16 ms·

It's worth pointing out how extremely far ahead Apple seems to be in terms of CPU power and efficiency. It tends to go under the radar a bit because it's not ea

• fsflover 7y ago
  > because it's not easy to run your own software on an iPhone Then why do I need all this performance? Perhaps to run js on shitty websites. Well, maybe for gaming.
  • godzillabrennus 7y ago
    Not that these Apple chips will run x86 or x64 code. Unless they bother with a Rosetta 2.0 then gaming is out of the question if Apple makes this jump.
    • MrBuddyCasino 7y ago
      Aren't apps already submitted as bitcode? They don't have to emulate, just recompile.
      • nagisa 7y ago
        LLVM Bitcode is still architecture specific. That is, bitcode generated targetting x86 is not compatible with bitcode targetting ARM is not compatible with bitcode targetting aarch. The reason they are distributed as bitcode is to enable additional optimisations based on what exact chip will run the code (i.e. -march=native equivalent)
        • pascoej 7y ago
          https://www.highcaffeinecontent.com/blog/20190518-Translating-an-ARM-iOS-App-to-Intel-macOS-Using-Bitcode https://www.highcaffeinecontent.com/blog/20190518-Translatin... This article appears to successfully statically translate arm64 binary into an x86_64 binary using bitcode.
        • imtringued 7y ago
          Using LLVM IR merely solves the problem of being able to compile to a specific instruction. You still need a compatibility wrapper for all the platform specific APIs like wine or Windows on Windows. If Apple puts in the effort then it might work out.
      • Wowfunhappy 7y ago
        On Mac? I don’t think so, no, because you can move App Store apps to different machines on USB keys. Also, a considerable number of Mac apps don’t use the App Store for distribution.
      • ken 7y ago
        No, the Mac compiler target has never supported bitcode. And LLVM bitcode doesn’t work like that anyway.
    • rsynnott 7y ago
  • TallGuyShort 7y ago
    A lot of people who don't need it already buy it for the image and social status. Apple doing this for workstation-type machines too just makes even more sense.
  • rooam-dev 7y ago
    A more future proof too.
  • rsynnott 7y ago
    The extreme slowness of the modern web honestly is a reason. I do a lot of my personal web browsing on an elderly (nearly 6 year old) iPad Air 2. For everything but the web, and for old websites, it's fine. For the "100MB of react crap" type of website, it's getting pretty painful, tho.
    • rvense 7y ago
      My personal laptop is a Lenovo X200s from 2008 or 2009 with 8GB RAM. I use it to write words, read PDFs, do some software development (most often C and "vanilla" web stuff), design electronics in Kicad, and casually browse the web... and I'd be happy with it as my only computing device forever, except for some websites it's starting to be a little slow, especially a well-known site that's meant for exchanging messages under 500 characters. There's some irony there.
      • KozmoNau7 7y ago
        Mine is an X220s, so slightly newer than yours, but still a modest i3, upgraded to 8GB RAM, 128GB SSD and dual-band WLAN. With a new 9-cell battery, I get ~6 hours of battery life, and it's fine for general browsing with a decent number of open tabs, even a few games (Darkest Dungeon and some emulated SNES games). As long as it keeps ticking and I can get whichever spare parts I need on eBay or something, I'm not going to replace it anytime soon. The only reason I have a more powerful desktop PC (still ~2011 vintage) and don't just use the X220s in a dock, is that it it struggles with a 1440p external monitor (full HD is fine, though) and I sometimes like to play more graphically intense games.
        • DeathArrow 7y ago
          >As long as it keeps ticking and I can get whichever spare parts I need on eBay or something, I'm not going to replace it anytime soon. Good luck with that as the Javascript crowd are in competition to slow the web as much as they can.
          • KozmoNau7 6y ago
  • mumblemumble 7y ago
    Alternatively, when you don't need to be running full throttle, you get better battery life. Which any consumer will appreciate. And is also perhaps an environmental win, if it allows them to get away with smaller batteries. That said, the experience of trying to browse a local newspaper's website with NoScript turned off on my 2015 MacBook Pro tells me that, yes, JS on shitty websites is a problem. And not one that most people would find to be particularly avoidable. Heck, even GMail is getting to be noticeably slow on that computer.
