To be clear on [Chromebooks wiping the floor] I think W10 is fantastic for developers, and I'll probably pick up another PC myself, but I could see a Chromebook Android-App effort hinging on utilizing a similar architecture that works out fairly well for most consumers, particularly the less technically-inclined. Nvidia seems interested in them as future gaming devices too, so that's not bad.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
i hope intel comes back to beat the m1
- Thread starter Ratsaremyfreinds
- Start date
- Sort by reaction score
You guys have this weird mythical belief in the supremacy of Apple silicon, literally take away the chip shortage (so TSMC 5NM instead of Samsung 5NM) and give X1 cores in an 888 a *bit* more cache as ARM themselves had intended in designing the core, and an X1 core is right there with an A13.
And that's ARM reference cores. They [Qualcomm] just acquired Nuvia, with a tape-out by 2022, so I'm betting we'll see lots of cope here come 2023-2024.
I fully admire the M1 owning X86, but the weird "it's apple not arm sh*t" is going to age poorly. Sure, Apple will probably still have some leads if only because they have a blank check to work with on teh production end (see using larger cores, big hit to yields owing to the defect scaling which isn't linear obviously) but I mean holy sh*t. I could easily see something +- 20% Apple's performance and likewise for efficiency from Qualcomm in 2025. And that's completely within the timescales being discussed here IMO
Last edited:
As for why I think ARM laptops will succeed: It really comes down to the ability to target similar or lower price points with similar performance but notably improved efficiency or aggregate core count.
chromeunboxed.com
the kind of efficiency I think will be readily-available on $500-750 craptops will be a gimme to Qualcomm et. al, unless Intel wants to lower margins, and lol
This full performance analysis of the Snapdragon 7c for Chromebooks is impressive
We're all patiently - or not so patiently - awaiting the full arrival of the first Qualcomm Snapdragon 7c Chromebooks into the market. The Acer Chromebook Spin 513 is already hitting some regions, but wide roll-out is still likely a few weeks away at this point. Additionally, detachable tablets...
the kind of efficiency I think will be readily-available on $500-750 craptops will be a gimme to Qualcomm et. al, unless Intel wants to lower margins, and lol
Sure, that's a crappy Snapdragon perf. wise right now, but again, Qualcomm has had this on the backburner. Just wait.
It's not mythical belief. It's been happening in the mobile space for about a decade now. And as others have pointed out, it's not just the chip design, it's the OS being fully optimized for the new hardware, it's the dev tools being ready to go, it's the vertical integration of product managers imagining a new feature and the OS and silicon teams working together to make it happen. Johny's team doesn't have to worry about meeting the needs of any other customer but Apple, Qualcomm doesn't have that luxury.
Apple has fully committed to the transition. MS has dipped a toe in the water, but their effort was somewhat lacking. The Surface Pro X looks great, but it's expensive and it doesn't perform well. At launch, there were few native apps, and x86 emulation was 32-bit only. They did tweak the Qualcomm chip a bit. They could have done what Apple does: they could have pushed their own software division to be fully arm64 by launch. At least Office. They have a game division, they could have tasked them with creating or porting a few titles for launch. They could have invited some devs of popular apps, especially iPad apps, in early to bring their titles to the new device. Then, when devices ship, new users would have been able to go to the Windows Store, and see a nice selection of arm64 native apps ready for them to download and use on their shiny new device. They could have, but didn't. So an expensive device got tepid reviews and MS barely mentions it.
Even if another vendor manages to offer up competitive chips, if they can't undercut Intel and AMD on price, they might not get much in the way of sales. It needs to be a concerted effort between the chip designer, the OS vendor, and software developers to pull it off. It may not be imminently profitable. Apple plays the long game and is willing to invest in projects that may not be profitable for some time. MS has the resources, they just need to commit. Google has the resources, but won't commit unless there is a way to tie it back to their core ad business. Even if the Nuvia folks can help Qualcomm be more competitive, it's just one piece of the puzzle.
The continued “just wait for Nuvia!” stuff cracks me up. Nuvia produced not a single chip. It’s 3 guys out of a hundred from Apple. There’s a lawsuit. etc.
