Apple Preps Next Mac Chips with Aim to Outclass Highest-End PCs
bloomberg.com
bloomberg.com
i.e UMA ( Unified Memory Architecture ), GPU, NPU stay intact. You are basically getting a M1XX, where M1X is the 16 CPU Core variants.
Consider the current M1, a 32 Core High Performance CPU Core and 128 GPU core would give you roughly 300W TDP Max, excluding all the memory controller and interconnect. A rough estimate is already at 500mm2+ Die Size, excluding NPU and all other bits and pieces, of course assuming the same processing node.
The Old ( Or the recent ) Mac Pro has a Maxed out system of ~910W, that is 28 Core CPU + 2 x Radeon Vega II Duo ( That is 4 GPU ).
We are back to TrashCan Mac Pro again, except it would be even faster than the 2019 Cheese Grater Mac Pro. We would have reached the thermal ceiling again, unless Apple does Water Cooling and push that TDP allowance to 400W.
Remember this M1XX is CPU + GPU TDP. For a PC system, or workstation power users are quite used to 300W CPU + 300W GPU. So I was sort of expecting Apple to be making CPU + GPU design, purely to create a monster as a successor to Cheese Grater Mac Pro. May be they have different Goal in mind, may be they think this is good enough for Mac Pro ( Which certainly is from many perspective ), and obviously iMac.
Same Design, scaling across iPhone, iPad, MacBook, And Mac Pro.
This brings me to the final point, we only have so much uArch improvement we could do, it is now all left to process node and heat dissipation of the system.
We are in the era of TDP Computing.
Intel has been saying this for years, like almost 8 years. But this is just what non-innovator say. It is an excuse for mediocrity and failure and stagnation.
There is tons of uArch improvements in Apple's chips. That is why they consistently beat even the other ARMs chips on the same process nodes.
AMD also has been significantly innovating with uArch.
I love the innovation we are seeing these days. We are definitely not in an TDP era (where only TDP matters! but of course TDP is important), we are in another golden age of CPU/GPU innovation and it is beautiful. Reminds me of the 1990s and early 2000s.
According to Wolfram Alpha, 5 nanometers is ~23 silicon atoms across (~2.2 Å each). But that’s also assuming they’re actually fabricating 5 nm gates - which they’re not.
The oxide/insulator thickness on modern process is pretty insane. For instance for the intel 14nm++ the silicon nitride and silicon oxide layers seem to 4 to 5nm thick.
-Edit- For instance here is 2 fins for from the intel 14nm++. The grains/spheres you see this image are the atoms. Some parts of the a modern transistors are even smaller than process number may let you know.
The A13, Qualcomm Snapdragon 865, and Mediatek Dimensity 1000L were all on DUV N7P (Hisilicon Kirin 990 5G was actually on the more advanced EUV N7+). For more information on chips and their processes: https://www.eetimes.com/apple-huawei-use-tsmc-but-their-7nm-...
Here are some benchmarks of how the A13 and 865 (and the 2020 updates) stack up: https://www.phonearena.com/news/apple-a14-vs-a13-vs-snapdrag...
From the charts, it looks like particularly for single-core and browser/JS performance, the Apple designs are very strong, and beat out the Qualcomm competition even on the same process node.
Intel is basically running a super legacy design and they are doing surprisingly well with it, but they are losing on all fronts at this time. What they have for them right now is momentum and enterprise deals.
That will hold up for a while yet, so it gives them time to get their act together.
[1] https://www.anandtech.com/show/15885/hpc-systems-special-off...
It's more likely that the future Mac Pro will use Lifuka which everyone in this thread is ignoring.
It would be interesting if Apple did make a discrete GPU, that would means they are making a separate GPU die for their Mac Pro and iMac with volume estimated to be no more than 1M unit per year. It makes sense from a technical perspective, it surely doesn't from a cost perspective. Although Apple could happily charge $1000 just for the GPU to make up for ROI.
The advantage of GPU having it's own memory is you can use super high throughput ram that the GPU needs, while using higher capacity/lower energy/cheaper ram for the cpu.
The downside of using separate ram is that you have to shuttle data back and forth from one ram to the other quite a bit, as well you get less flexibility in how your total memory is used.
Who do you think writes all the code for this stuff? It’s guys in tracksuits in Visual Studio C++ on Windows. The top games for iOS are also either written by a few guys in Unity or by an absolutely huge team of people porting an existing, low graphics quality PC game.
> We are back to TrashCan Mac Pro again... we would have reached the thermal ceiling again...
This is kind of drawing new lines to read between. :)
Apple almost certainly developed their current uber-expensive, trypophobia-triggering Mac Pro design knowing that they were also working on Apple Silicon processors. While it's possible they intend that whole case design to be a one-time thing, it doesn't seem likely to me -- and if that wasn't the intent, then it's also not likely they want to ship a teeny tiny Apple Silicon logic board in that huge case with the rest of the space going unused. I think they're going to do their best to make an Apple Silicon-based Mac Pro that can be upgraded and expanded. (With all respect to Mark Gurman, I think that "half-sized Mac Pro" is going to be a different product entirely if it ships, with a different branding. Mac Mini Pro?)
The ever illusive xMac, aka "the affordable Mac tower"
The iPad is at least an order of magnitude better than comparable netbooks (though chromebooks, depending on manufacturer, can be competitive thanks mainly to ChromeOS's reduced footprint). So even though the margins are high, the perceived quality, regardless of raw benchmarks, is still something. It's not just marketing to me to say the marriage of software and hardware is unique. (Gruber's observation about NSObject alloc's being a lot faster on Apple Silicon, for instance).
Now that I think about it, honestly Google is the only other company playing by these rules... ie pixelbook, pixel phone, etc. But they're much earlier in the evolution, and have less upstream control in software (especially since Fuchsia seems to be somewhat lower priority than before, though this is second hand knowledge).
As a technology enthusiast I'm excited to see what bonkers stuff they'll be announcing. But, since I am not doing climate models or rendering hollywood movies, I don't expect I will never work on any of them since I don't want to drop a new car's worth of money on one.
No, but unified memory does require starving the GPU of bandwidth. There's a reason high end GPUs use different types of memory than CPUs.