  • papermachete 7y ago
    Or run virtual machines like Android users do.
• rsynnott 7y ago
  > It's worth pointing out how extremely far ahead Apple seems to be in terms of CPU power and efficiency. Honestly, I'm not convinced they're THAT far ahead. They don't have a lot of the legacy baggage Intel has to contend with, and they're the only company making high end ARM chips (besides Amazon and a few other weird server implementations), but being able to match big core i7s in some benchmarks single threaded is to a large extent something that Intel's own low power chips can also do, at least burstily. There are a lot of challenges to big many-cored chips beyond single-core performance, and we really don't know where they are with that yet, as there are no publicly-available examples of Apple desktop chips.
  • wlesieutre 7y ago
    >There are a lot of challenges to big many-cored chips beyond single-core performance, and we really don't know where they are with that yet, as there are no publicly-available examples of Apple desktop chips. The A12X from 2018's iPad Pro is the sort of chip I'd expect to see in an ARM laptop, and its multicore scores are similar to the top end of 2018 Macbook Pros. 2020's iPad revision didn't get much in the way of processor improvements (just one more GPU core), so the new 16" MBP has pulled ahead with an 8-core i9, but when we get a new iPad based on an A13X or A14X I expect it to be back in that range again. And these are in thin fanless tablets. With a proper cooling system, there's got to be some extra juice to be squeezed out of them.
    • PascLeRasc 7y ago
      I wonder if anyone's ever done A-series chip "overclocking", or at least manual overvolting with active cooling. It'd be cool to see what kind of performance increase might be available.
  • ebg13 7y ago
    > and they're the only company making high end ARM chips That's, if you'll pardon the pun, an Apples to Oranges comparison. Apple isn't making "high end ARM chips" either if your comparator is powerful servers. You need to look at what Apple is doing within their power envelope and compare that to what everyone else is doing within _their_ power envelopes. The A13 Bionic is an uncooled 6W TDP chip blowing past 95W base TDP chips that require hefty active cooling.
• mtgx 7y ago
  Not to mention that with the latest few generations of CPUs Intel has also increased burst/decreased base clock while promising unrealistic power draws. In other words, Intel's CPUs are mostly about "burst" performance these days, too. They don't get anywhere the promised peak performance for any significant amount of time. https://www.anandtech.com/show/13544/why-intel-processors-draw-more-power-than-expected-tdp-turbo https://www.anandtech.com/show/13544/why-intel-processors-dr...
  • paulmd 7y ago
    Oh boy, here comes the AMD peanut gallery. The conclusion of your article is precisely about how motherboard manufacturers don't obey the official/nominal behavior and how that makes it irrelevant: > Any modern BIOS system, particularly from the major motherboard vendors, will have options to set power limits (long power limit, short power limit) and power duration. In most cases, at default settings, the user won’t know what these are set to because it will just say ‘Auto’, which is a codeword for ‘we know what we want to set it as, don’t worry about it’. The vendors will have the values stored in memory and use them, but all the user will see is ‘Auto’. This lets them set PL2 to 4096W and Tau to something very large, such as 65535, or -1 (infinity, depending on the BIOS setup). This means the CPU will run in its turbo modes all day and all week, just as long as it doesn’t hit thermal limits. Intel desktop processors will sustain boosts for an arbitrary amount of time. Yes, they will exceed the nominal TDP while doing so, so do AMD processors (AMD's version of "PL2", which they call the "PPT limit", allows power consumption up to 30% higher than the nominal TDP while boosting, and there is no official limit to how long this state may occur). These limits are of course observed much more strictly on laptops since power/thermal constraints actually make a real-world difference there. But overall, Ice Lake perf/watt is competitive with Renoir and its IPC and per-core performance is actually higher than Renoir. You just get fewer cores.
• mikorym 7y ago
  While I don't know much about phones and suppose you are correct there, I have been disappointed with the hardware that they pushed to their macs the last 5 or so years. For instance, why did they ship newer macs without upgrading to the newest Intel chips? I guess you can solve anything if you do it yourself, but that was perplexing for me that they would do that. I am not an expert on this, but this happened some time between 2016–2018. Edit: I stuck with my upgraded 2013 model and even today it's fast and good for the job. Even my 2009 mac is still running. So I am not saying they can't pull it off, I am just wondering whether they give enough attention to their non-iPhone products.
