I would like to see Apple take the flat memory mapping approach, where all processes live in a common memory space that is framed by the mmu to keep them from tripping over each other. With a 64-bit map (256Tb at present, 64Pb at ARM max), there should be enough room to fit them all together in the same space, and it would simplify IPC somewhat. This was experimented with in Australia a couple decades back, before 64-bit addressing was the word of the day.
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
64 PB may be enough, but 256 TB definitely isn't. A poorly configured memory allocator can easily exhaust 256 TB, until large allocations start failing due to address space fragmentation.
Because they are not going to rewrite external libraries from others they rely on for their code to work. People have been complaining to the developers for a while now also, but I think we know now they will never do it.
So now it is Windows only basically.
Maybe for simple "software" you find in the App Store, it is easily to recompile it to ARM or x86-64. But for other more complex software, it is not so easy due to external dependencies.
Sure, I get it if you are a video editor (and maybe a data analyst like in your case), the M1 or dropping of 32-bit is amazing. But on my side, it has broken many things and I sometimes have to resort to Windows.
Last edited:
Thanks for the example! True, if compiler is able to generate something like x86 lock xadd (which should be the case if you are using atomic exchange operations), you will have enough information to correctly transform the code. But this won't be the case if you are using load and store directly, as it simply generates mov for the x86 target, where ARM might need ldar/stlr. How does Windows deal with this?
That does sound rather unfortunate. I understand your annoyance. At the same time, I would still say that it's not Apple responsibility if others have difficulty designing and maintaining their software
You’re missing the point. At the enterprise end of the market, the market direction and market share is half a decade behind the tech.
Given the power efficiencies of ARM, there are huge sums of money to be saved in data centre power and cooling, AWS and Azure are looking to lean hard into their own proprietary ARM CPUs, and where Azure and Windows Server go, Windows on desktop will follow, especially once they perfect their x64 translation software.
NVIDIA are pretty much going all in on ARM and even if regulators block that acquisition the future of nvidia CPUs and SoCs is ARM, then you have Apple showing the world exactly what can be achieved with ARM and proprietary implementations of the ISA.
AMD are beating Intel from a technological standpoint on X64 and also have a robust ARM roadmap.
All of this also adds up to something else very important, the mindshare of ARM, which will reach a tipping point.
And then you have Intel, half a decade behind their X64 roadmap, lacking any really serious roadmap for ARM, completely devoid of innovation, rapidly losing their high achieving engineers to competitors and all of this means that the market right now in its current state doesn’t mean a damn thing for the health of Intel beyond this decade.
Aren't modern OSes all presenting a flat memory model to user processes using the MMU?
Wait, what? I thought that a "flat" memory model was one of the benefits of the "Unified Memory" architecture of the M1 SOC? So you have to be cognizant of *where* your app is writing in memory to prevent it from overwriting another apps' space? I must be misunderstanding what you meant? Thanks!
Rich S.
It's you that's missing the point, or rather thinking all PC markets are the same.
First, I agree those big data center want efficient chips and that's important to them, but Intel really hasn't been dominant in that market to begin with. Before it was mainframes and midranges, now I expect ARM-like processors already outnumber Intel processors in that space.
Big data centers and what they need has nothing to do with my job though. They sell services, and while I might buy those services I couldn't care less about it other than does it run what I need, and at a decent cost.
I work in a manufacturing plant, with no more than 50 PC's, and only about 25 user PC's. Efficiency means absolutely nothing to me. those 50 PC don't even rate a blip on our electricity usage. Just *spinning up*, not usng our A/C in the mill every days costs about $10,000. And we have lots of machines that use a lot more electricity than PC's!
What concerns me is -- does it run what I need, backwards compatibility is a MUST (capitals intentional). and does it run well enough to not piss off my users. That means it *has* to be either Intel or AMD, I don't really care which, but Intel definitely has the lions share of the market. People that use computers here, when buying their own home computers, buy the same type as here -- learning a new computer system isn't important to them.
This is where I'm saying intel has a huge lead, and it's all in the market numbers -- you can disagree, but you can't argue against the numbers, period.
For the future, my space changes a LOT slower because there's no ROI for it to change. Other segments don't matter to me.
