I don't doubt that Intel has the smaller (and better transistor) at 14nm, however minimum gate size is only part of the picture. As an example, that Samsung 14nm part has a reported die size smaller than the Atom in the surface 3.
Samsung owns the high margin SOC ? Not for long.. Android based smartphones are soon to be a dime a dozen commodity product.
Swap out that expensive high margin SOC, for a $5 part and samsung saves $25 on it's BoM...Swap out that high end screen, for a cheap one, samsung saves $60 on the BoM...
[And yet Samsung's quarterly profit just plunged by 39 percent and Apple is over 7 bucks off of its 52-week high. You just don't seem to be able to grasp the cost side of the equation. Units sold don't necessarily reflect profits made. Perhaps the financial side is just too complex for your micro-architecture only skill set. The economics side of fabrication and the maturing market are killing ARM. Everything you mention relates to the high-end smartphone market and just for two companies. The big picture for ARM is very troubling.]
I have already shown that the 'cost' for the Exynos 7 are well within range (indeed, very similar to previous Galaxy S launches). As for profits, given the price of the S6, how much do you think Samsung makes on each one?
Yes, it's all high end stuff, which is typical when you move to new nodes. No, I mentioned 4 companies, Qualcomm, Samsung, Apple and MediaTek.
[And how long do you think it will remain at TSMC? Does the ARM Marketing Department allow you to mention that the transition has been planned for quite a while?]
"Like I say, ARM is one Apple-Intel deal away from collapse. And the first deal (modems) will happen in 2016."
So, how can a deal were the modem is manufactured at TSMC lead to collapse?
When will this part transition? And will Intel still win that socket?
[And the first deal (modems) will happen in 2016.]
Actually, if the rumours are to believed, in 2015 (but still on TSMC 28nm node).
[ARM will make an excellent 2nd tier technology semiconductor company producing a lot of fairly efficient 28nm products.]
A quick google says analysts are forecasting the following:
70 million S6 for 2015
200 million iphone 6 for 2015
So that's a quarter of a billion app processors on 20 or 14nm on those two products alone.
But that isn't the full picture, as a good chunk of those 70 million S6's have a 20nm modem. We're also missing other Samsung products that include 20nm app processors (such as the Alpha, A7 and Note), Apples ipad sales, sales from QCOM (such as 808, 810 and gobi modems) and other 20nm parts, from the likes of MediaTek.
So, perhaps upwards of 500 million SoC's on 20 and 14nm for 2015 and you say the ARMy will be stuck at 28nm. I can't imagine what the numbers would look like once they solve the "capacity" and "economics" problems.
Analagous to Gillette razor blades. When will it end?
Technically, quite a bit further with the new interconnects coming and the fact that the A53 is so small. As for when it'll end, when the consumer stops buying the hype. Afterall, 10 cores must be better than 8, which is better than 4...
...and you thought 8 cores where a waste of space...
"MediaTek Unveils Helio X20 Tri-Cluster 10-Core SoC" over on anandtech. Also rumors of a 10 core SoC from QCOM.
[Go away, denial boy. I'm tired of your stupidity.]
Can't face the truth eh? You are wrong.
[You're being an earwig. Again. Go away, I'm tired of this stupid game you're playing. You suggest that there isn't a problem except it is well-documented in the tech press. Go argue with ArsTechnica, pinhead. Your denial has reached epic proportions.]
YOU were the one who said the A57 had issues, not ArsTechnica. Do you understand what the difference between a CPU and a SoC is?
"So, was the problem with the A57 or not? "
["a chip which uses the same combination of Cortex A53 and A57 CPU cores"]
So, what is your answer? Was it the A57 or not? It's a simple yes or no. As a reminder, Samsung used the same reference design at 20nm and at 14nm, and neither of those designs exhibit the problems reported with the 810.
[If you still can't figure it out, let me know. As far as your "ARM has designs for the future", well - that's the future. Skip the marketing buzz.]
You said the following:
"Except Qualcomm has more problems than you can imagine with overheating and lack of performance as a result of using ARM's reference designs"
So, was the problem with the A57 or not?
"You claimed the problem was with the A57. "
[Please post up where I claimed this.]
You said "Except Qualcomm has more problems than you can imagine with overheating and lack of performance as a result of using ARM's reference designs". That reference design is the A57.
[Everything you mention is design. Once again ARM is trying to compensate for the fabrication crisis with design.]
Yes, they are designs. Just like Intel first has to DESIGN something before fabbing it! It's called progress. Nothing stands still in this game. All those designs I mentioned are for 14 or 10nm.