  • koheripbal 7y ago
    "newest" might be unfair. I imagine there's a lot of product life-cycle / driver testing to ensure high quality. Whereas other PC shops typically will just throw PCs together and assume they'll fix issues found in patches. That's not going to fly for "just works" Apple product image - and why it's positioned and priced as a "luxury" product.
  • mikorym 7y ago
    Since I was downvoted, here are some discussions on what happened cerca 2016–2018: https://news.ycombinator.com/item?id=12816474 https://news.ycombinator.com/item?id=12816474 https://news.ycombinator.com/item?id=16130872 https://news.ycombinator.com/item?id=16130872 https://news.ycombinator.com/item?id=13364583 https://news.ycombinator.com/item?id=13364583 Don't bother with the links, the comments are more informative. In some of the comments people explain why it's Intel's fault and not Apple, in some comments people explain processor speeds. A lot of it is about the keyboard that they changed, but that is not related to this. My point was more that I think Apple should put the amount of effort into their laptops that they did in 2009, whether or not they do is subjective until we see their new processors.
• walki 7y ago
  > It's worth pointing out how extremely far ahead Apple seems to be in terms of CPU power... I agree that Apple's ARM CPUs are very competitive on simple scalar instructions and memory latency/bandwidth. However x86/x64 CPUs have up to 512 bit wide vector instructions and many programs use vector instructions somewhere deep down in the stack. I guess that the first generation of Apple ARM64 CPUs will offer only ARM NEON vector instructions which are 128 bit wide and honestly a little pathetic at this point in time. But on the other hand I am very excited about this new competition for x86 CPUs and I will for sure buy once of these new Macs in order to optimize my software for ARM64.
  • miohtama 7y ago
    For certain professions like media editing vector instructions help. But for your average Facebook / Netflix / Microsoft Word user, a kind of user that 95% users are, there are less benefits on vector instructions.
    • cozzyd 7y ago
      Sure but the same is true about performance in general.
      • nicoburns 7y ago
        That's not really true. Single-threaded scalar performance is still super important for the everyday responsiveness of laptop/desktop systems. Especially for applications like web browsing which run JavaScript.
        • gowld 7y ago
          Your UI is slow because of IO and RAM and O(n^2) code, not CPU. Look at your activity monitor.
    • vardump 7y ago
      Are you saying Facebook, Netflix and Microsoft Word don't require media processing? Pretty sure you'd see plenty of SIMD instructions being executed in libraries called by those applications.
• einpoklum 7y ago
  1. ARM-based CPUs have for years (decades) been much more power-efficient than Intel and AMD CPUs. This is regardless of Apple, and is not really about being ahead - it a different path in the design space. 2. If Apple were "far ahead" in terms of performance in general, they would probably have been using these chips in products which aren't smartphones.
  • itchynosedev 7y ago
    wrt point 2. I always assumed that it's not only chip performance or efficiency, but ecosystem and software compatibility that's a larger issue. It's not trivial to port x86 software to ARM, or even run an energy efficient emulator. Is that correct?
    • cozzyd 7y ago
      In most cases it's fine since Linux distros for arm have existed for a while. Getting proprietary vendors to support it is a different question.
    • blihp 7y ago
      I believe you are correct. I suspect a big part of Apple's increased popularity since moving to x86 has been the ability to run an alternate OS either via dual-boot or in a VM at native speeds. While it was possible to run Windows in a VM back in the PowerPC days, it was sloooow. While switching to ARM would be relatively trivial for Apple's own software and even OS X applications, the downside would be losing easy/performant/power efficient access to non-OS X applications. That said, I expect Apple to make that trade-off sooner rather than later... x86 compatibility isn't nearly as important today as it was 10 or even 5 years ago.
• umvi 7y ago
  Well, I've purchased Macbook Pros, iPhones, and iPads in the past, and every single one got incrementally more sluggish with each OS update until it was unsuable. This was in 2013-2015. It's been 5 years since I've owned an Apple product, but I just have a hard time believing Apple is heads and shoulders above everyone else.