No, macOS uses the Unix approach, where each app is given its own memory space that is essential a remapping of the same memory space as every other app. In a flat memory model, each app would occupy some unspecified location within the same map but would only have access to the region it is using (the MMU would limit what any individual process could see). In general, code is fundamentally written to work on data where ever it happens to be (absolute addressing is basically ancient history), go going to a flat mapping would be pretty easy and might simplify some aspects of memory management.
And "unified memory" simply means that the RAM is embedded in the SoC and used by all of the separate cores the work within the package (CPUs, GPUs, Neural logic, etc), as opposed to the GPU having its own private memory block.
Are you suggesting that Apple, were they to do this, would ship an OS with a poorly designed memory manager?
Each user process uses its own memory allocator. If the process keeps allocating and deallocating memory, its address space may eventually become too fragmented to make large memory allocations. That was very common with 32-bit processes, but I've seen it happen with 48-bit address spaces used by 64-bit processes as well. If each process has a separate address space, the effects of the fragmentation are limited to the process itself. If all processes share the same address space, a badly behaving process may make the entire system unusable.
The reason we don’t have flat mapping is it makes memory management MUCH more complicated. The app wants to have its memory region contiguous - item 235 on an array should be in the next memory location after 234. By using virtual memory, as far as the app is concerned, this is the case. If the memory is flat-mapped, then if my memory footprint for an app increases beyond its initial allocation, it will likely be given a non-contiguous extension. As someone else noted, this also causes fragmentation problems that only get worse as apps are quit, reduce their memory needs, etc.
It is far simpler and more efficient to give every app its own address space and to transparently map that to RAM. It also means all apps get the benefit of the full address space. If you flat map, then you need to divide the address space up so it can be shared by every running process. So a 64-bit program doesn’t get 2^64 words of address space, and, in fact, its available address space depends on what else might be running.
Yes absolutely - these are the rather academic cases it could not possibly deal with - i mentioned these in one of my earlier posts.
That is true. Fortunately for Windows users though, Microsoft itself has "difficulty designing and maintaining their software". They just announced that the next version of their flagship IDE Visual Studio 2022 will be 64bit and should be available as a preview version this summer.
Microsoft Windows XP Professional x64 Edition, was released on April 25, 2005 so it has taken a while.
Yeah, those kind of leads in tech can dissolve very quickly--little by little and then all at once. Think Blackberry.
That kind of malarkey almost put them out of business in the 90s.
Different market altogether. You're talking email vs. what industry runs on.
Business 101. This kind of stuff happens in every industry all the time. Tech is not the only thing that gets disrupted. However, Intel is clearly being disrupted by Apple Silicon right now. I can tell it’s bothering you, but there’s not much point trying to prove otherwise. The numbers are all there.
Of course business get disrupted all the time, but once a company has a near monopoly, it takes quite some time for that to change.
Now bothering me, that's the dumbest thing I've ever heard, you're talking to the guy that has multiple Macs, a lot of iPads and have been using the iPhone for a very long time. (my first one was a 3GS) I'm nearing retirement, and I know where I work isn't going to change platforms any time soon, it just wouldn't pay. I'm not worried at all, no matter what happens -- I just think it's WAY premature to be calling the death of Intel, lol! Kind of like how so many were optimistic that there would be a follow-on to the M1 this week -- totally outside Apple's typical MO.
And no, you can't prove squat with the current market, give it a few years for anything to make a difference.
100% agree.
Intel has enough of the market still to effectively coast for many years like they have already been doing, while AMD needs to be able to deliver something like 5x their current product in order to take the throne (last I checked, this may be off at this point), and offer more SKUs to fill niches they currently ignore. And now with foundries like TSMC at 100% capacity due to spike demand during the pandemic, and multiple industries (including automotive) impacted by an inability to source chips, Intel owning their own fabs is suddenly a benefit and allows them to meet demand that AMD simply cannot in their wildest dreams do right now. And we’re seeing folks in the industry expecting the shortage to potentially last for a couple years, which will hamstring pretty much everyone, but Intel the least.
