Qualcomm Nuvia vs Apple Mx

[crazy dave]

I think this interview on Anandtech was an interesting read but one thing in particular caught my eye:

Interview with Alex Katouzian, Qualcomm SVP: Talking Snapdragon, Microsoft, Nuvia, and Discrete Graphics

www.anandtech.com

IC: Nuvia’s design will be Arm-based, and the Nuvia team comes from people who built Apple’s M1 and talent from Google. So will that chip compete primarily against Apple because it's ARM-based, or against Intel because it's Windows-based?

AK: The devices that we're going to come out with based on the new CPU design and the new architecture, will compete head on with Apple.

IC: So when you do your comparison charts, you'll be comparing against Apple?

AK: Yes.

IC: Is that because Intel is no longer in the picture? Or is that just because that’s where the market is?

AK: The way we look at it is that we're not going after discrete designs: we're going after an SoC and the architecture that makes the best sense for the PC. Like I said, mobile traits are coming into the PC, and I agree with you that performance wise compared to mobile it’s much higher, it has more power dissipation capability, but we're going to try to make it a lot sleeker and have a lot more mobile based. So our comparison is to a company that can do both types of SoC and bring that capability to the PC, so we're really preparing ourselves for that.

It seems that when Nuvia-based SOCs finally release for PC the primary comparison Qualcomm will be drawing against will be Apple’s M-series SOCs rather than Intel/AMD chips.

That, to me, seems like a mistake. In the mobile space, Apple as the primary point of comparison makes sense. But while both Apple and Qualcomm will be making ARM-SOCs for the computer, the majority of customers that Qualcomm will be looking to acquire are going to be from the Windows-x86 side. For one thing, it’s a much bigger pool of people. I’d think they’d want to focus on giving those customers a reason to switch from x86 to their ARM cores.

Maybe they think their new SOCs will be so good that they won’t have to draw comparisons with AMD/Intel. But that’s not the answer he gave (despite Ian leaving the door open for that).

Anyway, these new chips are a long ways off but we’ll see how they perform and what Apple themselves have out by that point.

 

[Xiao_Xi]

Were Nuvia's server SOCs good?

 

[crazy dave]

Were Nuvia's server SOCs good?

Never released. Qualcomm acquired them first and Qualcomm wanted to focus on laptops. Ironic since supposedly the rumored reason that GW3 left Apple to found Nuvia was the he really wanted to build server chips that Apple didn’t want to make.

 

[Xiao_Xi]

Does Qualcomm think that the PC laptop market will transition to ARM faster than the server market?

 

[crazy dave]

Does Qualcomm think that the PC laptop market will transition to ARM faster than the server market?

Dunno. If I had to guess I think it’s more that they have more recent experience in the laptop market and are pushing into it already. They’ve said these cores will be used for everything though and no market is off the table, including servers.

Apparently the story of Qualcomm’s previous server efforts are a Greek tragedy wrapped in farce.

 

[Adarna]

I think this interview on Anandtech was an interesting read but one thing in particular caught my eye:

Interview with Alex Katouzian, Qualcomm SVP: Talking Snapdragon, Microsoft, Nuvia, and Discrete Graphics

www.anandtech.com

It seems that when Nuvia-based SOCs finally release for PC the primary comparison Qualcomm will be drawing against will be Apple’s M-series SOCs rather than Intel/AMD chips.

That, to me, seems like a mistake. In the mobile space, Apple as the primary point of comparison makes sense. But while both Apple and Qualcomm will be making ARM-SOCs for the computer, the majority of customers that Qualcomm will be looking to acquire are going to be from the Windows-x86 side. For one thing, it’s a much bigger pool of people. I’d think they’d want to focus on giving those customers a reason to switch from x86 to their ARM cores.

Maybe they think their new SOCs will be so good that they won’t have to draw comparisons with AMD/Intel. But that’s not the answer he gave (despite Ian leaving the door open for that).

Anyway, these new chips are a long ways off but we’ll see how they perform and what Apple themselves have out by that point.

Price points will determine how these Qualcomm PC SoC laptops/desktops will be perceived.

Hopefully they be priced below $999 for laptops & below $699 for desktops BYODKM

~80% of desktop & laptops sold worldwide & not just in your 1st world country are priced lower than $999 & $699.

 

[leman]

Price points will determine how these Qualcomm PC SoC laptops/desktops will be perceived.

Hopefully they be priced below $999 for laptops & below $699 for desktops BYODKM

~80% of desktop & laptops sold worldwide & not just in your 1st world country are priced lower than $999 & $699.

Yep, this is one of these cases where the adoption will be driven from the bottom rather than from the top. There is little doubt that x86 will remain the "premium" gamer and performance-oriented choice in the PC market for some time, but if the basic home and office machines will move to ARM it is going to have a huge impact on software availability.

 

[Adarna]

Yep, this is one of these cases where the adoption will be driven from the bottom rather than from the top. There is little doubt that x86 will remain the "premium" gamer and performance-oriented choice in the PC market for some time, but if the basic home and office machines will move to ARM it is going to have a huge impact on software availability.

