steal Windows share. The equation changes if they believe Qualcomm will have a chance to make inroads into Windows market share (I'm still skeptical)

 

[SpudLobby]

Depends on whether the LPDDR5X is going to be on a motherboard or on a carrier for tighter integration like Apple did. If the latter, then Qualcomm will be able to mark it up however they want.

So sure the SSD and display will be supplied by OEMs, and there will be some substandard displays for people who want to save money. Those same savings are available to OEMs making x86 laptops. Do you really believe Qualcomm is going to sell its stuff for LESS than Intel and AMD do? I would bet it is the other way around.

Again I don't think it is matters too much how they compare to Apple because Apple is not their real competition. Apple sells to a different market - people decide questions like "Android or iPhone" and "Windows or Mac" before they start shopping. There just aren't a lot of customers cross shopping Macbook and LG Gram or whatever because the application availability/software compatibility is different and has to be taken into account.

The only exception might be people who are just using their laptop for browsing and watching videos as that is a use case that's truly OS agnostic, but those people aren't buy M3 Max and won't be buying Qualcomm's high end thing either. They would be perfectly happy with what Qualcomm is selling now, or whatever x86 laptop, or a Chromebook for that matter.

Where did I say they would sell it for less than AMD and Intel? Do you even read the responses? I said more than the current generation including MTL, albeit less than LNL which is on N3B and Arrow Lake which will also be on that process at least partially, and probably less than or similar to Strix Point at worst.

I somewhat agree about Apple — there have been some marginal switchers that both sales data and my own anecdata line up for. But it’s true the majority of their base has always used Macs and/or is Apple loyal.