Atom is moved forward. DT broadwell is cancelled. Looks like Intel took 14nm capacity away from broadwell to get 14nm Atom out sooner.
Yes I see this as the motivation behind Intel doing it.
What other option do they have to fight Arm? Arm accomplishes "computing" with a much, much lower financial overhead than Intel.
They've already got the market share in phones, which will soon handily dock to a station and replace 90% of peoples' needs for a laptop or desktop. (Though even with atom this paradigm would negate the need for a desktop.)
1) What's the price of intel chips?
2) Are OEMs willing to give up freedom for more speed or slightly more battery life and end up like the PC OEMs battling for stupidly low margins with only a big guy eating the majority of their bom?
3) Do people need more speed on their phones? We're quickly approaching fast enough levels of performance and manifacturers need to just stop using anemic batteries (1200mah)
I'm using a phone running a dual core 1ghz A7 and it's almost there for me (facebook app loading time is my main gripe), factor a 2050mah battery and I get two days out of my phone with light usage and active data connection... that's a 130 huawei phone.
The biggest power hog of any phone is the screen, not the cpu.
AtenRa
Lifer
- Feb 2, 2009
- 14,003
- 3,362
- 136
If im not mistaken 20nm is not Hybrid, 14XM from GloFo and 16nm from TSMC are "hybrid" but they will also bring tremendous power reductions. Also at the same time ARM will be releasing the new 64bit SoCs. The combination of new 64-bit ARM arch, manufactured with FinFets will be extremely difficult for 14nm ATOMs to compete against.
20nm ARM SOCs will be available in 2014, perhaps a few months earlier than first 14nm ATOMs and in 2015 14/16nm FinFets will compete against 14nm second Gen ATOMs. So its not all that easy for Intels ATOMs as you might think.
They can clearly opt to shoot themselves in the foot there, if they so desired, but being on 14nm isn't what will drive the price.
In fact the production cost for a 14nm chip will be lower than the 22nm chips as neither node will be operating on a fully depreciated toolset.
This is a false dichotomy because the reality is the OEMs don't have such choices to make.
OEMs either adapt and evolve while jostling against their competition, or they get one-upped and die out.
When TI was the mobile chip-maker of choice I watched the last round of OEMs rush around like mad trying to one-up each other all while TI was proudly telling shareholders our margins were going up up up quarter after quarter.
No one in the OEM landscape actually has the luxury of opting out of the rat race. Everything is a choice between two undesirable options, and picking the lesser evil usually means taking the option that at least will pay the bills for another business quarter.
This part, if true, is true for everyone with a stake in the game, ARM and x86 alike.
If the market has matured and is saturated with "good enough" products then that will stifle demand for Intel as well as Qualcomm/Apple/Samsung/etc.
At that point it is a race to the bottom for cost, and again that is where having a true node shrink like Intel's 14nm will be to their advantage versus everyone else who won't get any cost-saving benefits in going from 20nm -> 16nm or 14nm at the foundries.
Intel's production costs will be lower than everyone else's when it comes to their 14nm vs the foundries 14XM and 16nm, if that weren't the case then Intel would be fabbing their chips at the foundries instead of inside their own fabs.
(something I also experienced at TI when TI management decided to move production from our internal fabs to TSMC and UMC because the foundry prices were lower than our internal fab costs)
StrangerGuy
Diamond Member
- May 9, 2004
- 8,443
- 124
- 106
We all knew how those terrible Prescott chips destroyed Intel and made AMD the dominant player in terms of marketshare back in 2004 because the mass market behaved like AT forums right? Oh wait, that didn't happened.
Simply because AMD did not have the fab capacity and you can't just build additional fabs on short notice. Athlon FX cost $1000 because of high demand and low fab capacity. mainstream dual cores were what like $600? And then Core2 came along.
One could also argue that the process saved Intel. because it was good P4 was not that bad. People also forget that performance wise the P4 was not that bad. it wasn't like 50% slower single-thread like current AMD chips and consuming double the power.
Intel expects a 17% perf/watt lead over ARMv8. Even Silvermont will be very hard to beat for them. 14nm Atom is out of reach for ARM.
AtenRa
Lifer
- Feb 2, 2009
- 14,003
- 3,362
- 136
Now that the first performance data of the 22nm ATOM Silvermont its becoming available, 14nm ATOM will save the day, every year the same old story
How can they know given that there's no ARMv8 chip on the market? It would be so nice if people sticked to facts instead of rehearsing Intel marketing. What about waiting for the first devices to be tested by people outside Intel?
Now that the first performance data of the 22nm ATOM Silvermont its becoming available, 14nm ATOM will save the day, every year the same old story
Nobody is waiting, 22nm Atom should be enough to beat them.
From ES samples for example. Or do you think they have to wait until the first device is purchasable for anyone? Same for ARM, they are surely aware of Silvermont, they don't have to wait till November.
StrangerGuy
Diamond Member
- May 9, 2004
- 8,443
- 124
- 106
Now that the first performance data of the 22nm ATOM Silvermont its becoming available, 14nm ATOM will save the day, every year the same old story
I heard consumers saying they won't buy any mobile products until those have 10nm chips inside so they can play Candy Crush without lag.
Vesku
Diamond Member
- Aug 25, 2005
- 3,743
- 28
- 86
"Krzanich and James said that under their leadership, Intel will give much more priority to its Atom mobile chips. In the past, Intel's most cutting-edge manufacturing resources were reserved for making powerful PC chips, with Atom chips made on older production lines.
