Qualcomm Nuvia vs Apple Mx

[retroneo]

What features differentiate a real-time processor such as Apple's R1 from those used by IPhone and Mac?

A real-time processor is one without an MMU (virtual memory). There's now a hybrid core called the Cortex R82 that has an MMU, so it can run a normal OS like Linux or in Apple's case xnu. xnu can be running and when a real-time event occurs, the processor can switch to handle the real-time event, then switch back to xnu. The Apple R1 is probably something like 8x Cortex R82 cores, DSP cores and a 16 core Apple Neural Engine and a large system level cache. Deterministic performance.

It would do the eye tracking and gesture recognition before the M2 even receives the image. LiDAR Mapping, oncoming person detection and fusion of the two camera images together would also be performed by it. These can't occur on the M2 as they need to occur constantly and in real-time. Better to use highly optimised code that can continuously had over a continuous stream of image data to the M2, with eye tracking and gesture recognition already done.

This means visionOS can work more like the windowing compositor on macOS, and inputs appear to visionOS like touch events on iOS.

 

[Xiao_Xi]

Apple R1 is probably something like 8x Cortex R82 cores with a 16 core Apple Neural Engine and a large system level cache

Is the neural engine suitable for sensor fusion? Are sensor fusion algorithms based on machine learning?

 

[retroneo]

Is the neural engine suitable for sensor fusion? Are sensor fusion algorithms based on machine learning?

It wouldn't need the Neural Engine for fusion of the main camera images together, but it might use it for gesture recognition. It might use DSP cores to do the fusion.

Also possible that the R1 instead has real-time microcontrollers for eye tracking and LiDAR as well as something like a cut down A15 with fewer CPU cores, GPU, NPU along with a stereo vision ISP. It would run a cut down OS based on xnu, like they have done with the touch bar and iPad HDMI adapter.

 

[Longplays]

Counterpoint

www.counterpointresearch.com

ARM laptops will materially harm legacy x86 PCs...

In just three years, 30% of PCs may be built on ARM architecture

A new report suggests that ARM may appear in many more computers in the near future.

www.digitaltrends.com

From March 2020 to March 2023 COVID forced an upgrade on laptops & desktops.

Apple observes that Macs are replaced every 4 years while Intel sees PCs to be every 5-6 years.

So the next replacement cycle will occur 2024-2029.

Qualcomm Oryon chip specs to be unveiled at Snapdragon Summit 2023

Qualcomm has high expectations for its upcoming Oryon processors.

www.windowscentral.com

By 2024 the 1st Windows 11 on ARM laptops using NUVIA tech will come out.

This ARM laptop will be using a 4-5nm SoC, ~$150 cheaper to make than x86 equivalent, longer battery life, better performance per watt & possibly a built-in 5G modem.

If Microsoft and Qualcomm executes this correctly they will mortally wound x86.

With reduced sales of x86 chips will result in higher MSRPs of AMD/Intel computers due to worsening economies of scale.

This will likely piss off PC gamers who insist on a AMD Ryzen 9 or Intel Core Ultra 9 as they are inflexible with "good enough" and demand "perfect".

This is similar to Nvidia hiking prices to cover the crypto crash resulting in less dGPUs being sold.

I hope Lenovo comes with a ThinkPad E-series selling for $799. Immediate corporate purchase for us.

I would not be surprised that by Dec 2031 ~80% of Windows 11 PCs will be ARM. This would be my expected Windows 11's end support date.

Windows 11 officially supports 14nm Intel 8th gen chips from Sep 2017 & 12nm AMD 2nd gen Ryzen chips from April 2018.

Qualcomm: Nuvia Acquisition Could Disrupt Desktop PC Market For Windows Computers

Qualcomm is now apparently planning to enter and disrupt the desktop PC market for Windows computers. Click here to see why QCOM stock is a Buy.

seekingalpha.com

Stock analysts hits it on the head when execution is flawless.

 

[Xiao_Xi]

ARM laptops will materially harm legacy x86 PCs...

So far, there is no competitive Arm-based SoC and the only one that could be has its viability in question.

ARM laptop will be using a 4-5nm SoC, ~$150 cheaper to make than x86 equivalent, longer battery life, better performance per watt & possibly a built-in 5G modem.

What makes you think that laptops with a Qualcomm SoC would be cheaper than laptops with an Intel CPU? Surface Pro 9 with Intel CPU is $200 cheaper than with Qualcomm SoC.

What makes you think that Oryon-based SoCs would be cheaper than AMD/Intel CPUs? AMD and Intel don't need to pay a license fee for the ISA they use.