Either Apple will be content with Mac Pro not having the absolute graphics performance lead, or they will make a discrete GPU with dedicated memory. I'm not sure which they will choose but I suspect the latter.
There doesn't seem to be any reason why Apple couldn't integrate a 16 TFLOPS GPU on an SOC using a 5nm or better process. The M1 GPU is 2.6 TFLOPS, they can x6 or more it, and obviously change the memory subsystem (HBM?).
Also, Xbox uses GDDR which trades off CPU performance for GPU performance. Apple probably wouldn't do that for a Mac Pro. The high end needs separate memory so the GPU can have bandwidth and the CPU can have low latency.
Also, GDDR does not have appreciably higher latencies than DDR memory when measured in nanoseconds. It's just more expensive than DDR and much more limited in terms of capacity.
If Apple does go the "huge SoC" route I'd expect to see HBM2 memory with socketed DDR4 or DDR5. It'd provide the best of both worlds - extremely high bandwidth and low latency for a small portion (say 32-64GB) of the memory, and high capacity for the rest (say 1-2TB), all without compromising the unified memory concept.
This is not without precedent - recent Xeon Phis, for all their other shortcomings, have had a similar memory hierarchy.
I agree that a combination of HBM and DDR sounds pretty good for a unified memory architecture. Are you imagining it as just another layer in the cache hierarchy or something actively managed?
So you get a certain baseline level of specs, but also have the opportunity to tack on a beefier GPU (which may or may not use UMA), additional slower memory, additional slower storage, maybe even additional CPUs, etc.
I don't think so, with the asterisk that I don't know enough to say whether mixing higher-speed and lower-speed RAM is really feasible in that scenario. (IIRC, if you do that in computers now, all the RAM will operate at the lowest speed.)
That’s correct, but irrelevant in this context. In that scenario, the memory is all on the same memory controller, and the memory controller used wants to run a single speed, so it has to be the highest universally supported speed.
Apple could have multiple tiers of memory with different characteristics, but just because you can, doesn’t mean you should. I don’t have access to such a test bed, so I can’t speak to what the experience would be like.
The Radeon Vega II Due has 128 compute cores and 8192 stream processors, so that maxed out mac pro would have 256 compute cores and 16384 stream processors. I don't have any benchmarks to work off of, but I think it's safe to say that a new Mac Pro with the 32-core SoC would still be substantially far behind a current Mac Pro unless it also had a dedicated GPU.
That is unless Apple decide to use DRAM in additional to HMB or GDDR5 as 2nd layer of Memory.
Yes, we can have power-efficient phones and laptops, but what about other product categories? I think the market should be more modular so that any company can build cool stuff with top of the line CPUs, not just Apple.
As a hacker, I think with Apple's dominance computing is becoming less interesting. It's like opening a LEGO box, and all pieces are already fused together.
And anyone can easily beat Amazon's pricing if they can just run a data center competently too - after all the building blocks are there.
Seriously?
Zen3 looks pretty on par with the Apple stuff, maybe it's finally time for OEMs to adopt AMD in a serious way so they can invest into R&D even more.
Also at the end of the day performance oriented people aren't going to care so much about process node idiosyncrasies. Does it perform or not?
Not that I would be mad if AMD meets or heck even bests Apple. Competition is good for all of us :)
It is extremely likely other companies will step up with new offerings. But it will take 2-3 years (where Apple will reap the benefit of being early).
I remember using laptops with LCD screens in 2010, and even before. They weren't even Apple laptops.
Dell had a 2560x1600 LCD in 2007. https://www.cnet.com/products/dell-ultrasharp-3007wfp/specs/ https://www.dell.com/downloads/emea/general/ultrasharp%20300...
Dell and Sony also had laptops going as high as 4K in 13" form factors around the same time. Yes they were high end. So too was the MBP.
When Steve Jobs announced iPhone on stage in 2008 he said it was a “revolutionary and magical product that is literally five years ahead of any other mobile phone,"
That was saying the quiet thing out loud and Apple is playing it much cooler on PR around competition these days.
But Apple Silicon combined with Apple software is very likely at least 5 years ahead of the competition, because there is not currently an assembly of organizations that can integrate to offer anything like what Apple is doing _now_.
What makes this article particularly interesting is it points at beating NVIDIA and AMD in graphics and reducing the size of the enclosure by half.
I find it interesting that Apple just redesigned the Mac Pro only to have to redesign it again for Apple Silicon.
Oddly, the failed trash can Mac Pro may have actually worked for Apple Silicon designs.
I suspect that Apple will provide web services to Apple developers that breaks pricing for Intel / NVIDIA based cloud compute and further burnishes the value of building for Apple platforms.
Apple can fab at max scale if it can consume its own chips in its own data centers that can flex between serving Apple or AWS or Azure customers.
It’s going to be a rough go for the PC market because not everyone is going to make it through this—-at least not as the companies they are today.
Why can't others just "get better" and make their own chips that actually compete?
That's not to say it would be impossible for another company to try to get themselves to where Apple is today—but it would be far from trivial.
It's not like they're an example or anything... I mean, it obviously doesn't work for every market, but I just see it as kind of strange that someone hasn't taken a page from Apple's playbook over the years when it would suit them.
When Steve Jobs came back as "iCEO" in the late '90s, Apple stock was roughly $8/share (pre-splits). They weren't a total nobody, but they certainly weren't interesting enough to any big fish that there was any push to put pressure on them to act like a more "traditional" company. And Jobs kept doing things his way for the next 15 years, until his medical retirement and untimely death, and those things increased Apple's stock price fairly steadily. Until the iPhone came out, of course, at which point it increased dramatically.
But by then, people were already used to the idea that Apple would do things Apple's way. (And some people were so used to the idea that Apple was Doomed that they still preach it every time there's any new Apple news.) And it's so utterly clear that Apple doing things Apple's way works—for Apple.
Any other company trying to do things Apple's way—or even its own unique way, rather than the way the markets want—would be fighting an incredibly strong current.
So many people, even those who should absolutely know better, seemed to get caught in the trap of believing that a company—any company—can only be "viable" if it can constantly outdo itself and all its competitors.
1985 - 1997
Anyway, thanks for the reminder that the period without Steve Jobs was less brief than I tend to think.
"We believe that a fundamental measure of our success will be the shareholder value we create over the long term...