• thefounder 7y ago
  >> By sticking to Intel or even x86 in general, there is ample evidence Apple is leaving a lot of performance on the table. Not just in MacBooks - but for the Mac Pro too. So we should expect lock-in on Mac Pro chips as well. Exactly what people were asking for. Another brick you can't upgrade without paying the cost of a new machine.
• api 7y ago
  I'm not sure the difference is so great. It's all about the node about the node... (to the tune of "all about the bass" :) I just got a 2020 MacBook Air with an Ice Lake 10th generation 10nm X64 chip in it. It's a quad-core, newer core rev, has AVX2 and a bunch of other stuff. It runs very noticeably cooler than a 2019 Air with a 14nm Amber Lake chip in it. Battery life is also noticeably better. When I read the spec differences I basically traded my 14nm Amber Lake Air for a 2020 Air (and also because of the better keyboard). It's a great machine. The difference between the Amber Lake and Ice Lake core designs is not substantial and the Ice Lake has twice the cores and a larger on-board GPU, so how is it so noticeably cooler? I can fully max out all four cores of the 10nm Ice Lake and it doesn't get as hot as the 14nm Amber Lake did at lower loads! The answer is obviously the process node: 10nm vs 14nm. It's a more efficient chip at the physical circuit level. ARM has some intrinsic power advantages over X64. The biggest thing is that ARM instructions come in only two or three sizes and it's easy to size and decode them, while X64 instructions come in sizes from one byte to 16 bytes and are a massive pain to decode. That decode cost comes in energy and transistors, but it's worth pointing out that this is a mostly fixed cost that shrinks as a percentage of the overall CPU power/transistor budget as the process node shrinks. In other words the cruftiness of X64 remains the same as you go from 22nm to 14nm to 10nm. Other than the ugly decode path the ALU, FPU, vector, crypto, etc. silicon is not fundamentally different from what you'd find in a high-end ARM chip. A lot of the difference is clearly in the fabrication. ARM chips have been below 14nm for a while, while Intel X64 chips have lagged. (Tangent: since the actual engine block is largely the same, I've wondered if Apple might not slap an X64 decoder in front of their silicon in place of the ARM64 decoder and make Apple X64 chips?! I am not a semiconductor engineer though, so I don't know how hard this would be and/or what IP issues would prevent this. Probably unlikely but not impossible. They certainly have the cash and leverage to muscle Intel into licensing anything they need licensed.) If Intel gets its act together with process nodes and/or starts using other fabs who are at tighter lower power nodes, the advantage will shrink a lot. AMD already has mobile chips that are close to Apple's ARM chips in performance/watt, partly because they are fabbed at TSMC at 7nm. BTW: I'm not claiming Apple's chips aren't impressive, and as long as they don't lock down MacOS and make it no longer a "real computer" I personally don't mind if they go to ARM64. Also: the fact that Apple's chips only get this great performance in bursts is mostly due to power and cooling constraints on fanless thin phones and tablets. In a laptop with better cooling or a desktop they could sustain that performance no problem.
• greendave 7y ago
  I'm always a bit suspicious about x86 vs. non-x86 micro-benchmarks. I remember all the fun people had with ByteMark back in the day, and while I assume that Geekbench doesn't play those sorts of games with compiler optimizations, I would really like to see data from something a bit more representative of a real-world CPU-bound application (short-lived is fine, just not synthetic).
  • qppo 7y ago
    Geekbench isn't a micro benchmark, it's a comprehensive test of the system using a variety of programs and workloads and aggregates the results. It's not a single program that one can play games with compiler optimizations.
    • DCKing 7y ago
      I should have clarified that Geekbench can be more accurately described as a set of microbenchmarks. It does test a lot of different kinds of performance, but it does not test sustained performance.
      • qppo 7y ago
        I don't like arguing semantics but I don't really think any of what Geekbench does is a "micro" benchmark. At least for me that typically refers to running a small snippet of code, like calculating a dot product or something. Geekbench tests whole program performance. It's not aida64 but it is a pretty decent metric, and consistent.