While Apple is shrewd in their buying up of all of TSMC’s 5nm capacity in the short term, they are also beholden to the same supply chain issues that AMD is, and has no ability to buy more capacity at the moment.
While I do think Intel is proving it’s in a rut, and is currently the weakest between Apple/AMD/Intel from an engineering perspective, getting that to translate to shifts in sales is a whole ‘nother matter. Intel has a lot of runway to get their act together. AMD hasn’t yet proven to me that their QA/QC is good enough to avoid facepalm-worthy launch bugs.
It's pretty hard to "effectively coast" in the PC space. They better not do it too much longer. They're already about 8 years late to this party. Life comes at you fast.
Why are you blaming Apple for the developers of virus TI not updating their software? That would be like blaming GM for a gas shortage...
I agree but less than 100%
. As you said, the chip shortage, in PC CPUs anyway, is due to hugely increased demand rather than a loss of supply. That means AMD is selling more chips than ever before and the gap between them and Intel should be widening rather than shrinking. The chips that are hardest to find from AMD are 5900s and up, which is a sector Intel isn’t even competing in and the supply of 5800s seems to have stabilized. And Intel’s chip that competes against the 5800 had to get its price slashed the first week in the market and while not a terrible chip the recommendation is still the 5800 despite the price reduction on the Intel chip. There was a report that in some quarter recently that Intel had clawed back a tiny bit of market share for the first time in awhile but that seems to have been a blip. We’ll see by the end of of the year. To be fair losing customers to AMD is a lot better than losing them to ARM and many have suggested that AMD’s success could be a lifeline for Intel in the long run. But it’s still *could* be not *will* be and they are losing customers to *someone* at almost every end of their CPU business now. The market growing helps hide this but it’s still there and causing executives and investors at Intel worry. And, as stated previously, collapses tend to be very slow then very very fast.
Now Intel is huge, x86 has a lot of backwards compatibility going for it, and AMD and ARM chip makers have their own business issues which could come to Intel’s rescue, but I’d rather be the CEOs of those companies than Intel’s - not that I’m qualified to be so for any. ?
EDIT: After posting saw this highly relevant thread on Twitter:
Basically Intel’s margins are waaaay down
Last edited:
I agree but less than 100%
Intel is still 80% of the laptop market, and 78% overall. Despite the erosion.
The point I was making is with Intel's current position, things still have to shift quite a bit to put Intel itself in jeopardy, and deprive them of the money to fight back or shove them out of the CPU market. And my point about the shortage is more that it's unlikely AMD will be able to buy the capacity they'd need to even be able to supply Intel's current customers when there's no capacity to buy, and it's not likely there will be any relief on fab demand for a couple of years. That gives Intel a couple cycles, and even gives Intel an opportunity if they can actually move on it.
But keep in mind that with the rise of laptops, and shipment numbers of desktops being a third that of laptops, you can't dethrone Intel using desktop chips. AMD could stand to be a bit more aggressive with their laptop chips, IMO. However, they are being plenty aggressive in the server space at least.
Why do you think Intel spent what 10nm fab capacity they had on Laptop chips?
My argument isn't that Intel is somehow in a good position. They aren't. It's more that there's no path in the short term to be able to supply CPU demand in the PC space without them representing a majority share of those CPUs. If Intel were to implode in the next couple years, it would likely be a bad thing for the industry in the short term.
Oh, I agree, as I'd rather not be Intel's CEO at the moment.
I happen to have upgraded from a 2020 Ice Lake MBP (16GB, 512GB) to an M1 Air with the same RAM and SSD specifications, a 7-core GPU, and while I agree, there is still something distinctly “heavy” about the system’s UI as opposed to Windows. It’s odd that iOS manages to perform well enough on old, RAM handicapped hardware (well that too is an issue with apple hardware but we’ll save it for another day) and yet Mac OS is only smooth on the absolute most exceptional, efficient mobile chips the industry has to date.
Is it too much to ask that Apple have something similar to Facebook’s “2G” days albeit for some kind of Macbook SE’s running cut-down M1’s, to better grok the performance of the software stack absent premium chipsets? It would probably aid in easing up the gluttonous, wasteful animations and transparency ******** that they tend to be addicted to like crack cocaine.