The machine is blind to the complexity. It builds what you give it instructions to build. But, this is a system where, just randomly picking three units on the die ,you could literally have one chip dead (or severely impaired), one chip that meets the specs, and one that WILDLY outperforms your best expectations of what should be possible. In a system like this, the simpler the thing you’re trying to produce, the greater your chances of success.
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.
A14X Chip for First Apple Silicon Mac and New iPad Pro to Enter Mass Production in Fourth Quarter
- Thread starter MacRumors
- Start date
- Sort by reaction score
This still doesn't explain why the complexity causes lower production yields. For instance, here's a simple pattern:
And here's a more complex one:
If the error rate per grid location is the same for both, then there won't be any more errors in the complex pattern, on average, than in the simple one. Perhaps what you have in mind is that the more complex chip is less fault-tolerant. I.e., an error in the more complex chip is more likely to lead to a fatal flaw than one in the simpler chip.
In any event, your original general observation is very interesting, and one I've not heard before: The fact that Intel is lagging TSMC in device minaturization does not, as most describe it, mean that Intel is behind TSMC in fab technology. Rather, it's that what Intel is trying to fab is more difficult.
Might you happen to have some sources that corroborate this? I ask because other articles I just read have given different reasons -- e.g., that Intel was late to go to EUV, and was instead trying to miniaturize using its DUV process.
Last edited:
So it sounds like you want a Mac Pro with a Core i9 instead of a Xeon 32xx? Even if you reduced the PCIe slots to four and RAM slots to six, with a smaller case/power supply, that’s still a $5,300-5,500 machine isn’t it?
And given the R&D, testing, manufacturing and ongoing SW and HW costs of simply designing, manufacturing, and lifecycle carrying costs of an additional model in the lineup (which is unlikely to expand Apple’s sales more than slightly imo) I wouldn’t be surprised if it would have to sell for $6k, or more. Hard to say. But it’s not a $2k, $3k, $4k or even a $5k machine, as far as I can tell.
So there might be some unknown demand for a Mac that Apple—who does a ton of market research and does know the demand—apparently has zero interest in making.
Last edited:
Never said I wanted it; wouldn't fit my use case right now (though it might in the future). Just trying to articulate what those who want the xMac are looking for, and why they want it; regardless of whether Apple will make it, I don't think their desires are unreasonable.
I don't know the parts and assembly costs and needed margins, so here's my approach: Start with the iMac.
A 27" iMac with the top processor (Core i9-10990) and GPU (Radeon Pro 5700XT w/ 16GB GDDR6) options, 1 TB SSD and base RAM (most will order their own rather than pay Apple's prices) is $3400. To turn this into an xMac, we need to change its existing case into that for a small tower. Let's call that a wash. Then we need to remove the 27" 5k panel and associated circuitry, and add an upgraded processor (Core i9-10990K, which retails for $100 more than the i9-10990) upgraded PS and fans, two PCIe slots (one for a GPU, one for a 2nd storage drive), and two to four more RAM slots.
My guess is that the 5k panel, and its assocated circuitry, costs more than what I just listed. So, at the same profit margin, it's about a $3k machine. The R&D costs for a simple machine like this should be much less than that for, say, a new MBP or iMac. And, countering the per-unit development cost of a potentially lower-volume product is the intangible benefit of strengthening the ecosystem, since a machine like this would probably be popular among developers, engineers, scientists, and independent video pros who can't afford a Mac Pro.
I've never argued Apple would make it. Indeed, the first idea I presented about this, which started this whole discussion, is that Apple likely won't, because it could cannibalize iMac and Mac Pro sales.
Last edited:
It has nothing at all to do with cannibalization imo, it’s that nobody would pay what that xMac would cost them. That machine (that you don’t want, at least not right now) would be closer to $5,500 imo. You suggest a slightly different Mac Pro—not much different—but you (not you) don’t want to pay for it.
The Xeon in the base unit lists at $749 while a high end i9 is something around 500-600 list. I doubt Apple’s price difference between an 8-core i9 and and 8-core Xeon would be more than $150, maybe I’m wrong.
Four fewer PCIe connectors saves maybe $5, 6 fewer RAM slots maybe another $5. A somewhat smaller PCB board maybe $20. You wouldn’t need a 1,400W power supply, but with 2 GPU slots you’re still looking at at least 1,000. So maybe you save $25-35 there?
