I see lots of people saying "If only $company had an ARM device", and I don't understand it. My experience of ARM device is slow laggy janky products - from competing smartphones, tablets, router and network device management interfaces, year on year, the pairing is always (arm + fucking slow).
It's only Apple who have managed to get "desktop class" performance, and at this point I'm more willing to attribute that to Apple doing black magic than to ARM being inherently superior.
Microsoft have piles of money, thousands of developers, they make hardware, they've been working with ARM since the likes of the Compaq IPAQ PDA and Windows CE 21 years ago, and the Surface RT in 2012, their ARM devices are nothing to speak of. Google, industry gorilla of tech with the finances to match - see the famous Gruber/DaringFireball about iPhones far ahead in single core JavaScript performance year after year after year. Tech giants like Samsung, Qualcomm, Sony, who have been designing and building chips for decades, with their own drivers, firmware, and customized Android builds, and building their own flagship products.
Apple whomps in from "nowhere" with the fastest ARM device anyone has ever seen, and the critics response is "anyone could do that if they wanted to".
Then why don't they want to?
[Edit: Consider also that Apple spent $278M on acquiring PA Semi chip design company in 2008, and $600M on Dialog Semiconductor in 2018, and Microsoft spent $2.5BN on Minecraft. I say these companies "are tech giants with finances to match", but it's not like Apple had to lean hard into their hundreds of billions cash pile to do what they've done. Instead they had to do something that appears to be "black magic" (i.e. desire + leadership + execution + long term planning + ???)].
> Apple whomps in from "nowhere" with the fastest ARM device anyone has ever seen
Not really from nowhere; they've already had the fastest (non-datacenter) ARM chips anyone has ever seen, fairly consistently, in their phones, for years.
The first A-series chip they officially released and named was the A4 in March 2010. That is 11 years ago. That is the beginning of M1. And the reason is originally in this amazing Ars Technica article [1]. Apple has always wanted to fully control the entire stack.
Any competitors in the same space lack either the vision, or the products, or the money to do the same thing. In the case of Microsoft and Google, their internal politics don't allow them to do the same thing (Google couldn't care less about hardware because the Web is where Google's money at; MS cares about hardware but traditionally they have relied on throwing money at partners to make them come up with solutions).
[1] Owning the stack: The legal war to control the smartphone platform, https://arstechnica.com/tech-policy/2011/09/owning-the-stack...
One of the ways we already know that beefy ARM processors can work well (and with better power efficiency than Intel CPUs) for heavier workloads/at higher scale than mobile or embedded devices is that it's already being done in the server space. Is scaling ARM processors up for the desktop radically different? (If so, I'd love to learn a bit about how )
The problem that remains for anyone who wants to sell PCs with something other than x86 is that desktop users want their apps, and for Windows users those apps are distributed as x86 binaries, and users aren't in control of that (especially for unmaintained or legacy software). Software publishers won't build or optimize for a new architecture if they suspect it's just going to be a flash in the pan, and this creates a chicken-and-egg problem.
But Apple can just declare that their new architecture is the only option for you if you're buying a new Mac, and a ton of their users will come with them no matter what. App publishers will just deal with it, knowing that they have to play along if they want their application to have a first-class experience on Mac. This is a huge deal, and not a position any particular PC manufacturer could hope to be in!
I'm sure Apple has done lots of technically impressive stuff that I'm not even competent to really appreciate in order to make this architecture change happen. But other vendors are more capable of technically impressive stuff than they are of forcing adoption for enough users to reach critical mass to get people who sell/distribute proprietary software to come along with them.
Apple is in a position where it can move swiftly and decisively. They have an army of software engineers who can port Apple's OS and associated Apps to a new architecture. They can force their entire ecosystem to move to a new architecture, and are emboldened by the fact that they've done it twice in the past (the first time they were more hesitant).
If Samsung wanted to pull down the latest Qualcomm ARM chip, slap a Macbook-sized heatsink on it, and go head-to-head with Wintel and Apple, they would fail. What are you gonna run? Linux? Some abortive version of Windows on ARM that doesn't support the vast number of x86 apps?
Even ChromeOS is missing barebones necessities like Photoshop, so professionals won't touch it. So the ChromeOS market is doomed (at present) to have "thin & light" and "long battery life" as it's only selling point.
I think we'll see M1-style ARM devices come out of the woodwork in 2021-2022, running ARM Windows and Linux and ChromeOS, simply to attempt to copy Apple. But I don't think they'll be as successful, simply because the apps aren't there, and no one can force developers to port to the new architecture as Apple can.
Thing about your comment is "Who would buy it" assumes it's basically the same and people are indifferent. My griping is assuming ARM is generally worse. But if we take this alternate world seriously then the reason for ARM is superior performance, so the answer to "who would buy it" and "why would developers switch" is driven by that - like when Apple was falling behind with PowerPC - it wouldn't be Microsoft pushing it on an unwanting market, it would be a demanding market pulling it with demand.