Because of our emphasis on the long term, we may make decisions and weigh tradeoffs differently than some companies...
We will continue to make investment decisions in light of long-term market leadership considerations rather than short-term profitability considerations or short-term Wall Street reactions..."
Maybe there's something to this approach?
[1] Bezos letter to shareholders, 1997 - https://venturebeat.com/wp-content/uploads/2010/09/amzn_shar...
But who's fault is it? I would tend to think that it's Windows Arm port that sucks since Qualcomm chips offer good perfs for Android phones.
My point is that PCs have offered high end specs at affordable (but above average) price points even before Macs had them. And the fact that many PC laptops still offer low res screens point to price being the deciding factor which moves the PC industry forward, not Apple.
I still wish Apple would release a reasonably priced stand alone display. The Pro Display is amazing but I could never justify it :/
I'm happy that Apple is showing that the king is naked, I hope that a new wave of ARM based devices (not just from Apple) will influx new life on the PC Market.
So other metrics like power draw get more prominence. Apple managed to get both with a chip design, but I don't see it as a trend yet. I wouldn't even call it innovative in itself, it is just a very good iteration of CPU tech. I think that is awesome because it challenges other companies again.
For performance intensive applications like gaming or CAD, GPUs have become more relevant.
There is a lot of special software for physical calculations (lenses, materials, etc.) that require raw CPU power for n-th degree numerical solutions, but these are niches that cannot sustain an industry.
My main device I use in private for a lot of development has a standard intel CPU and I only looked at power usage because I knew it would be fast enough. I wanted it to be as silent as possible. Turns out it was indeed fast enough for anything I wanted to do. Maybe I buy a new one in a few years, but it didn't yet fail me in any way.
Free dancing bears and network effects will always win.
My point was that most consumers don't understand these things, and just assume that "computer speed" is one-dimensional: it's either fast, or not fast.
I disagree. At least in the datacenter there are still lots of workloads that are CPU bound. And many of these workloads are not easy to parallelize. Right now there are about a bajillion startups in a Python codebase that are hitting scaling limits, where the single-core GIL limits are starting to hurt responsiveness.
Yes, they could re-write into Go and get higher performance and better concurrency at the same time. But most orgs don't have the bandwidth for re-building their software stack from scratch. They'd much rather throw money at hardware.
My guess is if you had a datacenter CPU with 2-3X the single-core threaded performance of top-level Xeons, it would sell like hotcakes.
Many companies would rather throw money at beefier hardware than play around with experimental JITs that may or may not break standard behavior in unpredictable ways.
The rave reviews that the M1 Macs are getting for performance and power efficiency shows that large step improvements in CPU performance are noticed and greatly appreciated. We just didn't have anything as good as the M1 before to really understand how much better computing could be.
This is just not true. I do NOT have an M1 yet. I have a group of co-workers that do. This group is normally what I would call the critics that are never happy with anything. With out fail each one of them has raved about how fast and responsive their M1s are. They say it nothing like anything else and is a game changer. They also rave about the fact that they can go a day without needing to find a wall socket. This is a group of devs and hardcore networking people.
From my view, no matter how powerful of an Intel system I have running Windows, macOS or Linux, the desktop UI are not up to the speed of an iPad. Getting that in a laptop...that would be great and I plan to try it out in the form of an M1 Air.
What I read: Desktop environment should have more UI effects like iPad.
What exactly is stopping, say, Qualcomm, to introduce "top of the line CPU"? Apart from engineering merit?
They did their level best to get maximum control and this was and is a huge turn off. If they made it as open as x86/x64, and pulled out all the stops to get software support equivalent to x86 like Apple have done with their system, then it would be a compelling proposition. As it is, it has woeful software support, and Microsoft did very little to address this.
Investing in a top end CPU is expensive and currently there is no buyer for a workstation class CPU other than Apple.
Microsoft's ARM efforts have been lackluster and sales has been lethargic. The last time Microsoft tried to jump to another architecture was an expensive failure. Who is going to invest hundreds of billions in jump starting this category with a track record like that?
Likewise, Linux. While Linux is a great development platform, Qualcomm isn't going to gamble tens of millions on developers migrating en-mass to a Linux/ ARM platform.
Someone has to put up the money and it's a bunch.
Currently, the only company willing to bet enormous amount of money and frankly the entire Mac platform to change status quo is Apple. Maybe (and I really hope so) Apple will push other ARM processors on the desktop as a side-effects - while Qualcomm is more or less "just" the packer of ARM designed cores, and if I understand the market it's entirely out of their reach to design M1-class CPU, maybe Apple will push ARM to design more efficient desktop cores. Maybe Amazon or Google will do that with their custom ARM silicon now when Apple demonstrated that processors doesn't need to be hot to be powerful. For so long Intel sold us basically the same thing, and we cheered when AMD was just a couple percent faster. However this ends, I'm glad that someone at least rocks the boat.
Nvidia is the most likely company to build a higher end ARM CPU to compete with the M1. With the ARM core and their GPU expertise, they could conceivably crank out something which performs much better than Qualcomm.
HP ships nearly 50% more PCs. Dell sells just a few more. I don't know that I would say they are dominant, after a release of a single low power, high performing part.
What many in the Apple ecosystem are missing is the steady, rapid improvement in performance and efficiency coming from AMD, who apparently few have heard of, because only Intel CPUs were found in Apple for 14 years (and we're not all gamers!)
In early 2020, AMD released massively dominate laptop CPUs, but OEMs have only tepidly adopted them. It looks like about 20% of notebooks are now being sold with AMD, and that number is going to grow in 2021, as the lead time for new designs using AMD will have elapsed.
https://www.pcworld.com/article/3569437/amds-notebook-pc-sha...
I'm not saying these AMD chips are better than Apples - the battery life is at best half!
Although, I can't help but read your comment as satire! "What if PC components were modular!" I also enjoy building sets from LEGO bricks, and putting together my computer from parts!
Huh? They're saying you won't be able to do that with an Apple CPU.
Amazon has Gravitron 2. Microsoft is doing something with "their" SQ2, though at this point they're mostly faking it by putting a badge on an uncompetitive Qualcomm part, but I'm sure they'll get there.
The M1 is a wonderful chip, and it has done loads to push the industry to really adopt ARM -- possibly the greatest fat-computing transition to ARM yet -- and it should help across every market, and loads of hardware from competitors.