The case is where you’ll save the most, who knows maybe a smaller case would save $50 or $75 or so? (More? Less? Just an educated guess on my part.)
So yeah, you’ve got a cut-down Mac Pro (that you don’t want), and Apple’s saved a few hundred bucks in parts. But saving a few hundred bucks in BOM cost doesn’t equate to a $3,000 discount.
imo you’re looking at a selling price somewhere around $5,300-$5,500. Nobody buys that box, even at $5,000. They’d get the full-blown Mac Pro for $6,000. And over the following 5-7 years they’d probably be very glad they did.
Last edited:
There was no need for the repeated snarky '(that you don’t want)' comments.* They make it seem like you're more interested in trolling me than in having a serious discussion. Which is disappointing because, in the past, I've generally regarded your posts as containing serious content.
Contributing to the appearance of non-seriousness is pricing out an xMac by starting with the base Mac Pro and subtacting a few components which, it should be obvious from the start, is inherently is going to lead to a skewed conclusion.
The reason this approach doesn't make sense is that the Mac Pro features a pro-grade frame designed to support four GPUs, 1.5 TB RAM, a 28-core Xeon, multiple drives, and other internal add-ins. This qualitatively different grade of machine than what the x-Mac proponents want. They, by contrast, want a consumer-grade machine—much more akin in parts grade and construction grade to the iMac. Hence my analysis. Which you've carefully avoided addressing.
*Why am I arguing the x-Mac enthusiasts' views even though I don't currently have a need for it myself? It's because I often see their views treated derisively on these forums—as if their desire for the machine is foolish, independent of whether Apple will build it or not—and I think it's important to stick up for those that are unfairly treated with derision, regardless of whether I am among their number. Hence I am arguing that their position is a reasonable one—that Apple *could* build that machine at a sellable price (but won't, because of cannibalization).
Last edited:
My “that you don’t want” was a direct response to you repeatedly claiming such. For someone who doesn’t want an xMac, you sure do spend a lot of time thinking, posting and arguing about the xMac 🙂
Starting with the Mac Pro for xMac pricing makes complete sense, because that’s the tower that Apple actually manufactures right now, and can make money selling (presumably). Yeah it’s an awesome case, but that’s not why it costs $6k.
It’s just math. It’s a niche product, with low volume, and can’t be equated to the iMac platform that sells a few million a year.
Apple sells a the full tower for $6k, and that includes their profit margin at the quantity they’ll be able to sell (maybe 300-500k?). That may only be $1,800 in BOM cost, who knows. But to make money, Apple has to price it at $6k. If they could sell it at $5k or $5,500 they would.
There’s just no possibility that an xMac that might have a parts cost of maybe $1,300-$1,500 instead of the Mac Pro’s $1,800 could be priced any lower than $5,000. The fixed costs of carrying such a niche product simply don’t allow it.
Would their be cannibalization at $3k? Sure! Because it would be underpriced by $2,000. It would be like selling the iPhone Pro at $600 and the Pro Max at $1,100, when they’re really not that far off in cost to manufacture.
There’s absolutely nothing wrong with wanting an xMac. The only thing “wrong” here is wanting it at a price that loses Apple money.
I’ve had this discussion so many times with different posters, I’ve got limited patience for it at this point, and I apologize for that. I know I can be irritating and short and overly sarcastic, and I know it doesn’t play well sometimes. I apologize again.
Last edited:
I don't recall every post I've made on macrumors, but I just did a search through my entire post history, cross-referenced with "xMac", and the only hits I got were on this thread. So your comment is erroneous. And the whole thing started because I made a passing one-sentence comment about cannibalization, using the x-Mac as an example, and three successive posters decided *they* wanted to argue with me about it (you being the most recent of the bunch). So it sounds like the ones that really want to argue about it are you guys, not me. I'm just responding, I'm not initiating.
Indeed, I also did a search cross-referencing your user name and "xMac", and I see you've commented about this on this on 9 (!) different threads. In summary--me: 1 thread; you: 9 threads. So I think you're projecting here.
I appreciate the apology, but note that you chose to come after me on this, not visa versa. If the issue so bothers you, and you have so little patience for it, such that you knew you couldn't discuss it without behaving poorly, then perhaps it would have been best not to attempt to critique me in the first place, yes?
But if you can put that aside: Your low-volume argument against my iMac analysis is contradicted by your later statement, that "Would their be cannibalization at $3k? Sure! Because it would be underpriced by $2,000." You can't have it both ways. Think it through: You can't say a price of $3k isn't plausible b/c of low volume, and then say a $3k price is a problem because, at $3k, it would cannibalize sales from other products, because cannibalization wouldn't be an issue with a low-volume product! Again, there's a logical contradiction here.
Last edited:
Just because I’m exasperated by posters not understanding how quantity, pricing, features and perceived value drive revenue, profit and resultant price tiers, doesn’t mean I’ve given up on trying to explain why people can’t have a $2k or $3k xMac 🙂But I do have less patience for it!
There’s no contradiction in my post, it’s simply a multifactorial system of equations. You can’t ignore the expected volume nor the lack of profitability at the arbitrary price you’d be willing to pay.
Building an xMac and selling it at $3k with a $2k loss (at the expected qty) is a light year from being able to build it, sell it, maintain the product for a 5-7 year lifecycle, improve it, and support it for $3k.
You want a $5k machine, but you’re not willing to pay for it. This is effectively your argument: The xMac should only cost me $3k because that’s all I think it’s worth, and Apple would still be fine, because I say so”. That isn’t anyone’s definition of cannibalization. Yet that’s apparently what you think, from what I understand of what you’ve written.
I’ve no doubt there are probably another dozen ways for you not to understand that Apple would lose money on the price you’re willing to pay for an xMac, but I’ve reached my limit and must therefore disengage from our discussion.
I can explain it to you, but I can’t understand it for you, yes? 🤷♂️ If you find that an unacceptable critique... well, sorry, not sorry.
You're just making stuff up here. I never said that, so this is an obvious strawman argument. And when someone resorts to such games, it's an acknowledgment they can't argue based on the substance
No one who actually does any sort of math-based modeling talks this way (throwing out disconnected jargon the way you do). You're pretending to have some sophistication in this area , while just delivering empty jargon. And yes, there is a contradiction in your post. And I now see I'm the one who is not able to get you to understand what you can't understand.
Yes, that would definitely be your best recourse. And please refrain from commenting on any of my posts in the future.
.
Last edited:
That's not what it boils down to; that's just how one side characterizes it. To sum up how an argument boils down, you need to present the essence of both sides (or all principal sides, if there's more than two).
Last edited:
If Apple and TSMC can achieve i9 performance with an ARM RISC processor it makes you wonder what they could do if the architected and produced a CISC processor.
Intel's architecture for a while has been in a sense RISC covered by a CISC backwards compatibility layer, it's more complicated than that as always but there is essentially a translation layer that takes the stable CISC instructions and converts them to the internal RISC architecture.
Where Apple have an advantage in a way that Intel doesn't is that it isn't tied to backwards compatibility for their instruction set. Apple have in the last two decades pulled off supporting a large number of architectures and implemented significant shifts (PPC 32 + 64 to Intel 32 + 64 to soon Apple Silicon; within the ARM space they've gone from ARMv6 to ARMv7 to ARM64/ARMv8). Apple have implemented custom instructions that they hide behind their Accelerate framework that also allows them the ability to innovate in hardware in a way that Intel could never achieve. Apple are not afraid of removing support for legacy that holds them back (Intel 32-bit, PPC before then, the various ARM transitions on the phone and watch) which enables them to focus on new functionality over compatibility.
If this wasn't enough, Apple's ownership of their APIs and operating systems tied to that same hardware allow them to start adding capabilities to existing programs that leverage their APIs that transparently makes use of the new hardware capabilities when available on the hardware the app is running on. Accelerate is an example of that as is Metal.
Where Apple have an advantage in a way that Intel doesn't is that it isn't tied to backwards compatibility for their instruction set. Apple have in the last two decades pulled off supporting a large number of architectures and implemented significant shifts (PPC 32 + 64 to Intel 32 + 64 to soon Apple Silicon; within the ARM space they've gone from ARMv6 to ARMv7 to ARM64/ARMv8). Apple have implemented custom instructions that they hide behind their Accelerate framework that also allows them the ability to innovate in hardware in a way that Intel could never achieve. Apple are not afraid of removing support for legacy that holds them back (Intel 32-bit, PPC before then, the various ARM transitions on the phone and watch) which enables them to focus on new functionality over compatibility.
If this wasn't enough, Apple's ownership of their APIs and operating systems tied to that same hardware allow them to start adding capabilities to existing programs that leverage their APIs that transparently makes use of the new hardware capabilities when available on the hardware the app is running on. Accelerate is an example of that as is Metal.
The much better question would be, what INTEL could do if they could drop their performance sucking decoders that are required ONLY for backwards compatibility? Of course, they can’t, they have to drag that cruft into the future with each generation. BUT if they could...
I just want the A14X iPad Pro right now! 😉😂 ***a real upgraded iPad Pro. *Better louder speakers, Dolby atmos with really good bass, mini LED, ProMotion, same 5.9mm thinness, 7000 aluminum, new iPhone 12 glass hard shell. same dual camera but it is fixed to look better, (more like iPhone 12 does.)
I doubt Apple is designing for your tiny niche market. Most of Apple's Mac revenue (and thus profits, and thus their primary design target) is currently based on the 2 or 4 core Intel chips that go in MacBook Airs, Pros and mid-range basic iMacs. So anything more than 4 cores is gravy. Maybe some future A chip (for 2022 or 23) will be targeted at the Mac Pro.
Doubtful Apple is aiming at a floor of just 4 cores. That is actually in part Intel's constraint. Basically borders on incompetent competitive analysis at this point to not include where AMD is going with future x86 Windows laptop systems (AMD is delivering 8 cores for performance mobile solutions). Besides, the A12X (A12Z) is already past the 4 core count .
" All eight cores can operate concurrently "
https://en.wikichip.org/wiki/apple/ax/a12x
Apple isn't going to go 'backwards'. So "more than four" isn't much of a prediction since Apple is already past that threshold.
The notion could be use that only the "Performance" (big) cores of the Apple Silicon count as 'real' cores in these cross vendor product discussions. That the "energy efficient" ( little) ones don't count. But the 'little' cores do make an impact on multithreaded generic benchmarks that people like to use in comparisons across CPU package product lines. Apple's marketing pages will probably use that combined core count as 'hay' to build a hype bonfire. Intel is likely going to be doing something similar 2H of 2021 also ( as they roll out big-little to compete in the "core count" wars in the mainstream. )
The rumors ( via Bloomberg with good track record ) are that Apple upped the game with 5nm and now the "iPad Pro" class SoC is in the 12 core ( presumably 8 Performance 4 energy focused) but if a bit cheesy those could be swapped in groupings. ) cores. That will match up AMD's 8 performance with Apple's 8 performance. [ there is likely a Gen 11 (Tiger Lake) SoC from Intel too that is just lower priority at the moment. Something to counter AMD iterative update probably coming in early 2021 and when can couple to volumes of LPDDR5 DRAM. )
Probably not going to be labeled an "A" series chip. The Mac Pro SoC is probably going to be substantially decuopled from the phone series chips. Not as much the core count , but also the I/O paramaters are about an order of magnitude bigger. Quad digit GB RAM capacity (versus single digit). In the range of 48-64 PCI-e v4 lanes of interpackage bandwidth ( where the A series is order of magnitude lower than that ). The lower end laptops may overlap highly with the iPad Pro class A-series implementation, but the desktop class (at least for 27" and 'up' ) is probably going to decouple from the A-series as to what is significantly in the SoC package. Powering one-port-wonder iPhones/iPads gets substantive savings in ignoring I/O.
Since Apple gave themselves as two year deadline to 'lfip' the Mac line up , 2023 probably isn't likely. Unless Apple runs into a major development hiccup. ( Possible. ) . The Mac Pro SoC that Apple will do also likely will get an iGPU. ( just Apple's thread of Unified Memory and also the "natively run iPad/iPhone apps mindset. ). IMHO, that will probably cap the max core count at under 32. Throw on top 4 of those being energy efficient and likely looking at 28 cores that are incrementally better than the 28 the Mac Pro 2019 has.
Additionally, probably looking at 2022 when the smaller Apple SoC packages move on to the 5p nm and/or 3nm process that the Mac Pro is still sitting on 5nm. The Mac Pro class SoC will iterate much slower. ( maybe a 3 year cycle, but given Apple's track record in this product space could be as long as 4-5 years. )