Comment by mdasen | Hacker News Reader

Comment by mdasen | Hacker News Reader

The question I have around Google's CPU ambitions is what their end game is.

I understand Apple's end game. They sold lots of iPhones and wanted to in-house the chips for that. Makes sense. Once they had that working well (and Intel was falling behind on their chips), they decided to put in-house chips in their macOS computers. Makes sense.

AWS rents more CPU time than anyone out there. Makes sense for them to want to develop some custom chips for their stuff.

What's Google's end game? They don't actually sell a lot of chips. I'd understand them wanting to create a server chip, but they seem to be targeting phones and laptops. Is Google looking to become a major player in Android hardware? I like the Pixel phones, but they're a very small part of the market. Chromebooks are usually built by companies like HP and Acer, not Google directly.

The article says that Google's Pixel sold 7M units in 2019 (their highest year) and they're looking for 50% more than that (so 10.5M) for the Pixel 6. In 2019, Apple shipped 215M iPhones (over 30x the Pixel). Is Google looking to make the Pixel a much larger part of the Android ecosystem? Are they looking to compete directly with HP, Acer, and others in the Chromebook market?

I guess I wonder what the end-goal is here. Maybe the premium on Qualcomm's Snapdragon chips is so high that they'd rather make their own. Maybe they want to become the primary hardware vendor in the Android ecosystem - or at least a much larger one. But Google never seems to push its own hardware aggressively so it's a bit curious.

Apple didn’t decide to design their own chips because they sold a lot of phones, it was a differentiation play not a volume play. Designing their own chips means they can offer unique features it’s hard for their competition to replicate, like the machine learning acceleration stuff and hardware enabled security features.

The Android OEMs are trapped in a perpetual race to the bottom, so they can’t afford to invest in premium hardware features. Google has realised the only way for them to compete head on with Apple on hardware enabled features is to design their own hardware because nobody else is going to do it.

> Android OEMs... can’t afford to invest in premium hardware features.

I think there are a lot of android phones that have had premium features: e.g. foldables, 90-120 Hz refresh rates (Apple only recently caught up), Vivo x70 Pro Plus w/ an amazing camera that beats the best iPhones (https://www.youtube.com/watch?v=n0r2rENgvwY), etc.. I think there's just a perception that Apple hardware is better, but I feel like they put in above average hardware overall w/ good polish. These are all hardware features. Sure, the performance of Qualcomm chips lag Apple's, but that's because Apple's engineering team is better than Qualcomm's (IMO). I am sure that Samsung, etc. would buy a faster chip for their flagships if it was available. E.g. Samsung, etc. typically fill their flagships with crazy amounts of RAM and Storage.

Disclaimer: My views are my own, and not of my employers.

It’s true that Android has had some competitive devices but you’re usually getting one good feature and a bunch of sub-Apple features: great camera but a slow CPU or years shorter software support, etc.

The problem is that Apple is the hardware vendor getting recurring revenue from usage. Qualcomm needs you to buy a new phone to see anything more, only Google gets a cut of you buying stuff on the Google Play store, etc. That gives Apple both much larger revenues per customer and an incentive to keep your old phone working as long as you’re buying stuff with it. Google trying to get into that model seems healthy from the perspective of getting better competition but worrisome for the level of resources needed to compete.

Why mention your employer?

Because if I don't, people end up alleging potential bias in their replies to my comment, even though I feel there's no connection. So, I just throw it in there pre-emptively.

acdha addressed the feature issue very well I think, it's not just about one or two features and when Apple does execute a feature they often do it better. ProMotion actually saves battery overall by often clocking the display down fo static images. Qualcomm's engineering team is just fine, the problem in their economics don't work for making truly high end chips. It's the most expensive engineering they do, but only a tiny fraction of the Android handsets sold carry flagship chips. Meanwhile all iPhones are effectively flagship devices.

A lot (but not all for sure) of the A-series chip performance advantage comes from a massive on-die cache for example. That's not fancy engineering, just brute force transistor count. Qualcomm engineers could absolutely do that, but Android handset economics won't support it.

Variable refresh rate display is adopted before Apple (like Galaxy Note 20).

Samsung had their own chip division. Not everyone is trapped, like laptop manufacturers.

I suspect their end game is building machine learning features into the silicon.

Most Chromebooks at the low price point use awful low-end intel CPUs that seem unable to cope with opening more than 2 tabs at once. If Google can create a CPU that is competitive in performance and price to those terrible Intel based ones, but also chuck in some hardware assistance for ML inference on-device, they're likely on to a winner.

What user-detectable effect would ML have on a Chromebook? They are just used for web browsing and word processing, aren't they?

Above 80% of Internet web devices are running Android. Apple made M1 which is very efficient power usage and speed this would be a Google counter move.

Consumers will get more power efficient devices.

Intel may get affected by this.

Google doesn't need a solid plan to do something, they have more money than investment opportunities and since they are so "skilled" at shutting down projects it wont be a long term cost if it doesn't work out.

> They sold lots of iPhones and wanted to in-house the chips for that. Makes sense.

The Intel situation is more obviously, but could you elaborate on this? Was it to improve margins or for CPUs optimized for phones?

Google wants everyone using Chromebooks instead of Windows laptops.

Because almost no one uses Chromebooks except Google employees, so far their strategy involves contracting all the developers they can.

Chromebooks are very popular in education.

| no one uses Chromebooks except Google employees

There were ~30M Chromebooks sold in 2020[1]. I think many of those were sold to non-googlers.

[1] https://www.statista.com/statistics/749890/worldwide-chromeb...

Shipments doesn't mean being sold.

Some of those landed here in German stores, where they usually get reduced every couple of weeks until finally someone buys them.

In the current situation I wouldn't be too surprised if prices for these Chromebooks went up tenfold with queues forming at the stores from midnight every night - everyone in the queue working for some auto maker desperate to get their hands on some (ANY!) electronics parts ;-)

They are huge in education, and I am guessing they are seeing the new hybrid work world as a huge new avenue where people might want a simple/easy/disposable laptop to give employees for working remotely.

US education system yes, everywhere else not much.

Here in Germany they tend to stay on some corner being increasingly reduced until someone finally buys them.

If they wanted Chromebooks to succeed they should let you use one of them without giving it your telephone number.

Seriously, try using a Chromebook in anything other than the crippled "guest mode" without giving it your telephone number, a telephone-number-linked gmail account, or jailbreaking it. You can't (unless it's part of a corporate/education site bulk purchase).

Completely moronic policy decision.

Without seeing a good reason I would expect it’s to have control of the entire vertical of ‘you’ from hardware through to surveillance ad tech engine.. yikes

Reply on news.ycombinator.com