All technological innovation is coming from Apple? That's a mind-boggling statement.
Apple released a great CPU, which replaced the last best CPU, and some other CPU next year will be better than both (it could be another Apple chip, it might not be).
At the end of the day, the average consumer wants performance insofar as their device can run apps they want (benchmarks means squat), and battery life. Apple moved the envelope forward here, but let's not pretend they landed on mars.
I think Apple just showed us that being less modular can bring huge benefits.
As an example, in a modular world, the CPU designer would add instructions that speed up encryption/video decoding/whatever, larger vector registers for speeding up ML stuff, etc.
In a M1 world, the system designer might say “we don’t need them in the CPU; the system will have custom hardware that can do that better”, or even “yes, that would speed up the system, but we better spend that transistor/heat budget on X”
If Microsoft doesn't go down this path for Surface, don't hold your breath any other vendors will.
PC's are not where innovation happens. Intel had USB for almost two years yet PC's still shipped with serial/parallel ports. Then the candy colored iMacs hit, and all the accessory makers started chasing them with candy colored USB accessories and then all the sudden the PC vendors woke up to USB. USB was a disaster on Windows 95, Marginally better on Windows 2000 but didn't get decent until XP.
Wifi - Apple not only released wifi on their laptop but a whole ecosystem - Access points too. One app to configure and manage everything. Push button simplicity. Wifi thrashed around on PC vendors and was relegated to PCMCIA cards with dorky external antennas before Apple baked it in and pretty much embarrassed the rest of the industry to stop nickel and dimeing everyone as an add on and just include it as base tech. Or at least bury the antennas in every laptop and still charge people for an optional internal card. How long did that continue for?
I'm sure there are some others I could think two but I vividly remember living through both of those as I had a foot in each camp at the time (actually I was more PC back then since before the iMac is was the craptastic, uninspired and overprice Performa years for the Mac - the PowerPC iMacs brought me back into the Mac fold).
PC has always been focused on high volume, low margin, least common denominator hardware. Big bins of unremarkable but functional interchangeable parts. Major Pro's of that strategy are pricing and choice. Major Con's are instability/incompatibility/driver issues and lack of moving innovation forward.
Apple has never (and will never) play in the value market. Remember how Apple was supposedly doomed because they didn't have a netbook? Who still uses netbooks (or even remembers them other than us geeks?) By not playing in the value market (or race to the bottom) it frees them up to do things like the M1.
Fast/Good/Cheap. Pick any two
I'd consider an Apple laptop if I knew any game I wanted to play would work, and I never play the latest AAA games. But so far I checked and StarCraft 2 does not seem to play well, despite being a 12 year old game. (That may change with software updates?)
Looks like at least some people have had good luck with StarCraft II.
I was all set to replace my 2015 MBA with an M1 MBA, but in watching his (and others) Cities:Skylines performance I'm now thinking I may switch to gaming on Mac full time. Only problem is 16GB of RAM is NOT going to cut it with C:S - so I now am waiting for the next batch. 32GB should be plenty and I would be shocked if the next group of machines didn't have at least one laptop that could go to 32GB.
Hopefully still HBM too.
I'm hoping a cloud gaming service takes off that works more like a dedicated remote desktop rather than an a la carte storefront the way they all seem to be.
The big advantage here is the ability to pull all of this functionality onto smaller and more highly integrated SoCs and packages. This is the antithesis of modular.
If you are looking for similar performance, the best you can hope for is a more accessible competitor. Someone like Nvidia or Qualcomm releasing a SoC with comparable features.
The big problem there is Qualcomm and Nvidia both have closed drivers so they are only marginally better than Apple. Since Qualcomm does such a lousy job supporting older CPUs, arguably worse (what's the value of a CPU with only 3 years of support?).
Still... the Fujitsu A64FX is really interesting but I doubt we’ll ever see it outside Japan.
Ampere Altra is similarly interesting but it’s hard to convince management to put faith into a startup with its own specialty CPU.
This is how apple has lost its way.
Apple should be more "modular". (and that would lead to what you want)
What people don't realize about steve jobs was that he was the great integrator. He got different folks to work together.
The mac pro from his era had pcie slots and worked with add-in cards from lots of vendors. You could add 4 hard disks.
Another way of looking at it is to read that silly book "The 7 habits of hightly effective people" from decades ago.
Basically said the arc of maturity is:
- dependent (child)
- independent (teenager)
- interdependent (fully developed)
After Steve Jobs, Apple regressed to independent. Instead of using its market power to foster the ecosystem, it has regressed to teenager level and just does what is good for itself. There are no other hardware manufacturers anymore and now their software developers don't feel like partners and tread carefully.
Thing is apple can probably do well for itself independently but... It's lonely at the top especially if you've turned your back on the rest of the ecosystem.
lol - the only reason the Apple II had slots is Woz threatened to quit if SJ won on excluding them.
The Original Mac wasn't designed to be opened by users. You had to have a LONG handle Torx screwdriver to open it. Job's was VERY much into appliance computing. Any expansion baked into any Mac was in spite of, not because of, Steve Jobs.
Many things I like/appreciate about the man - his obsession over all computers being sealed was NOT one of his more endearing characteristics. Obviously he was flexible - we had Mac's with slots back then and we do now too. But there has been no "regression" since his passing.
Hardly!
https://www.filfre.net/2014/02/macintosh/
> Jobs was a huge headache for Michael Scott, Mike Markkula, and the rest of Apple’s senior leadership, who received memos almost daily complaining about his temper, his dismissive attitude toward the Apple II platform that was the only thing supporting the company, and his refusal to listen to reason when one of his sacred precepts was threatened. Jobs’s headstrong authoritarianism had been a big contributor to the debacle that was the Apple III launch. (Traditional wisdom, as well as an earlier version of this article, would have it that Jobs’s insistence that the Apple III ship without a cooling fan led directly to the hardware problems that left Apple IIIs dying on buyers’ desks by the thousands. It does, however, appear that this version of events is at least questionable; see the comments section for more about that. Be that as it may, everyone involved would agree that Jobs did an already muddled project no favors.)
and yes, the reality distortion field was his preferred method for that.
