Intel and the big.LITTLE concept

[Fjodor2001]

It basicly get a 5 year update or how long it is now to its uarch. OoO is one part. Its a major overhaul.

Just what is known about this Intel Atom 5 year overhaul? What can we expect from it in terms of performance, uarch changes, etc?

I've been hearing about new revolutionizing mobile Intel Atoms CPUs and platforms for several years (Moorestown, Medfield, Cloverdale, and whatever). Yet when they show up they are always a letdown. So far it has not lived up to its promise. ARM still rules the mobile world.

 

[Arachnotronic]

Just what is known about this Intel Atom 5 year overhaul? What can we expect from it in terms of performance, uarch changes, etc?

I've been hearing about new revolutionizing mobile Intel Atoms CPUs and platforms for several years (Moorestown, Medfield, Cloverdale, and whatever). Yet when they when show up they are always a letdown. So far it has not lived up to its promise. ARM still rules the mobile world.

Give Intel time. Considering their 5-year old Atom that was designed to be cheap & low performing still managed to show up to the fight and do quite well against the very best ARM chips is telling.

 

[Olikan]

Majority of, if not all, ARM partners have no real interest in ARM. They will happily change if something better comes along. And certainly Apple, they ditched PPC for the same reasons.

for sure they are using ARM because is dirt cheap, nothing else

 

[ShintaiDK]

Just what is known about this Intel Atom 5 year overhaul? What can we expect from it in terms of performance, uarch changes, etc?

I've been hearing about new revolutionizing mobile Intel Atoms CPUs and platforms for several years (Moorestown, Medfield, Cloverdale, and whatever). Yet when they show up they are always a letdown. So far it has not lived up to its promise. ARM still rules the mobile world.

Performance should be covered in a previous post.

Atoms have come a serious long way already. I guess you only focused on performance. But Atom is now in smartphones and tablets on equal foot as ARM in terms of battery, performance etc. Next thing is to turn up the performance. Hence refer to previous posts.

 

[ShintaiDK]

for sure they are using ARM because is dirt cheap, nothing else

I wouldnt call ARM dirt cheap...

 

[podspi]

Performance should be covered in a previous post.

Atoms have come a serious long way already. I guess you only focused on performance. But Atom is now in smartphones and tablets on equal foot as ARM in terms of battery, performance etc. Next thing is to turn up the performance. Hence refer to previous posts.

Where are you getting your information on performance expectations? All the leaks I have seen just list cache sizes, clock speeds. They tell nothing of performance.

 

[ShintaiDK]

Where are you getting your information on performance expectations? All the leaks I have seen just list cache sizes, clock speeds. They tell nothing of performance.

For the tablet versions:

 

[Olikan]

I wouldnt call ARM dirt cheap...

for fabless guys yeah... but i bet that, for Samsung is a very good deal

 

[Fjodor2001]

For the tablet versions:

Well, it doesn't say anything about power consumption. So how do we know if it can match ARM in mobile phones?

 

[ShintaiDK]

Well, it doesn't say anything about power consumption. So how do we know if it can match ARM in mobile phones?

How do we know ARM can even match Atom? Its ARM talking about big and small cores because they got a problem with power consumption.

 

[Roland00Address]

what it will bring to table? all i know is OoO x86

Now this is marketing bullshit slides. But intel posted a slide saying the 2015 intel atoms are going to be 10 times faster than the 2011 intel atoms in a specific test called SPECint2000. If the tablet atom that intel was refering to is the 1.66GHz Intel Atom N570 CPU then that processor performs about 7,700 MIPS in that test. If the 2015 intel atoms are 10 times faster then that makes them 77,000 MIPS which is faster then the phenom ii amd x6 is able to perform in the same test according to intel.

Who knows real performance 3 years down the road with the intel atom. Like I said marketing bs but very tasty bs.

 

[Cookie Monster]

How do we know ARM can even match Atom? Its ARM talking about big and small cores because they got a problem with power consumption.

Well the A15s for starters are looking quite good given their high performance with relatively low power envelope and the fact that these are being manufactured/shipped gives them the edge already.

Id think Intel has alot more to worry about with regards to power consumption than ARM in the mobile space, something thats been killing them off in this particular market for a long time and the reason why no one wants the current Atom based offerings by Intel.

 

[Exophase]

Now this is marketing bullshit slides. But intel posted a slide saying the 2015 intel atoms are going to be 10 times faster than the 2011 intel atoms in a specific test called SPECint2000. If the tablet atom that intel was refering to is the 1.66GHz Intel Atom N570 CPU then that processor performs about 7,700 MIPS in that test. If the 2015 intel atoms are 10 times faster then that makes them 77,000 MIPS which is faster then the phenom ii amd x6 is able to perform in the same test according to intel.

Who knows real performance 3 years down the road with the intel atom. Like I said marketing bs but very tasty bs.

SPECint_rate works by running the same single threaded benchmark over several threads. It scales very well but isn't representative of real threaded workloads.

