50 comments

[ 2.8 ms ] story [ 117 ms ] thread
This certainly fit into certain type of perspective that is common on the internet and HN. Absolute performance regardless of energy usage.

Another problem is people blindly trusted Apple's number on GPU. And while Apple's benchmarks in terms of CPU and Battery life tends to be fairly accurate if not often under stating it. Their GPU numbers and graphs are a marketing spin to say the least.

Surely it is all about the Watts? OK, I think the M1 iMac is a lame duck, but for laptop use, the M1 is just more useful. Let's see what the M2 (or whatever) delivers in a Mac Pro (which is what pros use anyway).
As the article alludes, most of the benchmarks for the M1 presented in the article are meaningless as the software is either running virtulised under rossetta (AND) OR not utilising the GPU/Neural Cores via the Metal API (Cuda equivalent). Expect things to even out over the next couple of months as software is patched and recompiled to run under an ARM architecture.
The benchmarks are meaningful for people who need to use that particular software. They also won't extrapolate well to software that is better optimized for ARM.
No. You shouldn't. This is clickbait. All their benchmarks are meaningless, as they're comparing native x86 using GPGPU to virtualized Rosetta using no GPU acceleration.
People want to run x86 code.
Not me. I have assembly-olfactory synesthesia and x86 code smells like the bad parts of gasoline.

Kidding aside, unless you’re looking to play games, most apps have either ported to Apple Silicon, or have a Beta of their port that is releasing soon. I think the M1 Pro/Max release was the kick in the pants that companies/projects like Adobe, Unity, Unreal and Blender needed in order to go “OK yeah, professionals use these, time to get porting”.

(I know some of these betas preceded the release of these laptops, but for a few months before these laptops came out it was clear they were going to be nuts)

That's fine. We absolutely need benchmarks comparing native x86 performance against virtualized performance thru Rosetta 2. A lot of code isn't getting re-compiled for arm anytime soon.

All I'm saying is this article is clickbait because it frames these comparisons in the dumbest way possible as if to say 1 processor is better than the other, with no explanation that these are not 1-to-1 comparisons. It is a complete lie by omission.

No they don’t. People want to run fast code. Rosetta is just a stopgap.
Original announcement: https://www.nvidia.com/en-us/geforce/news/geforce-rtx-laptop...

As far as I can tell, the benchmarks are done with software packages^ that support a GPU rendering path for NVIDIA chips via CUDA / OptiX, but not yet with M1 via Metal (or are even CPU-emulated on M1 via Rosetta). So it's "fair" in the sense that this is what the market currently offers for users, but it's not an apples-to-apples comparison on the hardware.

^ e.g. Blender 3.1, currently in alpha, adds a Metal GPU backend to Cycles, which in early reports gives about a 3x - 4x improvement in render times vs the CPU path on M1 machines (https://devtalk.blender.org/t/cycles-apple-metal-device-feed...).

>but it's not an apples-to-apples comparison

No, it's an apples-to-nvidias comparison

While I'm not a fan of Apple in most regards, I think their approach to mobile hardware is much more coherent than NVidia and Intel right now.

Generally speaking - The CPU just isn't the performance bottleneck on most machines that people are intending to carry with them during the day.

The performance bottleneck is the battery and the temperature.

I think it's incredibly telling that from this chart (produced by intel, designed to promote intel products) my takeaway is intel has missed the mark entirely. The relative performance of their chips just doesn't matter when you realize that the M1 chips are SO MUCH BETTER at power usage. The ranges for power draw barely even overlap - the M1 is getting you ~100% of the performance, using half the power of the lowest power intel chip.

That's the win. Full stop.

It means the machine with that CPU is going to run cooler for longer. It means that cpu isn't going to be thermally throttled so you can actually use all of it (unlike that intel). It means you still have a running machine 15 hours later, instead of having a dead brick after just 5.

Ethically - I refuse to buy any general purpose machine Apple makes, but I'm sad Intel is so far off the mark here.

What are your ethical issues with Apple? Why mention them in the first place without further explication? Your other points rang true for me but this seems a little tangential.
For me it's the hostility towards developers and users. I can't even lower the dang contrast on my iOS or MacOS devices to make them a little less dopamine inducing. Sure you can switch to B&W but then I lose usability (Is that link clicked?). The massive cut they take from creators is another pain point. These are just off the top of my head, I'm sure others can chime in.
No doubt, Apple is more opinionated when it comes to UI/UX than others — for better or worse. It's a trade-off.

Fwiw, Apple does provide more control over the display tucked away under Settings/System Preferences => Accessibility (lots of goodies there, actually).

Not clear if you knew that and still found that too limiting, but perhaps others will have fun poking around in there.