I had meant, based on your previous comment, that he had used the field on you :-p
"If they live up to expectations, they will significantly outpace the performance of the latest machines running Intel chips, according to people familiar with the matter who asked not to be named because the plans aren’t yet public."
It's not entirely obvious that "the latest machines running Intel chips" here probably refers only to the latest Macs running Intel chips. We already know that the M1/Firestorm beats Intel handily in single thread, so this just means they'll use enough cores to beat the 28 core Xeon-W in their current Mac Pro. This is of course completely uninteresting, since 1) they've long announced they'll be Intel-free by 2022 and 2) they're not going to release new Macs any slower than the current Intels.
Going by Geekbench they need probably a 16 to 20 Firestorm CPU to beat the current Xeon W-3275M Mac Pro. A 32 Firestorm CPU would completely obliterate it. To make a statement, they'll probably do just that.
The more interesting question to me is if they're going to go for the absolute performance crown, currently the AMD EPYC 7H12, a 64 core Zen2 CPU which benches around 60k in Geekbench5 (vs 19k for the 28-core Mac Pro). By mid-2021, that that will be probably a 64 core Zen3 EPYC at around 72k (+20% IPC vs Zen2). Apple can probably match that with a 48 Firestorm CPU. A 32 core CPU won't do it, even with a generous IPC increase next year.
First of all, that's ludicrously expensive. You need an active interposer that's larger than the chip to do something like that. 4x HBM stacks means 4096 wires to run between your compute-device and the RAM, that's a lot of wires in a tiny area, and only possible with advanced packaging (expensive) methodologies.
Second: HBM2 based designs have poor memory-capacity: 8GB per stack means 4x HBM2 caps out at 32GBs under most reasonable configurations.
Going into 2021 with a ludicrously expensive 32GB machine seems a bit weak. I don't think anyone in the high-end market is targeting that level of memory capacity. Even if we're generous and HBM capacity doubles between now and then, I'm not sure if 64GBs is enough for workstation tasks.
HBM works for supercomputers, because when 32GB isn't enough, you just buy another node. A64FX, the current CPU-based supercomputer that uses HBM, only has 32GB/node. Then Fugaku bought 150,000 nodes.
I think Apple would want to release a 64GB or 128GB Mac Pro... but only to upsell its customers to a 256GB or higher Mac Pro. The ability to configure more RAM into a system is key for marketing (even if this RAM is soldered onto the motherboard: like LPDDR4x, it wouldn't be too expensive to make different motherboards especially in a niche $10,000 computer like the Mac Pro).
But if Apple did go with HBM2 itself... well... then its physically unable to go above 32GB (today), or maybe ~64GB (if bigger HBM2 modules come out next year).
--------
Sticking with LPDDR4x just seems like the most obvious, and best, decision for Apple. So that's my expectation. But then again: making a 8-way decoder + fat-as-heck execution pipeline wasn't part of my expectation either. So maybe Apple will surprise us again.
So you have a high-bandwidth, but no better latency, L4 chip. And that's kind of hard to deal with and optimized against.
Xeon Phi had a DDR4 + HMC (where HMC was a early competitor to HBM), and getting good performance out of that was unusual. DDR4 ended up having slightly better latency, so the HMC cache would hurt some workloads.
They really only have to be competitive with the lower core consumer chips to accomplish their goal.
I think they should try for the performance crown though and the server business. There's a lot of money in that. In servers power consumption is an important consideration and they have an advantage there as well. ARM based architectures are also a lot more relaxed in that they don't have to present a nearly sequentially consistent view of cache and memory between multiple cores like x86. That may give them an edge scaling the core count. It's horrible to program for if you are coding lock free algorithms, but almost nobody does that and fewer should.
Their OS is pretty unsuitable for server tasks[1], and they've abandoned most of their server tools, so I think they don't care about that market anymore.
[1] No synflood mitigations, easy to panic by spawning threads that sleep.
There's a reason Apple don't make Time Capsules any more. Servers are usually commodities, and Apple don't make commodities if they can reasonably avoid it (anything they make that's even close to a commodity still, at least, can be turned into a fashion accessory, like the aluminum keyboards and stuff). Commodity suppliers get put into bidding wars, and operate on razor-thin margins.
Apple have basically opened the market for ARM servers by producing a workstation to go with them. Amazon is probably feeling pretty vindicated for Gravitron right now. Everybody else will catch up with them over the next five years or so, I'm sure.
[1] Excepting those big enough to build their own
I don’t think Apple will do it considering how far behind they are, but I wouldn’t be surprised if they start an AWS competitor using Apple Silicon.
Cloud computing is never 100% a commodity because there’s always business logic to sell.
Sure, S3 and EC2 are by this point commodities, but cloud products like Babelfish for Aurora PostgreSQL are not.
Businesses are plenty interested in single source cloud computing, there are tons of shops that basically can’t leave AWS without a monumental effort.
ARM chips are also commodities nowadays. I lost count of how many companies make them. Yet Apple chose to make their own.
Charging circuitry chips are also commodities and Apple bought the division of Dialog semiconductors making such chips to have in house knowledge.
Now Apple bought the modem division of Intel so we're gonna see GSM modems from them in the future.
Not in any non-Apple device, that is. The rest of the world will be stuck with with Qualcomm, Broadcom and some crap from Mediatek.
It's a high margin business, the opposite of a commodity.
Throwing in to that market with an ARM processor would limit you to portable software, putting you into competition with not only Intel and AMD but also Amazon and potentially Qualcomm and others. Which could actually hurt Apple if it gives ARM competitors a wider market for their processors and therefore more resources to spend on development that carries down to PCs and mobiles.
Also, they can sell you a monthly subscription to iCloud storage instead. I'd rather be in that line of business than selling an overpriced and under-featured NAS. Anyone who cares that much is probably technically savvy enough to prefer a proper NAS anyway.
Replacing the Intel machines with something that runs (a) faster and (b) with less power, is a double-win. I'd frankly be amazed if there wasn't a N-core or N-module server setup somewhere within the spaceship undergoing tests. And it's not that far from there to selling a Mac-based server. Which they've done before, after all...
Or, in Elon Musk speak, a "hardcore smackdown", with x86 playing the role of the gas engine.