Take the chief architect and two senior designers at random from any CPU team I ever worked on and send them to qualcomm and you know what they’d produce? Maybe a slightly better qualcomm chip.
I mean, talk about weird mythical beliefs. Apple has, year after year, produced 20% improvements - a slope much better than the industry as a whole. Nuvia produced nothing, but “just you wait and see! Nuvia FTW!”
Forget Nuvia, we have the X1 reference core and with a bit more cache (which is obviously going to happen for laptops) and TSMC 5NM (so no chip shortage - hardly the fault of the engineers you call into question) and the X1 is right there with the A13. Apple's firestorm cores are just flat out larger/wider/have more cache, lack X86 baggage, and have TSMC's finest producing them.
Qualcomm can at least hit two of those, if not all three down the line. What is it you really think is so special about muh M1? I'm sure there exists some proprietary bullsh*t streamlined into the OS wrt paging and memory, maybe.
I mean, at some point we get to Everest regression and "he's faster because he has a longer stride". Apple deserves credit, but I don't think that what they have will remain ahead for as long as the fruit acolytes in these among other fora seem to believe.
I would love to forget Nuvia, but everyone keeps bringing them up!
As for your theory re: X1, it’s easy to say on paper, but until I see silicon that proves the point I don’t give it much weight. It’s easy to produce paper designs, and hard to produce working silicon. Apple has, every year for the last decade, averaged more than 20% improvement, year after year. Nobody else has done that. And they do that by producing a physical design that is 20% better than qualcomm (really the only current design team that does physical design about as well as apple is AMD). M1 can handle a tremendous number of in-flight instructions, and issue a lot at once, indicating their micro architecture is also either very good, or is precisely tuned to MacOS.
Qualcomm uses an ASIC design methodology. They just cannot compete with Apple’s hand-crafted layouts. And if what Nuvia brings to Qualcomm is “we should use the AMD/Apple-style design methodology” then it will be a few years before you see any benefit from that, because, as the guy who was in charge of creating AMD’s methodology, I can tell you it is going to take a dedicated team a year to get it in shape to use.
Fair enough, but surely you realize not all of your interlocutors mentioning Nuvia are of the disposition that lends toward hyperbole, no? E.g. I’d suggested with proper investment and a similar process node, Qualcomm would be competitive with Apple, but maybe we’ll have to see, certainly the track record isn’t great.
That said I maintain that various ARM vendors could eat into Intel’s market share on $400-800 craptops.
As for AMD, I keep seeing this line about TSMC 5NM as the only differentiator between it (mobile Zen 3 chips like the Ryzen 7 5800) and the M1 - as though it is only Intel’s designs that are behind on PPW, not AMD. This [trope on Reddit and HN] just seems absurd, the TDP differential would remain and the M1 would still wipe the floor, even without the GF IO Die or whatever.
Though OTOH, with an updated architecture (Zen 4) and TSMC 4/5NM, I'd be excited to see what AMD has, even if the PPW lags a bit
Last edited:
Related but @ c maier why do you think AMD isn’t going for a heterogeneous architecture yet? Windows scheduler & the myriad of already small*er* Zen cores in Ryzens?
I dunno
After getting my Ph.D. and spending a decade and a half designing microprocessors, I went back to school and changed careers, so I don’t know why their roadmap is what it is now.And I agree that sticking any other chip on 5nm isn’t going to magically make them P/W competitive with M1. Certainly no x86 chip can achieve that.
Honestly, the best way for anyone to come close would be to jettison 32/16/8 bit compatibility and remove a huge chunk of the instruction set. The 64-bit extensions are far easier to decode, and you might be able to remove a pipe stage or two at the decode end, and issue more than four instructions to the ALUs per cycle.
Not sure what you mean. SQ1/2 is still up to today the fastest 7W TDP SoC, which you can get in any Windows machine. So it is very competitive. Its competition are Lakefield SoCs and Alder Lake SoCs.
Saying the SQ1/2 is not competitive is like saying the M1 is not either, because Threadripper is much faster.
Would i like to have a more advanced core than A76 in my Surface Pro X? Certainly, but on the other hand i have the machine already for 1.5 years and there is still nothing better yet.