 

[bobcomer]

If Microsoft and Qualcomm executes this correctly they will mortally wound x86.

They wont, and it wont even if they get better than they are now. Arm laptops and desktops just don't sell well for a reason. They're slower, and they don't run everything x86. Why would I want to buy them for our company given those constraints when an x86 of any kind would be that close in price. (and I don't think qualcomm would be cheaper as it isn't that way now)

 

[Bodhitree]

I’m surprised Qualcomm have been so quiet about the Nuvia-powered chips. They just gave them new names within the existing naming scheme and said release towards the end of 2023 and then all went quiet.

It looks like they are expecting it to not make that much of an impact. And we will have to see how the lawsuit with ARM turns out.

 

[Xiao_Xi]

With reduced sales of x86 chips will result in higher MSRPs of AMD/Intel computers due to worsening economies of scale.

I find it very funny that economies of scale only apply to Intel/AMD, but not Qualcomm. Intel/AMD would be forced to raise their prices as they lose market share, but Qualcomm would offer competitive prices from the start with zero market share.

 

[Sydde]

Qualcomm would offer competitive prices from the start with zero market share

Not quite zero marketshare, as they have MS-inertia/momentum to ride on, but a common business practice is to price a product where you think it will be sustainable and start out losing money while building sales until you reach the proper volume/profit balance.

 

[Sydde]

They're slower, and they don't run everything x86. Why would I want to buy them for our company given those constraints when an x86 of any kind would be that close in price.

They will still continue to pants x86 in perf/W, and there is basically no path for x86 to get even on that. They might be able to get kinda close, but then ARM or RISC-V will jump ahead a little more. This is a metric that is becoming more important, even for datacenters.

And they do not have to be able to run everything. The vast majority of users can get the job done without having to use exclusive applications, and the situation is becoming even more pronounced: by later this decade, we will have local-running ML-based tools that can construct the application you need based on "this is what I want", so platform relevance will contract rapidly.

 

[Unregistered 4U]

ARM laptops will materially harm legacy x86 PCs...

I would imagine that Google’s not the only company that’s replacing Intel laptops with ARM Chromebooks for their employees that are just checking emails and completing internal online request forms. Mainly because Google’s likely willing to give a “better than free” deal for the administration of these.

Then folks go to conferences, speak of their “Chromebook success stories” that got them a promotion/more pay and the process repeats. ARM doesn’t HAVE to run Windows to replace laptops that spend most of their time in a browser or email client.

 

[bobcomer]

And they do not have to be able to run everything.

They do if they want to compete in the x86 market. That market expects compatibility the most.

by later this decade, we will have local-running ML-based tools that can construct the application you need based on "this is what I want",

That will probably happen eventually, but not as quick as you think and it wont be really trusted to begin with. Once you get people out of the equation, it kind of becomes pointless, if nobody has jobs, nobody will have money to buy anything. That's something corporations will have to think about when they will produce product like that. I mean, why would you need an office staff can be replaced by one person -- you wouldn't, and pretty soon you wont need that one person either.

One has to always think about the consequences....

 

[Longplays]

I would imagine that Google’s not the only company that’s replacing Intel laptops with ARM Chromebooks for their employees that are just checking emails and completing internal online request forms. Mainly because Google’s likely willing to give a “better than free” deal for the administration of these.

Then folks go to conferences, speak of their “Chromebook success stories” that got them a promotion/more pay and the process repeats. ARM doesn’t HAVE to run Windows to replace laptops that spend most of their time in a browser or email client.

Correct in markets that are open minded about desktop OS other than Windows.

 

[Xiao_Xi]

They will still continue to pants x86 in perf/W, and there is basically no path for x86 to get even on that.

None of that matters if laptop manufacturers continue to make more money with x86. First, Qualcomm needs to provide a good product at a competitive price and later, match Intel's "help" to laptop manufacturers. Do you think Qualcomm would succeed where AMD hasn't?

Not quite zero marketshare, as they have MS-inertia/momentum to ride on

What marketshare? The marketshare of the Lenovo ThinkPad x13s and Surface Pro 9?

 

[thenewperson]

Do you think Qualcomm would succeed where AMD hasn't?

Definitely. For one thing, Qualcomm probably won’t have absurd availability issues wrt their SoCs.

 

[Xiao_Xi]

Qualcomm probably won’t have absurd availability issues wrt their SoCs

Didn't AMD's notebook CPUs have availability issues because they bring in less money than phone SoCs and cloud CPUs? How could Qualcomm solve the availability problem?