My guess is that any multivariate binning will be lowest-common denominator, e.g. an 8-performance core / 12-GPU core part with a single, second-tier bin of 6 performance cores and 10 GPU cores.
They're fabbing with TSMC, which has a really established history in high yields. Just think about the A13/A14 CPUs, which are probably the most voluminous single CPU models every fabricated by a company, yet have never been subject to binning to other lower or higher end models. As the chips get more complicated naturally there will be more fabrication issues, so I expect we'll see more binning, but it doesn't look like there's much cause for concern.
[1]: https://www.anandtech.com/show/16028/better-yield-on-5nm-tha...
A quick look in ifixit's HomePod teardown (2018) reveals an Apple A8 processor, last seen in the iPhone 6 (released 2014): https://www.ifixit.com/Teardown/HomePod+Teardown/103133
I wonder how long they manufacture their processor lines for.
Yes, you bin the whole SoC; but all the performance differences are expected to be concentrated in the CPU and GPU. Expect a high bin where all cores and all GPU subprocessors are working, a medium bin where 1 or 2 CPU cores and 10% of the GPU is either missing or running at a lower frequencey, and a value bin where perhaps half the CPU cores and up to half the GPU are downgraded.
I expect these monsters to have oodles of HBM, and the Mac Pro with a TB or more could approach house-levels of pricing, like SGI workstations of the past.
After market RAM upgrades on Mac laptops haven't been possible for quite some time now, so I don't think a strategy to try and leverage more profit that way is a change. We'll have to see how they approach that for the higher end kit.
Pretty much everything they were paying Intel and AMD (for graphics) is now profit in their pockets.
The most comparable laptop processor is AMD Renoir: Zen2 8-core + iGPU for 10-billion transistors.
As such, I'd expect a "hypothetical high end" 32-big core Mac to compete against a 64-big core EPYC in terms of #transistors. Would the high end market be interested in a lower core count computer but higher single-thread performance?
I somehow doubt that. Server / workstation loads scale well to many cores: indeed, any computer that actually scales to 32-cores would probably also scale well to 64-cores (or SIMD-compute), and the M1 is only 128-bit wide SIMD.
------------
There's also the issue of the lack of hyperthreading: the Apple M1 does NOT implement SMT in any way what so ever. Those 32-cores would likely only support 32-threads, while the 64-core EPYC supports 128-threads.
I dunno. "Just scaling up" the M1 doesn't seem like it'd be a winner in the high-end market to me. Maybe if the core gained a wider SIMD-width and SMT (2x threads/core), that'd help a lot.
SMT isn't an unalloyed good, it helps for many workloads and hurts for some. That said, the very wide cores in M1 are asking for SMT to increass utilization, I'd guess it'll happen eventually. OTOH, the weaker memory consistency rules on ARM seem to allow for a lot more out of order execution, which seems to keep things busy.
I don't think we'll see Intel and AMD increase the width of their cores so much; they need their cores to scale to 5Ghz (even if it takes 1KW of cooling) for marketting and competition reasons, and wider cores are harder to clock so high.
I agree for sure. But the tradeoffs are pretty steep. Scaling within a core is traditionally sqrt(size), while scaling number of cores on a problem is also sqrt(cores).
Balancing between size-of-core vs number-of-cores is a big question. And that also depends very strongly on the task at hand. Web-servers scale better than sqrt(#cores), because each worker-thread is basically fully independent.
> I don't think we'll see Intel and AMD increase the width of their cores so much; they need their cores to scale to 5Ghz (even if it takes 1KW of cooling) for marketting and competition reasons, and wider cores are harder to clock so high.
I think there's a good argument that AMD / Intel can't scale their decoder much larger than 4-way / 6-way with uOp cache, due to the instruction set issue.
But I've seen some interesting innovations: the most recent Intel Atoms have double-decoders (2+2 decoder) on SMT-based Atoms. Such a thing would make SMT more efficient without much increase to power.
Its strange to see Apple take the single-thread performance crown. But... I don't think they've proven themselves in the high-end server / workstation market yet. Even if they do scale up, their decisions don't seem to be an obvious winner at the high end.
Is there even an API to take advantage of that yet? It looks like an impressive DSP but without an API, then that stuff is locked to Apple-only software / hardware.
> video coding acceleration
Is that really better than another 96-threads that you'd get with a 64-core Threadripper Pro? The thread-deficit of 32-core Apple vs 64-core/128-thread Threadripper is going to be pretty apparent.
When you have another +96 threads available, I'm not sure if "video coding acceleration" is even needed. Just run that in software.
Has been for at least three years now (maybe more): https://developer.apple.com/machine-learning/core-ml/
I guess that API is good enough for the deep-learning hype train however.
Of course other frameworks could leverage it too. As for direct access, I'm not sure. I think I did mix up API and framework in my response :)
https://developer.apple.com/wwdc19/718 https://developer.apple.com/wwdc18/701
Intro of CoreML: https://developer.apple.com/wwdc17/711
Being Apple I'm (sadly) not surprised they don't expose the Neural Engine hardware more directly.
> Those graphics chips would be several times faster than the current graphics modules Apple uses from Nvidia and AMD in its Intel-powered hardware.
I don't even remember the last Apple hardware with Nvidia graphics.
> Advanced Micro Devices Inc., which has been gaining market share at Intel’s expense, offers standard desktop parts with as many as 16 cores, with some of its high-end chips for gaming PCs going as high as 64 cores.
Threadripper is NOT a CPU for gaming.
> Company tests faster performing processors than Intel’s best.
Well if I'm correct AMD currently sits on top of the CPU ranking (and Intel 11th gen doesn't seem to be changing that as far as the leaks go), so AMD is the one to beat.
> Apple Preps Next Mac Chips With Aim to Outclass Highest-End PCs
Weird conclusion to draw considering the statements above. I would expect the performance to be on-par with the high end chips, but with lower energy consumption. The article seems to imply that it will outperform 280W+ CPUs.
Essentially all this article is really saying is "Expect Apple to release a half-size mac pro in 2022 as the last, most powerful new product". That's it. There's no clue about discrete graphics, NUMA, power budgets etc. It could be that that new mac pro is heading back towards the non-expandable trash-can style design. All they really know is that there's going to be something in the mac pro space in 2022 and that's the last direct replacement to Intel Macs Apple are making.