Consider Apple's focus on ObjC/Swift. In regular C or C++, the code uses simple JSRs/BLs (JSR does 3 things, BL does 2, and maybe the third thing later in the subroutine, if necessary). ObjC uses a lot of method calls where other code uses the basic calls. Method calls in ObjC tend to be slightly more costly than basic calls, but there are a lot of them, some very small (e.g., getting or setting an instance variable). If AS has hardware logic that reduces this performance cost, that would be big. Huge. ObjC/Swift is very flexible and convenient to work in, so just this tiny bit of software integration would be profoundly non-trivial.
That's very specific. Sure, it's the best you can currently get in a Windows machine. As long as you have that qualifier, then it's a true statement. Of course, when you compare it to M1, it's a different story. Either the new iPad Pro or the MBA will outperform it easily.
The other points stand. It took way too long for MS to have native software for it. If the company were truly committed to WoA, then Office, Edge, and other core apps should have been native at launch. Factor in WoA x86 emulation compared to Rosetta 2, and it's even more embarrassing for MS. They should have aligned the entire company around the decision to support ARM to give it the best possible chance of success. Microsoft has become very risk averse, and it shows.
I am absolutely certain about this - yet we are talking about machine, which is about 2 years older than the upcoming iPad Pro.
You really like to compare apples and oranges. Rosetta 2 is x64 emulation only - yet you compare it to Windows x86 emulation. If you would compare it to Windows x64 emulation, you would notice that it is very comparable.
PS. Office and Edge both were native at launch.
Last edited:
My point is that the performance hit running x86 Mac software on M1 under Rosetta 2 is far less than running x86 Windows software under SQX.
Apple has handled the transition better in every possible way. But hey, at least a 64-bit version of VS is on the way.
Rosetta 2 doesn't run Mac 32-bit software. (nor does Big Sur on Intel.)
It needs to be in the discussion because Windows still has a lot of 32-bit software available, and sometimes its even preferable for compatibility with older custom software.
You have no point, as Rosetta 2 is not emulating x86 at all - it is x64 only.
It is perhaps because Microsoft did not handle any transition...
Last edited:
And it doesn't matter to me at all, they could take it down to 1 atom wide level and the first thing I'd ask is does it run my software...
It is x86-64 only because Apple did the smart thing and ditched 32-bit compatibility ahead of time. Apple makes choices that allow it to move the technology ahead (at the cost of compatibility with older software). You can argue whether it’s right or wrong, but Rosetta 2 being so performant follows as a consequence of such decisions.
Nope, Apple ditched 32-bit and went fully 64-bit. Most app developers recompiled their apps and they worked.
Apple isn't interested in supporting niche markets that rely on legacy software. If that is your use case, then you should go with Windows or Linux. Apple has made it very clear for over a decade now that they are willing to provide a bridge when technologies change, but they will not support legacy software in perpetuity.
MS should go back to two versions of Windows: one that maintains backward compatibility and receives little more than security updates, and another consumer oriented version that will change and adapt as technology moves forward. The memo went out long ago that the age of releasing an application and then just leaving it alone is long past. It's time for developers to get with the program.
I just stated facts - i did not argue at all, why something is as it is - or if something is right or wrong
One of these facts was, that x64 emulation is much faster than x86 emulation - as we can clearly see with Windows on ARM, because we have both emulation technologies available there and can compare. So if we restrict ourself to x64 emulation, then x64 emulation under Windows and Rosetta 2 under MacOS are very comparable performance wise. There is a bigger hit for x86 emulation though.
I still find it somewhat disturbing, that you are answering to my post and not to the guy with all the wrong facts - but yeah its an Apple forum after all...
Last edited:
I just stated facts - i did not argue at all, why something is as it is - or if something is right or wrong
One of these facts was, that x64 emulation is much faster than x86 emulation - as we can clearly see with Windows on ARM, because we have both emulation technologies available there and can compare. So if we restrict ourself to x64 emulation, then x64 emulation under Windows and Rosetta 2 under MacOS are very comparable performance wise.
I still find it somewhat disturbing, that you are answering to my post and not to the guy with all the wrong facts - but yeah its an Apple forum after all...
Yes, x86-64 emulation should be faster for various reasons than x86 emulation. Wasn’t trying to say you were wrong about anything. Just replying to you because you were the most recent post on that thread.