Mediatek, Qualcomm, HiSilicon (Huawei), Samsung and Unisoc are the top Android chip brands that collectively ship more than 1 billion units worldwide in 2021.

Intel/AMD shipped less than quarter billion units in the same period globally.

Odds are those brands can ruin Intel/AMD/Nvidia provided that Windows 11/12 on ARM has the same user experience as those of Intel/AMD.

I would not be surprised that Intel/AMD SoC will eventually ship as many Macs as Apple. They'll be like IBM's mainframe business where in legacy support is key. This is why Intel is building up its GPU business to be more competitive in the future of SoC.

The market that will be harmed the most would be those who want PCs with replaceable parts as volume will go down as typical consumers will insist on tech that is the same as Macs but at a price point affordable to them.

 

[cmaier]

Big talk from a company that never made a chip.

 

[JMacHack]

I’ll believe it when it’s in hand. So far competing chipmakers can’t match the raw performance of the A-series, nevermind performance per watt.

Still, I’m open to the idea. At the very least Chromebooks will se a significant performance uplift.

 

[MayaUser]

Big talk from a company that never made a chip.

do we really know for sure that they never ever made a chip? maybe a modem chip, a car chip etc
Apple never made a chip, same thing...apple is like the architect in constructions...they design the chip...but the actual construction company build the "architects project"

 

[crazy dave]

do we really know for sure that they never ever made a chip? maybe a modem chip, a car chip etc
Apple never made a chip, same thing...apple is like the architect in constructions...they design the chip...but the actual construction company build the "architects project"

He’s referring to Nuvia which was bought by Qualcomm. While the individuals involved are all highly experienced ex-Apple and Google chip designers, @cmaier is dubious as Nuvia was bought before bringing a product to market so has not yet backed up its claims with shipping silicon.

 

[Krevnik]

do we really know for sure that they never ever made a chip? maybe a modem chip, a car chip etc
Apple never made a chip, same thing...apple is like the architect in constructions...they design the chip...but the actual construction company build the "architects project"

Keep in mind Nuvia never shipped a product before they were acquired by Qualcomm. They literally have no track record at all.

And as far as I know Qualcomm used off the shelf ARM core designs up to this point.

 

[crazy dave]

Keep in mind Nuvia never shipped a product before they were acquired by Qualcomm. They literally have no track record at all.

And as far as I know Qualcomm used off the shelf ARM core designs up to this point.

They used to design their own cores years ago.

 

[Krevnik]

They used to design their own cores years ago.

I stand corrected.

 

[crazy dave]

I stand corrected.

I looked it up to make sure and how long ago also depends on also server vs smartphone. The original Smartphone Kryo cores were fully custom (2015), but all the ones since have been semi-custom ARM core designs (2016-today). Their server Centriq products used custom Falkor cores but after that died in 2018 there were no no more fully custom designs. So they had fully custom cores a little more recently than I thought.

 

[kevcube]

Does Qualcomm think that the PC laptop market will transition to ARM faster than the server market?

as someone who works in the field of running software in the cloud - they will. engineers all over are switching to ARM. Customers all over are unknowingly buying arm machines in their next MacBooks. Enterprises are running VMs that they've had since 2016 and aren't going to take on the light task of rearchitecting their workloads to run on ARM. Even though it's a minuscule change.

However I always push for ARM first and I run on AWS Graviton2 CPUs wherever possible.

 

[mi7chy]

Idea isn't new. That niche market is the no man's zone that sits between <$100 Raspberry Pi like ARM SBCs and >90% x64. No one buying x64 will consider ARM instead and lose access to the richest software ecosystem. And, anyone looking specifically at ARM is better off getting x64 when prices creep too far north of a few hundred $ since, for example, most ARM lack AV1 hardware decode that x64 have. That's why Qualcomm is chasing Apple instead of x64.

 

[Krevnik]

as someone who works in the field of running software in the cloud - they will. engineers all over are switching to ARM. Customers all over are unknowingly buying arm machines in their next MacBooks. Enterprises are running VMs that they've had since 2016 and aren't going to take on the light task of rearchitecting their workloads to run on ARM. Even though it's a minuscule change.

However I always push for ARM first and I run on AWS Graviton2 CPUs wherever possible.

There‘s also the argument that it’s a lot easier to make inroads in the server space if you are developing on the same hardware. So ARM on desktop could well help ARM in the datacenter, especially for folks using tools like Docker/Kubernetes for deployment.

 

[Xiao_Xi]

Customers all over are unknowingly buying arm machines in their next MacBooks. Enterprises are running VMs that they've had since 2016 and aren't going to take on the light task of rearchitecting their workloads to run on ARM.

Companies may also use ARM in the cloud without knowing. It seems most of AWS managed services use Graviton instances.

Qualcomm used off the shelf ARM core designs up to this point.