Even this piece of info can be extrapolated from statement's Apple has made public. They've said they were migrating their entire product line within 2 years. The Mac Pro is part of that product line. The only questions are how Apple is going to get there and on that front it's pretty clear this article is pure speculation.
Will the Mac Pro have integrated graphics or a discrete GPU? Maybe someone at AMD knows this based on order volume? Likewise, are they using external RAM, soldered on the logic board, or user-serviceable RAM?
How is performance going to compare to current top end processors?
Those are the only questions, and the answers here are pure speculation. It's likely only a handful of people at Apple even know the answers to these things.
All of the retina generation but for the very latest (« mid 2015 »).
I run two gaming VMs for my kids on my Threadripper server.
Do you have a link on how to setup something like that?
I use NVidia video cards for VMs, AMD cards (other than BigNavi) have VM reboot issues.
https://www.reddit.com/r/VFIO/comments/jturbd/vendorreset_ne...
I currently run a Windows gaming VM and a bunch of Linux VMs on top of Proxmox using a single GPU. Planning to add a second GPU for transcoding in the near future.
X<3|4|5>70 boards should allow 2 GPUs in separate IOMMU groups. My old system was a 1700X on ASUS Prime x370 Pro with 2 gaming VMs. Only issue with x?70 boards is the IOMMU groups for USB controllers. On the Prime x370 Pro, only one USB controller is in its own IOMMU group, so second VM needed a USB passthrough.
On the Threadripper, due to the vast amount of PCIe lanes, on the Asrock Taichi x399, I got two USB controllers in their own IOMMU groups and each of the 4 PCIex16 slots it's in its own IOMMU group. It's a dream virtualization motherboard.
While this is true for intel CPUs, this hasn't been the case for AMD Threadripper. The 2990wx has marginally higher single core performance than the 2700x (https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_threadri...), while the 3990x has only slightly lower single core performance than the 3700x (https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_threadri...).
The rest is just him trying to make a long enough article out of one sentence of information.
The article itself is essentially devoid of information and I'm surprised that it is on the front page here. Regardless the source, it is asinine to consider as news a prediction that Apple plans to continue in the same direction they've been going for some time. What's next, sun will rise again tomorrow, rain is wet, or Microsoft is planning a next version of .NET?
I was told this by a highly reliable source in Apple’s executive team, Tim Cook when he declared their roadmap openly at WWDC.
This is how I understand the article too. Bloomberg probably tried to recruit a source within the Apple Silicon team who wouldn't provide much information, but did say "well the Mac Pro will have a CPU better than any other high end PC".
Bloomberg wanted to run with that line since they have "unique" information so they could capitalize on it. So the author was told to write an article about it, with really only half a sentence of fuel to run with. So the author states the half-sentence, and then proceeds to use conjecture and speculation to fill out the requisite 500 words needed per the Bloomberg Style Guide that they are forced to file for submitted work.
And a Ford Mustang isn't a sports car. It's a muscle car. Regardless, both are faster than my civic.
Threadripper might not be the best gaming CPU but any threadripper would be an improvement over my current gaming rig.
So, you can beat highest gaming PC at gaming, but that doesn't reflect your power at running "productivity" workloads.
There are also a few games (Factorio) where ram limitations can bottleneck before cpu/gpu limitations, where Threadripper's extra memory channels can really help.
The Mustang isn't a 'muscle car', either: it's a 'pony car'.
this is a really unimportant distinction
I would like to see a world class GPU from Apple but again, I repeat this often, it won't mean much more than marketing materials if they don't get developers on board. This means more than just publishing oriented developers; video, pictures, and audio. That market is not sufficiently large to pull in other software as has been proven over the decades let alone get people to pay a premium for a system; because people do shop based on $$$ outside of gamers who wont' even look if there game isn't supported.
across Black Friday there were many good windows laptops at $500 or less with SSDs, some with dGPU, and most with 16gb ram. people see that and also that they know software just works and that is what Apple still is up against, hardware superiority or not
A pure gaming PC these days doesn't need a very powerful CPU
I've never heard any game complain they had too much CPU
Yeah, it's definitely non-public that Apple's about to put best-in-class chips in their desktop machines. Totally secret. Don't tell anyone, Bloomberg!
https://www.epsilontheory.com/why-am-i-reading-this-now/
Reminder: it was also Bloomberg who ran that very likely bogus and unsubstantiated story/psyop about the scary backdoor chips on the SuperMicro boards, and refused to retract it. When they do very obvious marketing fluff pieces like this, they lose even more credibility as a "legitimate news source" in my view, even if this ad they're running for Apple's upcoming hardware is 100% factually accurate.
Those that herald the M1 as unequivocally the next great thing are very, very rarely in the same (anecdotal) set of internet users that acknowledge that AMD exists, and makes CPUs. (Or it exists, but the M1 outperforms it in every metric!)
I suspect that many Apple-focused journalists are similarly constrained by tunnel vision, limiting their view of the world as "things that Apple created in the past" and "things that Apple creates now."
For the already Apple faithful, it's not a bad way to go. "Is this new Apple Silicon desktop going to outperform the only option I have now, an Intel Xeon-based Mac Pro?" So it's not too surprising that Intel is their benchmark, even if it ignores much higher performing alternatives available to those outside of Apple's walled garden.
It’s more likely they offer TR like performance at much lower watts, temperature and cost, like they’ve done to AMDa best mobile CPUs.
But overall, yeah, it's not as though the RAM is part of the chip, it's just part of the package. Not a lot to see here.
Just seems elegant to do 4 M1's and get 64gb total memory and 4x gpu/cpu and seems hard to scale up a single chip to match current multi chip / discrete gpu systems.
Maybe even dimms for main memory and current on package memory acting just as a very large level of cache.
Just because Apple currently have a (for the sake of argument) 20 watt part, that can outperform an Intel 45 watt part, it doesn't mean this advantage is inherent to ARM or sustainable once the tech cycle evens out.
The bottleneck right now for essentially all chip performance in mobile devices is heat. If you bypass the power throttling and fan control on my 10th gen 6 core Dell XPS 9500 and let it slurp 80-90 watts (and it will), and sit at 100 degrees centigrade (and it will) it will handily beat an M1 multithreaded score in Cinebench.