At least on the Mac, I can almost always find a third-party tool that can accomplish anything Apple didn't expose.

I largely agree they need to reduce their App Store fees — it is damaging their goodwill for relatively limited benefit, but more-so it's attracting government scrutiny that will likely lead to worse outcomes for Apple and its users if they don't relent.

The dopamine inducing element is addictive apps like Facebook, not the color levels of the display.

You can simply delete the apps that have that effect.

Tell that to my apple mail icon.
You can delete that if you don’t like it.
> What are your ethical issues with Apple?

I don't believe you can genuinely purchase an Apple phone at the moment, and I think they're trying pretty hard to move their laptop/desktop machines in the same direction.

Namely - Much of their mobile ecosystem is utterly hostile towards non-Apple developers - including the owner of the device.

They lock hardware/software features behind entitlements that require Apple's permission (and of course an accompanying fee) to use. Good luck using multicast on your personal device on your personal network. Turns out you can't. The hardware can, of course, but only if Apple allows it.

In fact, turns out almost all of the software Apple provides to run their hardware treats you like a user, and not an owner. They riddle (and cripple) the product with digital locks, and do not provide the keys to the owner.

Instead they provide keys only to apps submitted for review to their store, signed with their certs. So again, it's hard to say I actually own the damn thing.

Is technological lock-in, or "hostility" toward a user as you put it, an ethical issue?

I would think the use of Chinese labor abuses to make iPhones is the ethical issue. Funny you didn't mention that.

Yes, but if one used that as their moral standard they wouldn't be able to purchase anything made by an international company. It's really easy to bash iPhone on things they do exclusively, but harder on industry practice.
Wrong.

You are making the "all or nothing" fallacy.

You absolutely can make the ethical choice when you have the choice. In some cases, you are simply forced to make the lesser evil choice.

Transparency is the first step. Yes, in some cases it is unavoidable, but to translate that into overall inaction, IMHO, is lazy.

That's reductive of my point. The overall being one's choices for a new smartphone are iPhone, Android, or abstain. When the parent comment elected Android for "ethical" reasons those reasons would not include the labor conditions of the phone's manufacture since that doesn't vary between most Apple and Android phones.
> That's reductive of my point.

> they wouldn't be able to purchase anything made by an international company

I suggest you don't make sweeping generalizations, then.

> The overall being one's choices for a new smartphone are iPhone, Android, or abstain.

That's why I took care to point out "In some cases, you are simply forced to make the lesser evil choice."

> When the parent comment elected Android for "ethical" reasons those reasons would not include the labor conditions of the phone's manufacture since that doesn't vary between most Apple and Android phones.

You're not allowing for someone to still consider the ethics of an issue: it's not like subtraction, if A has it and B has it than neither have it, so why consider it. IMO that's reductive. It can still be considered.

From my understanding, the factories that inflict these terrible conditions are shared between companies. The lesser of two evils isn't something that is easily findable. Apple's public controversy was specifically an investigation of Foxconn factories in China who have clients other than Apple. You've yet to convince me that the same workers in the same factories with the same terrible working conditions are somehow more justified if they're not putting an Apple logo on the products they manufacture.
Wouldn’t be able to go with pretty much any one then. Probably wouldn’t be able to wear much of your clothing either if labor abuse is an issue.
Gustin and Everlane.

These two clothing companies provide 100% transparency on their sourcing and labor practices.

You absolutely can choose to wear clothing that doesn't enslave people. Maybe not 100% (shoes are still difficult to source), but it doesn't mean throwing your hands up and walking away. Or being cynical toward someone who is trying.

Sorry if you do, do that. I’m almost every other case, people are not consistent with this sort of stuff. I was wrong with the way I wrote my comment and my attitude.
Apple has the strongest protections against unethical labor of any tech company. The main reason we know of any issues is because apple is literally the only company that audits and publishes the results of supplier audits.

If you think that somehow the people working on apple hardware construction are treated worse than the people working on the vastly higher number of android devices, etc you need to provide some evidence.

I recall a case where apple audits found a supplier using child labour, included that in their report, included the remediation (including cases where they outright dropped suppliers). The headlines weren’t “apple found child labour and stopped it”. The headlines were “apple uses child labour”.

The reason you don’t get that for other manufacturers is that they a) don’t audit, and b) don’t publish details even if they somehow become aware.