• dzonga 7y ago
  having faster processors, is what's wrong with our computing industry in general. processors on laptops | desktops have become 1000x faster, but has the software kept up. no software has gotten slower by the ages. Apple, Qualcomm, Intel can make as much faster processors with x cores, but do we have software able to utilize those, eh! run a JS heavy site | app and you see most processors heat up these days whether mobile or desktop. most programming languages can't easily delegate work to cores with smoothness like how Erlang n Elixir do it. in Python threads, were a nightmare but now with concurrent futures or dask at least we can utilize all cores. tldr - we need to make faster software
  • DeathArrow 7y ago
    >tldr - we need to make faster software Just use compiled, strongly typed languages. No: Javascript, Python, PHP, Ruby Yes: Java, C#, C, C++, Rust.
  • na85 7y ago
    It's not even "faster software" so much as eliminating the culture of Developer Convenience at the expense of User Experience. That's what got us Electron. I've actually seen comments on HN unironically describing the web as "the perfect app platform". Unfortunately, the majority of users seem conditioned to accept software with awful performance, so there's no impetus for developers to upgrade their skills.
    • jarjoura 7y ago
      Don't worry, the future is wearables and more efficient mobile phones. Electron just won't run on your Apple Watch.
      • saagarjha 7y ago
        The only reason Electron isn't on your Apple Watch today is because of App Store Review.
      • jamesgeck0 7y ago
        The mobile equivalents of Electron (PhoneGap/Cordova/Ionic) are already extremely popular.
• jayd16 7y ago
  How was Apple able to get here? What secret sauce does Apple have in it's chip that's beating out Qualcomm, Intel and AMD?
  • tarsinge 7y ago
    Isn't the current Intel Core line an evolution of the Pentium M (2003), itself an evolution of the Pentium III (1999)? Apple starting a fresh design with up to date constraints may have given them room for improvements I guess.
    • jayd16 7y ago
      That doesn't explain why AMD and Qualcomm can't keep up either.
    • hajile 7y ago
      Sandy Bridge was the last really big architectural change. It seems heavily inspired by the Alpha EV8 design (Intel bought Alpha from HP in the 2001) with of course, a very different decoder section (they wrap the x86 decoder around a RISC architecture). https://www.realworldtech.com/alpha-ev8-wider/ https://www.realworldtech.com/alpha-ev8-wider/
  • chillacy 7y ago
    Vertical integration? Even if their chips end up being a bit slower in the end, they'd probably increase their overall profit margins and get increased flexibility aligning their development cycles.
  • paulmd 7y ago
    Apple is willing to trade die space (i.e. manufacturing cost) for performance in a way that their competitors are not. Anandtech architecture reviews are helpful, as always. Worth reading for a page or two from the linked page: https://www.anandtech.com/show/14892/the-apple-iphone-11-pro-and-max-review/2 https://www.anandtech.com/show/14892/the-apple-iphone-11-pro... https://www.anandtech.com/show/13392/the-iphone-xs-xs-max-review-unveiling-the-silicon-secrets/2 https://www.anandtech.com/show/13392/the-iphone-xs-xs-max-re... The short of it is: massively wide execution units, massive amounts of SRAM, massive amounts of cache at all levels. There's no real "secret sauce", they're just willing to pay to make an incredibly fat core and Qualcomm and company are not. They have 16 MB of system cache on A13 for 2 high-performance and 4 low-performance cores, which is as much as a 9900K gets for 8 cores and as much as Zen2 gets for 4 cores. Plus another 8MB per big core on top of that (so up to 24MB per core in single-threaded mode), and 4 MB per small core. It helps that they're a vertically-integrated company, they don't have to sell their processors on the open market at competitive prices such that an OEM can also make a profit selling a finished product at competitive prices, they just sell the finished product.
• pmart123 7y ago
  How will this change relying on libraries like Intel's MKL library if Apple is using their own chips?
  • wmf 7y ago
    Apple will say developers should have used Core Whatever instead of Intel libraries.
  • ben_w 7y ago
    One of the general downsides of Apple development is how often they change absolutely everything. Like the time they switched from 68K to PowerPC. Or MacOS classic to Mac OS X. Or PowerPC to Intel. Apple provides brief windows of automated compatibility, but code untouched since 2003 won’t run on a mac today, and code untouched since 1986 wouldn’t run on a 2003 mac.