Single core performance beats even Intel’s highest TDP processors, I’m pretty sure AMD’s too, in real world workloads. Yes, AMD and Intel still have an edge when you get above 8 cores, but that’s kinda to be expected, no?
M1 has 16B transistors on 1 "chip", a 3950X has fewer than 10B across 3. What else would you expect, for now?
Actually this article is from Mark Gurman who has arguably the best track record of any journalist when getting the inside scoop on future Apple products. He has worked for Bloomberg for a couple years now.
If this is a supposed "good article" then I really don't want to see the bad ones...
This piece is written for a complete tech illiterate audience, if anything it makes me want to get an apple machine less
Read Anandtech if you want tech details, but then it's only for tech already in the market.
[0] - https://www.bloomberg.com/news/articles/2018-04-02/apple-is-...
But Bloomberg NEVER explained how a hypothetical 0402 capacitor could contain a backdoor. They literally took a picture of the smallest chip on a motherboard and said "this could be a backdoor" in big-bold letters. Even from a paranoid anti-China perspective, that article was awful.
"Chinese Backdoor in server equipment" wouldn't be surprising news to me (especially since important groups: military, telecommunications, etc. etc. use equipment like that). But when it comes to writing security-based articles, you need to be very specific about the nature of the threat. If you take the tiniest chip (a passive capacitor) and claim that there's a backdoor in it, you deserve to be laughed off.
That’s a very serious allegation.
They then went to a 2nd researcher to talk about "small chips", which is ambiguous. Some researcher was like... well... those 0402 ceramic capacitors are called "chips" (which they are: they're "ceramic chip capacitors").
The authors then got confused, and wrote about how 0402 capacitors were security flaws. Or something. I don't know, but that's how an article like that comes to be written in my head.
And now that the authors of the article are talking about stupid chip capacitors that can apparently take over the computer, they're completely off track and there's no reason to read the article with any level of seriousness.
-----------
> So are you implying that senior execs at tech companies were lying?
All I mean to imply, is that its bloody obvious that an 0402 chip capacitor can't hack your computer. That Bloomberg's authors couldn't understand this fact before publishing greatly diminishes the reputation of the article in question.
https://assets.bwbx.io/images/users/iqjWHBFdfxIU/i9VdsjZLS_P...
That's a 6-pin chip capacitor, is it not? Please, explain to me how a passive noise-dampening component can hack a computer.
Whether it's doing well or not doesn't really factor into that. That's more a matter of market and availability.
There are more than a few games in the iOS app store....
I think the grandparent means that using an M1 implies using macOS, which had its share of potential privacy issues recently (regularly phoning home on launching apps, having Apple service circumvent VPNs and application firewalls).
X86 is on a completely different level, see: https://01.org/linuxgraphics/documentation/hardware-specific...
None of the above is at the same level as phoning home or ignoring the user's firewall.
I put Windows telemetry into same bag as the Apple calling mothership.
You are really jumping through hoops to try to make a platform designed to protect your privacy equivalent to one designed to sell it.
Maybe that feature was implemented correctly after all and the objectives are different than what the marketing says.
This is not a privacy-vs-convenience battle, sadly; this is a very carefully planned circumstance where you can maintain privacy, or you can be effective in life. It's already impossible to use most mobiles in a private fashion, and within a few years, effective desktop computing will be exactly the same way. You're going to have to provide ID to do anything with a "normal" phone or computer.
Recently I learned that Apple will only give out NetworkExtension API entitlements for apps (that are required for VPN apps to work on macOS) to apps distributed via the Mac App Store. Aside from the obvious antitrust issues here (you can't self-distribute VPN apps, or run a competing app store that can distribute apps with the same functionality as Apple's), this means that you can't get access to e.g. the WireGuard app on macOS without IDing yourself to Apple to get an Apple ID and installing it via the Mac App Store.
Want privacy via a VPN? You're going to have to show ID first.
As has been discussed in various threads the last few weeks, the difference in power efficiency is not that large compared to recent AMD Ryzen APUs, accounting for a move from 7nm to 5nm node.
One could well argue that ARM/RISC is more elegant. But x86_64 besides backwards compatibility has the large benefit that it is an open platform when it comes to operating systems. Moving to ARM will be an excuse for many vendors to lock bootloaders to their blessed software (such as the iOS/Android ecosystems).
- The server and desktop markets move to ARM. The AArch64 instruction set becomes the default that people develop and deploy on. Optimizing for x86_64 CPUs becomes an afterthought. Intel only sells CPUs to organizations that have legacy systems somewhere that cannot be emulated. Intel has to enter the highly commoditized ARM market to stay competitive.
- The server and desktop markets largely stay on x86_64 as a result of Intel licensing x86_64 to three other vendors, increasing competition. The market is not commoditized, because Intel can decide who they license x86_86 to and also control extension to the instruction set. They lose market share due to the increased competition. But they can compete on their own architecture that they control.
Of course, these are extremes. But given that the current fastest supercomputer uses ARM and Apple showed that ARM is fit for the desktop, I am pretty sure that Intel is (or should be) worried about the first scenario.
In any event it's too late. They maybe had an chance of fending off Arm a few years ago - by the time any licensing now kicks in Arm will have established itself in the server / PC market.
So the only sensible strategy for Intel is to make the most of x86 whilst they can.
That’s not a feature of x86 itself. That’s just how the computers themselves are assembled. X86 machines could be just as locked down as Apple machines, if the vendor wanted them to be (and Microsoft tried!). ARM builds for multiple OSs exist, including multiple Linux distributions and Windows.
https://www.theverge.com/2020/6/24/21302213/apple-silicon-ma...
(In PRACTICE, you probably want to virtualise instead, because of a lack of support for the hardware.)
> As for Windows running natively on the machine, “that’s really up to Microsoft,” he said. “We have the core technologies for them to do that, to run their ARM version of Windows, which in turn of course supports x86 user mode applications. But that’s a decision Microsoft has to make, to bring to license that technology for users to run on these Macs. But the Macs are certainly very capable of it.”
He did a demo where he virtualized an OS. That doesn't mean you can't boot another OS.
When it comes down to it, Apple is still (largely) a hardware company. If someone wants to buy a Mac to put Linux on it, they still sold a Mac. And I think that having the Linux community testing their hardware would do nothing but make their hardware stronger.