(comment deleted)
Apple acquired a large number of Intel chip designers and architects about 2 years prior to M1 launch. That might not have been enough time to impact the overall architecture of M1, but I think it is possible M1 carries a lot of Intel design influence. I'm thinking it is a little odd for Intel to attack the M1 when it was likely designed by Intel engineers.
Two years is not long enough to do a major redesign — Apple started hiring back towards the end of the PowerPC era and they were delivering very competitive designs years prior to the M1 announcement. When they announced M1 most people had been predicting it because the phone/iPad side was so strong.
No, a few engineers 2 years ago are not going to change basic architectural concepts that apple has clearly Cavour that run pretty contrary to intels design concepts.

Also designing a cpu takes time: a chip that is released in year X was likely functionally locked at least a year ago, and general architecture would necessarily predate that by a few years.

No CPU manufacturer is designing a cpu for year X, and then starting work on the following years chip. They will be simultaneously designing chips for sequential years, years in advance of release dates. That’s why Intel and AMD can produce their “next years chip” talks. The chip isn’t ready for final production, but there won’t be any significant changes by that point.

There’s also a 8-12 week latency on manufacture, so if they are targeting a release date, they have to have started manufacturing 3 months before then in order to have any chips to sell.

This. Basically ALL CPUs put out by Intel in the last decade all throttle when under load on ultra mobile platforms. Cool Intel you can keep up with Apple for 10 minutes before your CPU slows to a crawl, no thanks.
Well, of all the reasons this might be the biggest one to buy Apple.

Must feel amazing when your competition does your marketing for you.

have you seen the charts apple uses in presentations? they deserve any mockery going their way
Does it really matter? People that spend more than $1,000 on a laptop buy Apple.

Edit: Some outliers really want me to write “overwhelmingly most” or find that old market study? Maybe. But I can also spend 4 points so fire away.

False. I like windows and have been using XPS laptops for a long time.
do you really think this is true? because it isnt.
This is a pointless distraction from the article. Professional users pay more for their laptops, no matter what platform, and given that the kind of laptop I need to do my job has gone down in price by at least a factor of two that price matters less than whether it does the job well.
I had no idea my Lenovo Thinkpad X1 was actually made by Apple! TIL!
Interestingly it is the base M1 that is still the most interesting device, lower cost, lower heat, good single core performance, capable of most tasks, and reasonable graphics for a lower power device. Other fast laptops are all about the heat and the fan noise (my tiny work laptop is the loudest device in my room) A number of times I have got off the train with a laptop almost on fire in my bag, because windows failed to notice the laptop lid was closed. Ill be getting the M1 laptop next time I need one.
i have an M1 macbook, Intel/NVidia can lie all they want they can't hide facts

M1 > to what ever they produce TODAY and tomorrow, it beats the hell out of them

and let's not talk about how macOS puts to shame windows 11

thanks journalists for spreading their corporate propaganda, i wouldn't mind if it was to improve everyone's life, but spreading crappy bloatware.. nope

this article is on par with the verge "stadia bad but xcloud good" let me laugh for a moment.. lol

I switched to Apple when they dropped PPC for Intel. I've had just about every x86 Macbook they built, and have built a handful of high-spec hackintosh desktops.

When the M1 came out, I ran out and grabbed the cheapest one I could (8gb M1 Air) to play around with. After about 30 minutes, I decided to transfer my daily driver to this little passively cooled notebook. The next day I bought the M1 Mini to replace my overclocked hackintosh. My high-end x86 gear just couldn't compete with the low-end M1s.

I don't care what Intel and Nvidia say. I've used all their products. The M1 is black magic.

One thing I’ve noticed about a ton of benchmarks is a focus on throughput instead of latency. Apple has had some of the lowest latency computers around and the M1 has only improved the situation with QoS features on the CPU and such.
Looks like Intel's "AC performance" marketing worked on PC World's executive editor.
I’m not sure comparing benchmarks of yet to be released hardware is super useful, especially given intels recent track record. I’m more likely to believe nvidia’s shipping estimates there.

I’ll given them both props for not pulling that ludicrous bullshit graph that apple had in the original M1 announcement.

An interesting feature of SPEC that strongly disadvantages Intel is the a number of SPEC tests use ‘long double’ which every platform except Intel is always 64bit double running on modern execution units.

For Intel (and AMD) the ABI for 32bit, and 64bit on linux and macOS is to use 80 bit floats. That necessarily means x87. So no SIMD, a vastly slower execution engine, and higher precision. There is no case where x87 performance comes close to the primary floating point pipelines.

The M1 Max is pretty awesome and I can’t recall anything from Intel as of late that amazed me all that much. Intel made plenty of incremental improvements but nothing necessarily leaps ahead of their previous generations imo. The M1 and M1 Max were a refreshing change to years of underwhelming releases at crappy price points. For what I paid for my Intel-based MacPro my MacBook Pro M1 Max runs circles around it at a much lower price point.

If they really beat the M1 Max my question is at what cost in dollars and battery life.