Which companies besides Apple design custom ARM SOCs?

chipmakers can’t match [..] performance per watt.

Not even server SOCs?

 

[Bodhitree]

So Nuvia started getting publicity in 2019, and according to Anandtech (here) had been going for a year at that time. They are planning to sample silicon late in 2022 and have something worth shipping in 2023. The purchase by Qualcomm was announced in January 2021.

Keeping in mind that the early work that Nuvia did was on bringing a server chip to market, how likely is it that they will have been able to pivot to doing a SoC for laptops in that timeframe? Are they going to slip to 2024? Are we going to get some weird hybrid chip?

 

[crazy dave]

Which companies besides Apple design custom ARM SOCs?

Technically all of them build custom SOCs. They may use off the shelf core designs, but even then they often have some degree of customization in clock speed, caches, even things like reorder buffer. And the arrangement of the cores, sometimes the fabric depending on what they’ve licensed from ARM, etc … is all customizable even if many of them opt for similar configurations. But if you mean fully custom core designs:

1. Nvidia does
2. Qualcomm used to and will again with Nuvia
3. Samsung used to but has no plans to do so again that I know of
4. Ampere plans to (servers)

There might be more but I can’t think of anyone else off the top of my head. I think there may be startups and the like (similar to Nuvia) as I remember people listing other (smaller) companies Nvidia could buy if they wanted to expand their ARM chip design teams.

 

[crazy dave]

Idea isn't new. That niche market is the no man's zone that sits between <$100 Raspberry Pi like ARM SBCs and >90% x64. No one buying x64 will consider ARM instead and lose access to the richest software ecosystem. And, anyone looking specifically at ARM is better off getting x64 when prices creep too far north of a few hundred $ since, for example, most ARM lack AV1 hardware decode that x64 have. That's why Qualcomm is chasing Apple instead of x64.

I see what you’re getting at but that makes less sense to target your comparisons to Apple as Apple doesn’t even compete in that market segment. What would you even compare yourself against? I could make a case that if they’re targeting the ultrabook segment (which they have been doing but is more expensive), then focusing on say the MacBook Air might be the most fruitful point of comparison. It is after all high profile so if you want to grab attention, I get that. But even then, as you yourself said, truthfully the hardest job is that you have to convince Windows people to switch to ARM and at the very least risking running much of their current and especially legacy software at reduced performance.

To me, that says the primary point of comparison should be against the x86 Windows market with particular emphasis on demonstrating that customers won’t notice the loss because “our new ARM cores are that much better at low power” while reaping the benefits of long battery life and so forth. (Obviously this is from Qualcomm’s marketing view) Not for nothing the Qualcomm rep has said things like AV1 are coming so by the time of Nuvia’s release next year it may be included. Of course by the point they’ll be competing against Zen 4, Raptor Lake, and depending on when in 2023 it is released, Meteor Lake. So it better be good even ignoring Apple.

 

[crazy dave]

So Nuvia started getting publicity in 2019, and according to Anandtech (here) had been going for a year at that time. They are planning to sample silicon late in 2022 and have something worth shipping in 2023. The purchase by Qualcomm was announced in January 2021.

Keeping in mind that the early work that Nuvia did was on bringing a server chip to market, how likely is it that they will have been able to pivot to doing a SoC for laptops in that timeframe? Are they going to slip to 2024? Are we going to get some weird hybrid chip?

Apparently the transition from server to laptop chip is not that difficult or not as difficult as one might think according to some tech observers - it’s more a question of taking some features out and generally for servers and laptops having die and power efficient cores is beneficial to both, so it may not have caused much if any redesign. More problematic is what is mentioned in the interview - Nuvia isn’t just a core it’s an SOC so they have to develop all the things around the performance core like efficiency cores and fabric. Any custom E cores are unlikely to have been designed for the servers and we have no idea what was the fabric they had planned on and what Qualcomm will be able to repurpose from their current technologies - I don’t believe they use ARM’s fabric which they wouldn’t anyway for Nuvia as I believe you only get that if you’re using ARM’s cores. In fact if I remember right there’s a complex web of restrictions if you only partially use ARM IP (like ARM E-cores but your own custom P-cores) which caused inefficient multicore designs in some early Android CPUs.

 

[huge_apple_fangirl]

It doesn't matter how good Nuvia is. If Microsoft doesn't start giving a s**t about Windows on ARM it will still be a failure. Of course they are trying to compare themselves to Apple- they will be able to make themselves sound more favorable just be getting close to Apple in performance, whereas Intel and AMD can say "we can run all your software, guaranteed." In reality they aren't competing with Apple at all, so even if they declare victory it is meaningless.
Same as Intel trying to focus on "beating Apple" to distract from TSMC crushing them in fabrication and AMD in data center.

Also, according to https://www.semiaccurate.com/ which predicted a lot of Intel's delays before they were public, the Nuvia acquisition is unlikely to bear much fruit. So let's wait till Qualcomm stops hyping vaporware and actually releases something.