While it's certainly true that you can compensate to some extent for having a "worse process" but the inverse isn't true, a better process can't compensate for a bad design (ie, contra revenue with bay trail). If you have a bad process and a bad design, well, you have to look at AMD.
The ARM camp has twice invaded Intel's traditional markets, servers and Windows tablets. They were thrown out of Windows tablets and they continued to be ignored by the server crowd. In contrast, Intel has captured a meaningful share of the tablet market and is beginning to introduce fully capable smartphones, both which are core markets for ARM. Newer, more capable and affordable tablets continues Intel's advance in ARM's core businesses and leaves ARM no market unchallenged.
The ARM server story has years yet to run, and you shouldn't dismiss this threat so lightly - Intel certainly isn't (Paypal has recently deployed ARM based servers). Yes, Intel has done well in tablets, but without contra revenue, would it have done so well? We shall see how well Intel's based smartphones go this time round...:)
[Want to explain Qualcomm's huge continuous negative press on the issue? Your approach is to pretend the problem doesn't exist? Denial is thy name.]
No. I am not denying anything. You claimed the problem was with the A57. I just asked you to explain why Samsungs 20nm and 14nm processes (which use the exact same core) don't have these issues?
[Want to explain Samsung's plummeting profitability? Intel's fabrication superiority will allow them to discontinue Atom in favor of the better performing line. What will you offer up then? More marketing mantra? ]
So, are you suggesting that Intel is going to start selling Core for Atom prices? The ARMy is already moving forward, it's not marketing. In terms of high end CPU cores, from ARM, you'll have the A72 , Ares and Prometheus. From Qualcomm Kryo, from Samsung you'll have 'mongoose' and Apple you'll have the A9.
[A while back you would have said why doesn't Intel have significant tablet share. A while back you would have said why did Microsoft choose to offer RT. A while back you would have said Intel will never match ARM in power efficiency. A while back you did say that there would never be an Intel based Windows phone. Things change, dufus. You are on the wrong side of the prediction curve.]
lol - me on the wrong side of the prediction curve, yeah right!
[What is this? Your new marketing mantra?]
Yes, do you like it?
[Except Qualcomm has more problems than you can imagine with overheating and lack of performance as a result of using ARM's reference designs.]
Want to explain why Samsung didn't have an issue with the same ARM's reference design at 20nm and at 14nm? Want to explain why Qualcomm is using ARM's latest core (the A72) as well as designing its own high end one?
[Intel is now killing ARM at the low end. The WinTel alliance is coming back and ARM has no where to run and no place to hide.]
Intel has just launched a product. Let's wait until we see some sales numbers. Want to explain the rockchip relationship and why Intel are using ARM's GPU ? The WinTel alliance is coming back? Really? Want to explain why Microsoft is still using Qualcomm for its Windows 10 phones?
[ARM's fabrication continues in crisis with new nodes that cost too much and offer mitigated performance improvements.]
Want to explain how Samsung's Exynos 7 is smaller in die than Intel's latest 14nm Atom, yet it still offers more CPU and GPU performance?
[The ARM world is looking pretty AMD-ish to me.]
[You must be buying #$%$ TV's.]
How much do you think an Apple phone actually costs?
So, basically you're saying that Intel can't compete in this market given their business model v ARMs.
Short, honest answer. No. That said, you never say never in tech...
Also, while I'm a 'bear' on Intel's mobile ambitions, I am a 'bull' for other parts of Intel's business (servers) and if Windows 10 is good (from a consumer point of view) there might be a nice uplift for PC's as well.
As a side, while the SoC in Apple's devices are 'cheap' ($20), the phones themselves aren't and can certainly cost more than a TV.
Who knows? It may not be very much as the Fabs that do business with end users based on ARM designs may have to either charge end users less, or reduce ARM royalties. Even a 10% reduction may hurt. Every little bit helps.
How about QCOM, Apple and now (supposedly) Samsung who do their own CPU designs? Is Intel going to be able to "diminish their capacity" as well?
With the above 'top' end designs, along with the various ARM CPU designs you have a variety of CPU designs across processes all hitting different feature and price points. It's going to be hard for Intel to hit the right "target" year after year to enable this reduction in R&D capacity that was suggested by ideal_invst. As a shareholder, how much are you prepared for Intel to lose year on year in this endeavor? The ARMy isn't AMD:)
FYI, ARM earns (typically) between 1 and 2% of the SoC's selling price.