"We see that Atom is now at the same importance, it's launching on the same leading edge technology, sometimes even coming before Core (Intel's line of PC chips)," said Krzanich."
http://www.reuters.com/article/2013/06/28/us-intel-ceo-idUSBRE95R0XK20130628
A generally good sign that Intel can proceed quickly from CEO mission statement to actual company changes. Should lead to more device variety, at least in the short term.
"We see that Atom is now at the same importance, it's launching on the same leading edge technology, sometimes even coming before Core (Intel's line of PC chips)," said Krzanich."
http://www.reuters.com/article/2013/06/28/us-intel-ceo-idUSBRE95R0XK20130628
A generally good sign that Intel can proceed quickly from CEO mission statement to actual company changes. Should lead to more device variety, at least in the short term.
Tremendous compared to what, exactly?
And you've misread what he stated: look up "BEOL" and "FEOL."
Yes, TSMC and the Fab Club's 1st gen FinFETs will surely beat the 2nd generation FinFETs of the long-running industry leader.
Listen, you can trash Intel all that you'd like, but it is simply foolish to pretend that anybody will be able to match Intel at the game they've been winning for decades: semiconductor fabrication. We can safely assume that Intel will have the best 14-16nm class process, especially in light of the fact that none of their competitors will achieve significant scaling improvements. Meanwhile, Intel is on track to roughly double the density compared to their 22nm process, not to mention their 450mm wafers going into production during the relevant time frame.
As far as 64 bit ARM goes, there simply is insufficient data to be making any claims in regards to its performance. However, this is a great opportunity for me to point out that Atom will be 64 bit capable well before ARM.
A more accurate portrayal:
It'll take 20nm ARM SoCs to match Intel's 22nm Silvermont. Not much later, 14nm Airmont will replace it... leaving the ARM club in the dust for a lengthy period of time. The ARM club will finally catch up in 2015 using second-rate 14/16nm processes, however Intel will quickly answer with Airmont's successor (a tock, mind you), where Intel will be easily be the leader for an extensive period of time.
Apparently you have not heard of an important metric called "cost."
Time will tell. As it stands bt already lost before it even began looking from the current landscape. Intels blurps is nothing but the old style b2c communication from traditional cpu market.
If they could deliver the important design wins for 22mm bt i highly doubt we would hear this talk about future-future-future products. There would be no need for all the talk. How serious is that. Typical signs of weakness, while qualcomm is just executing and probably have a future portfolio release so frequent the tick-tock talk looks funny.
If they could deliver the important design wins for 22mm bt i highly doubt we would hear this talk about future-future-future products. There would be no need for all the talk. How serious is that. Typical signs of weakness, while qualcomm is just executing and probably have a future portfolio release so frequent the tick-tock talk looks funny.
Where I come from, schedules being moved up in the tech industry are almost invariably the exact opposite of a sign of weakness.
Vesku
Diamond Member
- Aug 25, 2005
- 3,743
- 28
- 86
Even when the new CEO talks about trading away some "perfectionism" for faster rollouts?
That was the CTO, not the CEO.
Regardless, until you have an idea of what corners are getting cut in order to speed up time to market, your question is meaningless.
Talking about future products is not a weakness. companies like intel,amd,nvidia like to talk abt future products. qualcomm or apple doesnt. its just the way these companies are
cytg111
Lifer
- Mar 17, 2008
- 26,837
- 16,109
- 136
Around the time desktop and atom is merged
No wonder why so many goes x86 now.
IMO not too far off.
We cant keep adding transistors and cores with no regard to power consumption.. it seems reasonable to focus on perf/watt now while devs struggle with concurrency, i predict those two (devs & concurrency / cores and watts) is in a somewhat of equilibrium for the time to come. (and possibly a third componet to moores law*)
*) there's two sides to "moores law"...what it was intended to describe, and that to the letter. Religious folks will follow the latter.
Last edited:
Then at least talk straight and not through "rumours" like this is b2c cpu market and this forum is the customers.
What happened to bt? Why talk about what comes after the future?
We have waited 5 years for a new Atom arch. Either its going to succeed or 14nm whatever is not going to make the slightest difference.
Last edited:
Dont quote me out of context.
Its a sign of weakness when your current product is performing really bad on the market and have done so for years, and you start releasing "rumours" about successors comming after the next replacement. Looks like GF method to me just without ppt. Perhaps even more pathetic. Bad PR imho no matter the quality of your next product.
Why not? The fact that they will have 14nm Airmont in Q3 and possibly a new architecture before the end of 2014 bother you?
Last edited:
So AMD detailing Steamroller at last year's hot chips is a sign of weakness, correct? Because by your logic, it would be.
Unlike Steamroller, however, there's nothing but positive news surrounding the next generation Atom processors. Huge performance increases. Massive power reductions. To top it off, the 14nm successor to Silvermont is rumored to be coming sooner than expected. Meanwhile, Kaveri has seen delays -- however I don't believe Kaveri is necessarily doomed to fail.
I have a suggestion for you, and it's one that would be wise to follow: ignore media buzz as a sign of whether or not a company or product is doing well or not.
Where's your proof that this "rumor" was purposefully "leaked" by Intel?
You're picking and choosing the facts you are seeing and believing. Your contributions to this discussion are not only unhelpful, but counterproductive.
Let me make sure you understand something: the primary reason why Intel's Atom strategy has "failed," as you call it, is because they haven't updated the architecture in over 5 years. Your entire argument is based off the patently flawed assumption that Atom will continue to fail, despite the fact that it is becoming a first-class citizen in Intel's eyes.
To reiterate, the sole reason why Atom "sucks" is because Intel didn't pay much attention to it. Now that they're giving it their full attention, you're saying that it's still going to suck? Please tell me how you've come to this conclusion.
Last edited:
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter