Marginal market?!? While Mac sales may be smaller than iDevice sales, I will point out that Apple did $26 billion in Mac sales last year. That’s almost 3X the market cap of AMD.
Marginal market?!? While Mac sales may be smaller than iDevice sales, I will point out that Apple did $26 billion in Mac sales last year. That’s almost 3X the market cap of AMD.Yeah if you start with 1 horse is easy to get to 10 hp. however if you start with 300hp it's a lot harder or shall I say cost prohibitive to hit 3000hp. Relative comparisons sometimes really are meaningless. If your 10% gain means a higher absolute performance gain than the competitor doubling in performance, who actually is the winner? Hint: the gap increased.
Note: I'm not defending Intel just saying it's always easier to have huge relative gains if you are starting from a very low baseline.
Why invest into chip design for a marginal market? Mac market is tiny compared to iPhone and iPad meaning you now have to divide the chip design effort over a rather small amount of devices. Not worth it. Probably cheaper to put pressure on Intel and I'm sure Apple does so you know, by spreading rumors like this.
LOL Is Apple a Chinese company yet or are they US and just making all their products in China?Not happening, Apple is not manufacturing anything on its own nowadays. But they are supporting their manufacturing partners to some degree.
Especially since one horse produces about 13 hp.Yeah if you start with 1 horse is easy to get to 10 hp. however if you start with 300hp it's a lot harder or shall I say cost prohibitive to hit 3000hp. Relative comparisons sometimes really are meaningless. If your 10% gain means a higher absolute performance gain than the competitor doubling in performance, who actually is the winner? Hint: the gap increased.
Note: I'm not defending Intel just saying it's always easier to have huge relative gains if you are starting from a very low baseline.
Why invest into chip design for a marginal market? Mac market is tiny compared to iPhone and iPad meaning you now have to divide the chip design effort over a rather small amount of devices. Not worth it. Probably cheaper to put pressure on Intel and I'm sure Apple does so you know, by spreading rumors like this.
Marginal market?!? While Mac sales may be smaller than iDevice sales, I will point out that Apple did $26 billion in Mac sales last year. That’s almost 3X the market cap of AMD.
Depends on how important single-thread, general application performance is.It's not the sales figure that count but number of devices. The fewer devices, the more each chip will cost.
Mac: roughly 20 mio per year
iPhone: 50 Mio
iPad: 45 Mio
However this doesn't tell the whole truth as Macs use many different type of processors from dual-core laptops to 18-cores in the Mac Pro. So I would say to cover this range apple would need to make at least 3 designs or completely abandon certain markets. I guess the amount of 18-core Mac pros sold is rather tiny. It would simply not be possible to still offer such a product. That could very well be their strategy, but it shows the downsides of moving over to in-house ARM chips for Macs.
Why invest into chip design for a marginal market? Mac market is tiny compared to iPhone and iPad meaning you now have to divide the chip design effort over a rather small amount of devices. Not worth it. Probably cheaper to put pressure on Intel and I'm sure Apple does so you know, by spreading rumors like this.
They make more money from Mac than from iPad, $6.9bn vs $5.9bn per quarter. They can subsidise development of chips for the iPad by reusing them in Macbooks.
EDIT: Bear in mind that Apple don't necessarily need to make their own chips for every Mac. They could buy in server-grade CPUs from Qualcomm or Cavium to use in their Mac Pro replacement.
Revenue isn't making money. I can have a revenue of 100bn and still loose money.
Second point is entirely possible but if you are still going to buy the CPU/SOC from 3rd party why even switch away from Intel? Seems much more likely they will copy AMD by scaling using an MCM. But how that would work with graphics is a whole other question. Also does nv /amd even provide drivers for ARM systems? Not that I know of meaning Apple would also need to make their own GPUs (yes I know, probably already in the works)
Still it seems like a huge effort for little benefit.
I find it very strange people think Apple SoC aren't any good for desktop workload.
https://browser.geekbench.com/v4/cpu/compare/7795892?baselin...
This is an Macbook Top Spec from Intel on a Fanless Design, against an iPad Pro also a fanless design. Let's ignore the small difference of TDP for a minutes because both are limited by its Fanless design, and Geekbench do runs on for a few minutes, so it is an indication or how close they are in these specific workloads.
The Intel Core i7-7Y75 can turbo boost to 3.4Ghz if needed, comparing that to max 2.4Ghz of the A10X. If we also ignore the frequency, iPad Single Core performance is with 10% of Intel's best Core processor at this TDP. The A10 core here is a generation older then the A11 used in iPhone X and iPhone8, and A12 coming in this September will likely have even better IPC.
The JavaScript testing performance on both Mac and iOS Safari shows performance is similar and A11 even edge out Intel in some cases.
So no matter how you spin it, in certain workload Apple A11 has already matched up or exceed Intel within a Fanless TDP design. That is excluding the advantage Apple will have on Multicore when it has 4x Core compared to 2 Core 4 thread in Intel.
Intel's 14nm++ is also matured and better then then TSMC 10nm, which is merely a testing node for its 7nm.
i.e Assuming Apple wanted to, they could have a 7nm Quad Core A12X shipping in 6 months time that is better then Intel's Core at Fanless TDP or <15W design, at 1/3 of the cost.
The reality is Intel hasn't been executing its plan for a few years. And If this rumours is true it is only themselves to blame. When Apple were designing its 2015 Macbook, Intel's roadmap were clear, three years later in 2017, the Macbook was suppose to have Quad Core in its design. And looking at all the latest roadmap, it doesn't seem there will be a 10nm Quad Core CPU shipping this year either. Not in fanless design. We are looking at 2019 march, four years since its introduction to get that through.
Then there is the LPDDR4 memory support. You wanted 32GB Laptop Memory? Well Intel doesn't allow you to have it. Not even in 2018. The delay in 10nm, the little to no improvement in IPC, AVX2 is very much a niche. As a normal customer I dont really have a reason to buy Intel anymore unless you want absolute Price / Single Core Performance, otherwise AMD is a much better choice. So I dont think we are the one who are frustrated. Apple is likely too.
I also wonder how this moves means in the Modem space. Which is the much more important pcs for Apple. I have often argued one reason Apple didn't make the Mac to ARM move earlier was because Apple needed Intel's modem to fence off Qualcomm's "double dipping". Now that Broadcom failed to acquire Qualcomm, Intel is being dumped from Apple. Does that mean Apple is going back to Qualcomm modem soon? Likely along with a deal of using Qualcomm Centriq on servers and likely later in Pro lineup of Mac?
Previously Apple's rumoured Project McQueen were to bring all the Server inhouse, using less resources from AWS, Azure and Google. Apple is a large Server customers on its own. If Qualcomm has Apple to kick start its Server CPU business, along with Apple push to help PC transition to ARM, it may be well worth Qualcomm to lower its royalty and modem prices.
Yeah if you start with 1 horse is easy to get to 10 hp. however if you start with 300hp it's a lot harder or shall I say cost prohibitive to hit 3000hp. Relative comparisons sometimes really are meaningless. If your 10% gain means a higher absolute performance gain than the competitor doubling in performance, who actually is the winner? Hint: the gap increased.
Note: I'm not defending Intel just saying it's always easier to have huge relative gains if you are starting from a very low baseline.
Why invest into chip design for a marginal market? Mac market is tiny compared to iPhone and iPad meaning you now have to divide the chip design effort over a rather small amount of devices. Not worth it. Probably cheaper to put pressure on Intel and I'm sure Apple does so you know, by spreading rumors like this.
So what I envision is using the same chip family across the iPhone, iPad, iPad Pro, Macbook, Macbook Pro and standard iMac (running at higher frequencies in the desktop, maybe more cores and a wider memory interface, but otherwise the same CPU/GPU family as the iPad and iPhone).
The chip family is irrelevant when it comes to taping out and creating all the expensive masking layers. Each chip needs it's own tape-out and expensive masking layers no matter which family it is in.
You end up taping out/masking nearly as many models as Intel, and if you believe Bloomberg, Apple is only 5% of Intels business. Intel gets to amortize those tape-outs/mask among 20X the market that Apple will.
Also you now have Apple spending multiple times the development work on chips for Macs when they are a fraction of of the business, of the iPhone, where Apple only does ONE chip. That doesn't make a lot of sense. There will need to be massive cuts to the Mac lineup to limit all this chip development.
IMO if Apple makes this transition, it will be accompanied by Slash and burn of the Mac Lineup. The High end will just get dumped (likely many will just flee at the first sign this is true).
They will build two Mac chips:
Mac Low Power: MLP
Mac High Power: MHP
The iPad chip will be used in Low End laptops
MLP in upper range Laptops.
MLP in low end iMacs
MHP in upper range iMacs.
There will be no Mac Mini or Mac Pro.
There will be drastically less Mac Models, and lower sales.
I don't think that there will be any need for a "MHP" chip designed by Apple, and I don't think that they would need to kill Mac Pro. Just buy server chips from Qualcomm or Cavium for the "MHP" stuff.
Isn't Geekbench one of the worst and most overused benchmarks for testing CPU perfomance? Personally I think it is extremely ridiculous to only use and rely on a single benchmark that really doen't test anything.The only people who still think Intel desktop chips are faster than Apple chips are people who not only know nothing about CPUs and technology but also refuse to read or understand benchmarks. They avert their eyes from geekbench and quickly run to hide when it appears like it was the sun and they are some kind of vampires.
So, Apple is so desperate to flee Intel and their crappy chips,that they run into the arms of Qualcomm and their crappy chips?
I thought the point was Apple controlling it's own destiny and bringing everything in house?
We can also look at AMDs approach with Zen: with just one die they can cover 4C - 32C. Add the recent KBL-G experiment for graphics IP and there goes to road towards covering everything from fanless to workstation. It may not be the immediate solution, but the trend is clear.Instead of dealing with a vendor's crappy chips on all of their product lines, they only have to deal with them on the super low-volume niche products that they don't particularly care about. (See also: Mac Pro not being updated since the last Ice Age.) The high volume products that they actually give a **** about would be on Apple chips.
Isn't Geekbench one of the worst and most overused benchmarks for testing CPU perfomance? Personally I think it is extremely ridiculous to only use and rely on a single benchmark that really doen't test anything.
There are good reasons to sneer at GeekBench.
Even so relying on a single benchmark is rather shortsighted. CPU(and GPU) performance should be tested using a wide variety of benchmarks, not just only one.You might be thinking about Geekbench < 4. Geekbench 4 has been recognized as a pretty good benchmark by a lot of informed people, including Linus.
You said :It's not the sales figure that count but number of devices. The fewer devices, the more each chip will cost.
Mac: roughly 20 mio per year
iPhone: 50 Mio
iPad: 45 Mio
However this doesn't tell the whole truth as Macs use many different type of processors from dual-core laptops to 18-cores in the Mac Pro. So I would say to cover this range apple would need to make at least 3 designs or completely abandon certain markets. I guess the amount of 18-core Mac pros sold is rather tiny. It would simply not be possible to still offer such a product. That could very well be their strategy, but it shows the downsides of moving over to in-house ARM chips for Macs.
Even so relying on a single benchmark is rather shortsighted. CPU(and GPU) performance should be tested using a wide variety of benchmarks, not just only one.
Shouldn't CPUs and platforms be judged by the actual applications and usage in the first place? I'm not sure if synthetic benchmarks even have a place at the table if at all.Unfortunately there isn't much else that is cross platform, and Geek-bench will always be somewhat tainted as they used to be completely different benchmarks depending on your platform, making them completely untrustworthy as a cross platform benchmark.
If you look at new CPU reviews for AMD/Intel. No one cares at all about Geekbench. They care most about application and gaming performance. Here is a recent AT CPU review and the benchmark overview:
https://www.anandtech.com/show/12425/marrying-vega-and-zen-the-amd-ryzen-5-2400g-review/4
A huge page of benchmark overview. Most of which are application/usage/gaming tests.
No Geekbench, because it is a pointless synthetic benchmark.
Shouldn't CPUs and platforms be judged by the actual applications and usage in the first place? I'm not sure if synthetic benchmarks even have a place at the table if at all.