• mpweiher 7y ago
  > extremely far ahead Apple seems to be in terms of CPU power and efficiency RISC finally coming into is own. For the longest time, Intel was able to fend off much better architectures simply by being a fab generation or two ahead, more clock speed, more transistors, more brute force. Not to belittle the engineers wringing out seemingly impossible performance from the venerable architecture, but the the architectural limitations always mean extra work and extra constraints that have to be worked around. And now that Dennard scaling is dead, Moore's law wheezing and not helping all that much for our mostly serial workloads, they just can't compensate for the architecture any longer, at least not agains a determined, well-funded and technically competent competitor that's not beholden to Wintel. I remember when the Archimedes came out and just offered incredibly better performance than the then prevalent code-museums, 386 and 68K variants, at incredibly lower transistor counts. The 486 and 68040 were able to compete again, but with vastly larger transistor budgets (and presumably power budgets as well, but we didn't look at that back then). Oh, and can we have our Tansputers now? Pretty please, Xmos?
  • imtringued 7y ago
    I'm not sure how you can call that a victory. ARM and AMD are only ahead because of superior manufacturing processes. That's exactly the thing you accused Intel of doing.
• cptskippy 7y ago
  > It's worth noting that Geekbench is a pure microbenchmark. This isn't just a note, it's an important clarification. Microbenchmarking is used in-lieu of proper benchmarking because you can't do proper benchmarking.
• naravara 7y ago
  >It's worth noting that Geekbench is a pure microbenchmark. The iPhone will not sustain performance as long as the others. The point is that Apple could solve this when moving to bigger devices. Is that inherent to the architecture or is this a self-imposed limitation by Apple since it has to sip power and run without any active cooling? Also, the A-series chips seem to fall down in comparison against the Intel Macs on multi-core performance which seems like it would matter for anyone who needs a desktop.
• mlinksva 7y ago
  > Core i7 8086k TIL of this commemorative naming. Sole comment I could find on HN 2 years ago https://news.ycombinator.com/item?id=17409849 https://news.ycombinator.com/item?id=17409849
• bsder 7y ago
  > It's worth pointing out how extremely far ahead Apple seems to be in terms of CPU power and efficiency. Nope. But Apple put a MASSIVE amount of cache on their chips. It's not rocket science, but it is expensive.
• zrm 7y ago
  > By sticking to Intel or even x86 in general, there is ample evidence Apple is leaving a lot of performance on the table. That isn't necessarily true. Having competitive performance at lower power isn't always the same thing as having better performance, even assuming these benchmarks are representative. Processors designed specifically for low-power make different design trade offs. One of those is to exchange maximum clock speed for IPC (because higher clocks burn watts). The A13 maxes out at 2.65GHz, the i7 8086k hits 5GHz. Chances are you can't just give the A13 a 95W power budget and see it hit 5GHz, it would have to be redesigned and the kinds of changes necessary to get there would generally lower IPC. Apple is also riding the same advantage as AMD -- they're using TSMC's 7nm process which is better than what Intel is currently stuck with. Even AMD is still using an older process for the I/O die. We don't know what that's going to look like a year or two from now. Meanwhile the renewed competition between Intel and AMD makes this kind of a bad time to move away. They're both going to be working hard to take the performance crown from each other and Apple would have to beat both of them to claim an advantage. And continue to do so, or they'd have a lot of pissed off customers and developers after forcing a transition to a new architecture only to have it fall behind right after the transition is over.
  • imtringued 7y ago
    >Chances are you can't just give the A13 a 95W power budget and see it hit 5GHz, it would have to be redesigned and the kinds of changes necessary to get there would generally lower IPC. It's not really a difficult concept to understand. If your CPU runs at 5GHz then maximum time a single cycle is allowed to take is 0.2 nanoseconds. CPU designers have to make sure that this limit is never exceeded anywhere on the chip. If you make a even the slighted mistake in some unimportant corner of the CPU you will end up limiting the maximum performance of the entire CPU. Most CPUs are optimized for a specific clock frequency and going beyond it is not possible without sacrificing stability.