If Samsung could push a better-than-Intel chip, would Microsoft not want a piece of it?
If Samsung Chromebooks were suddenly more powerful than an Intel i7, would developers not sit up and take notice?
My contention is they can't, because ARM isn't that different, and it's Apple's black magic that is the real difference.
I remember the jump from the 486 chips to the Pentium being that dramatic. Heck, 486DX2-66 to P100 brought Linux kernel compile times from 45 minutes down to like 5 minutes.
Back then, I think a big part of these jumps wasn't just CPU design. I think a lot of it was that those CPU design changes were often accompanied by massive improvements to overall system architecture. So the I/O buses, memory, peripherals, all of it got a lot faster as part of the same upgrade.
Last time it felt like there was a jump this big was around 2005-2006 when the Core 2 Duo started becoming common.
Is it really a gamechanger for surfing, coding, videos, etc?
Curious because I might grab the base mac mini 8gb for music production, surfing, Java coding etc.
Also you said music production so verify before hand that your DAW, etc of choice will work fine. I use Reason and the visuals in the DAW itself are still laggy, but it isn't M1 native yet.
Other things like the fact that World of Warcraft runs at medium settings on a 1440P monitor feel kinda magical given I can barely feel any heat leaving the M1 Mini.
In the end I returned my M1 Mini (8gb/256gb) because my personal experience still felt half baked at this time. Once we hit the next generation a lot more software will have caught up.
Apple's memory and storage options tend to always be on the disappointing side.
The impressive thing about the M1 isn't just raw performance but performance / watt. There are obviously faster high-end many-core x86 chips, but the power use difference is wider than the speed difference. The M1 destroys mid-range Intel and AMD offerings at a fraction of the power consumption.
Check out this madness:
https://browser.geekbench.com/mac-benchmarks
The MacBook Air trounces the top-end Mac Pro on single core performance. No this is not the top-end Intel you can get, nor is it as fast as AMD, but it's a MacBook Air and consumes a small fraction of the power.
Again... the power efficiency is the thing that really blows me away. 5nm accounts for only some of that.
On multi-core the same MacBook Air is just below recent Mac Pro models on total computational throughput. To beat the MacBook Air you have to go up to newer-generation Xeon chips with 8 full-speed cores. The Xeon is branded as a server chip and uses as much power as probably eight to ten M1s.
It's just insane. If Apple puts, say, 16 full-speed core in one of these chips it's absolutely over for all other vendors. Speculation is that the 16" Pro will end up getting 8 performance cores and 4 low-power cores. I wouldn't be surprised if it ends up beating the top-end Intel Mac Pro.
I'm not sure that they can easily add more cores to the M1. If they can, that would also raise the TDP. And adding more cores does not increase performance linearly, even if the architecture is made for it.
Also keep in mind that comparing it to 14nm Intel processors with an architecture from a decade ago is a fair comparison when comparing M1 vs prior Apple products, but is not a fair comparison when comparing it against what x86 can do in general. You would need a modern x86 processor targeting the same watt usage to have a completely valid comparison. As discussed in the thread above.
So yeah, it's a great processor for its target. And it likely will also be a very strong processor when scaled up - it bodes well for the future of ARM. But don't extrapolate the performance too linearly, that is likely to be misleading.
I expected the M1 to be like a much lower power and maybe a bit faster version of this chip, and was really blown away by it being far beyond that.
Heck, he might have used an old dual core with an HDD from all we know. Of course the new M1 models would make a big difference then. So would all other modern laptops.
But in that context of x86 and ARM Zen3 is absolutely the only sound comparison.
I would have some sympathy with that - but that's not the same as discounting the parent's perspective because Apple somehow weren't using "modern processors" previously - because that simply isn't true.
And I still think it's very valid to not forget how old hardware used for subjective comparisons will be. E.g. the old Macbook air sold very well all that time.
It's likely the M1 is better in that category, but it does not look like a huge difference if you factor in the higher performance you get from the Ryzen 9 5980HS.
Multi core better but still lower but not too surprised as it's got 8 large cores vs big.LITTLE 4/4 for M1 - not clear when those cores will start throttling down.
Suspect single core performance is what drives apparent responsiveness so I'd say not too surprising that the M1 gets rave reviews from users.
It’s the first time in a over a decade I’ve felt a revolutionary step forward in a chip and the only time it has ever come with less power.
Process node is a factor but I find it hard to believe it’s the only factor. I know enough about the X86 legacy tax to know that is not the case.
Apple didn’t do black magic. They just took a cleaner superior CPU architecture and made a desktop class muscular chip with it. Other ARM manufacturers could equal or exceed the M1 if they wanted. X86 has been delivered a death sentence.
IMHO the greatest threat to AMD is ARM not Intel, and vice versa.