 

[Longplays]

Here's something to mull on by Nvidia cheerleaders.

Nvidia's doing an Apple by creating their own SoC.

NVIDIA Grace CPU Superchip

The breakthrough CPU for the modern data center.

www.nvidia.com

It was brought up 1+ year ago in this video.

APUs & SoCs are the direction everyone, even Intel, is going towards to. Users are having a cow over this because it is Apple among the most prominent companies doing it 1st but when companies like AMD/Intel/Nvidia/Qualcomm are eventually doing it for the very same reason of efficiency and performance per watt then they tend to quiet down and point to PassMark scores of these future APUs & SoCs.

 

[dmccloud]

Didn't AMD's notebook CPUs have availability issues because they bring in less money than phone SoCs and cloud CPUs? How could Qualcomm solve the availability problem?

That doesn't make sense. Availability would not be affected by profit margins. It WOULD be affected by demand outpacing supply. If AMD chose to manufacture fewer laptop CPUs because of lower profit margins, then availability is affected as a secondary effect, with lower production being the cause.

 

[bobcomer]

Availability would not be affected by profit margins.

Sure it does. when you have to hire someone else to make your CPU's. The bigger, more paying customer is going to get first dibs. (I'm talking about the fab, not AMD itself.). That's the drawback to not having your own foundry, you have to wait in line.

 

[thenewperson]

Didn't AMD's notebook CPUs have availability issues because they bring in less money than phone SoCs and cloud CPUs?

I dunno, you’d have to ask them why they can barely ship those ¯\_(ツ)_/¯

How could Qualcomm solve the availability problem?

Probably by actually shipping.

 

[Xiao_Xi]

Probably by actually shipping.

Are you suggesting that the real reason AMD couldn't ship its APU to notebook manufacturer was because AMD so decided, not because TSMC couldn't provide enough wafers or because Intel "helped" the notebook manufacturer?

By the way, Intel received an antitrust fine in 2009 (and subsequently overturned) for blocking AMD's market by giving rebates to Dell, HP, and Lenovo.

 

[leman]

Are you suggesting that the real reason AMD couldn't ship its APU to notebook manufacturer was because AMD so decided, not because TSMC couldn't provide enough wafers or because Intel "helped" the notebook manufacturer?

The process was mature enough by the time AMD was producing their CPUs, so I very much doubt it’s a yield issue. Most likely not enough orders/limited production for other business reasons.

 

[senttoschool]

They wont, and it wont even if they get better than they are now. Arm laptops and desktops just don't sell well for a reason. They're slower, and they don't run everything x86. Why would I want to buy them for our company given those constraints when an x86 of any kind would be that close in price. (and I don't think qualcomm would be cheaper as it isn't that way now)

You list the problems blocking ARM from taking Windows share then state that it doesn't matter if those problems are solved over time. Weird.

Anyways, I think Qualcomm will make a huge impact in Windows laptop world if their Nuvia cores are competitive with Apple. They'll start targeting the office/student market first with SoCs that are drastically more efficient than AMD/Intel and applications that do not depend on deep x86 optimization. Over time, anyone building applications on Windows will compile for both ARM and x86.

It'll be a long grind to take x86 marketshare in Windows world. But Apple Silicon is so good that Microsoft has no choice but to make Windows on ARM a priority if they don't want to lose market share to Macs over time. In other words, Microsoft needs Qualcomm to compete. Microsoft can't rely only on AMD/Intel anymore.

 

[bobcomer]

You list the problems blocking ARM from taking Windows share then state that it doesn't matter if those problems are solved over time. Weird.

No, I didn't say they were "solved" over time. I said they'd get better. I also said that it had to run everything, and I mean everything, that x86 runs for it it to take over significant market share from x86. I don't think they will ever be solved, so I don't see ARM taking up much market share. Some, yes.

It'll be a long grind to take x86 marketshare in Windows world. But Apple Silicon is so good that Microsoft has no choice but to make Windows on ARM a priority if they don't want to lose market share to Macs over time. In other words, Microsoft needs Qualcomm to compete. Microsoft can't rely only on AMD/Intel anymore.

I don't share your enthusiasm for Apple Silicon nearly as much. I can get faster over in x86 land and on the efficiency front, intel and AMD are getting better. There's nothing magic about ARM...

 

[Xiao_Xi]

The process was mature enough by the time AMD was producing their CPUs, so I very much doubt it’s a yield issue.

It looks like AMD chose to sell server CPUs.

AMD and Intel CPU Market Share Report: Recovery on the Horizon (Updated)

Hope is on the horizon, mostly.

www.tomshardware.com