You can see in 2012 the score roughly doubles because of a move from one core (Oak Trail) to two cores (Clover Trail), a clock speed boost from 1.5GHz to 1.8GHz, and improvements in memory performance along with very minor improvements in uarch. Intel is calling a 37.5% improvement after you take out the doubling from the second core, which I find slightly optimistic based on what we've seen so far but in the right ballpark.

The huge jump in 2013 is almost definitely facilitated by a 22nm Silvermont based SoC that has 4 cores. If you divide the score by 2 to negate the core scaling you get individual cores that are about 62.5% faster (based on the size of the bar). This huge boost will be partially due to better IPC from a new uarch, but most likely also due to a clock speed increase.

In 2014 you see the "tock" to 14nm giving a 9% performance improvement, probably entirely from clock speed increase. And in 2015 you get another massive IPC improvement, presumably from more extreme uarch refinements in their tick on 14nm. But that's so far into the future (especially in 2011!) that I doubt Intel had such a clear picture of performance and was surely very roughly estimating it.

nVidia has presented projected improvement curves that look at least as stunning but I don't see anyone here gushing over them... Maybe because nVidia is already known to lie in their marketing material, but I'm not so sure I trust Intel either with their extremely ambiguous SPEC2k numbers that eschew proper submission...

 

[Olikan]

Now this is marketing bullshit slides. But intel posted a slide saying the 2015 intel atoms are going to be 10 times faster than the 2011 intel atoms in a specific test called SPECint2000. If the tablet atom that intel was refering to is the 1.66GHz Intel Atom N570 CPU then that processor performs about 7,700 MIPS in that test. If the 2015 intel atoms are 10 times faster then that makes them 77,000 MIPS which is faster then the phenom ii amd x6 is able to perform in the same test according to intel.

phenom x6 will be 5 years old when this atom show up man...
it's not insane to think that atoms will reach that, specially when intel don't talk about number of cores, specific tdp, with HT or not

if you pay atention in the slides....These atom will have similar igp performance of last year bobcat, kinda patetic no?

 

[Idontcare]

These atom will have similar igp performance of last year bobcat, kinda patetic no?

AMD spent how much money in order to have that lead over Intel? Some $5B?

The output from project management is only pathetic if you are talking normalized inputs.

AMD got what it paid for, as did Intel. The igp better be higher performing for AMD.

 

[Olikan]

AMD spent how much money in order to have that lead over Intel? Some $5B?

The output from project management is only pathetic if you are talking normalized inputs.

AMD got what it paid for, as did Intel. The igp better be higher performing for AMD.

yeah, but still....
a chip from 2015 @14nm, won't straight beat an 2011 @40nm chip is unimpressive...
let's not forget that intel is putting alot of R&D lately on IGP

 

[ShintaiDK]

yeah, but still....
a chip from 2015 @14nm, won't straight beat an 2011 @40nm chip is unimpressive...
let's not forget that intel is putting alot of R&D lately on IGP

Did bobcat suddenly run in the miliwatt area? 28nm ARM chips cant beat Bobcat either, but again, different target.

The market already decided. And AMDs GPU advantage is not something people want, hence the 100mio$ writedown.

 

[CHADBOGA]

Id think Intel has alot more to worry about with regards to power consumption than ARM in the mobile space, something thats been killing them off in this particular market for a long time and the reason why no one wants the current Atom based offerings by Intel.

Things are changing fast.

If Intel can already do the below, I hardly see why they are going to have much difficulties in taking on ARM when Intel's OoO Atom architecture comes out towards the end of next year.

http://www.mobiletoday.co.uk/News/23063/razrI_cuts_b2b_market_motorola_mobility.aspx

Motorola Mobility’s new flagship smartphone, the Motorola RAZR i, is leading a push into the b2b market with the manufacturer reporting strong b2b demand for the Intel-powered device.

The launch of the Motorola RAZR i last month has been backed by a multi-million pound marketing campaign, the largest yet for its smartphones. Although the campaign is aimed at consumers, Andrew Morley, UK and Ireland general manager of Motorola Mobility, said the device was also creating a substantial amount of interest among business customers.

Morley (pictured) said: ‘We are finding since the launch that there has been a very strong reception from the b2b channels within our carriers and key customers. The fact that it is a small device, with a 4.3-inch screen, powered by Intel and with a fantastic battery life, carries a very powerful message in the b2b sector.’

Earlier this year, Motorola signed a ‘multi-year, multi-device’ partnership with chipmaker Intel, with the RAZR i, which packs the 2GHz Intel Atom, the first fruits of the team-up. Morley said the Intel partnership gave the RAZR i device more credibility in b2b channels. He said: ‘We are already very big in b2b in the US market. Intel also has a very powerful presence in b2b, so our partnership is certainly driving demand in that space, where trusted names are important.’

Morley said the RAZR i’s battery was a key selling point, with operators increasingly demanding longer battery life. He said: ‘We are finding that operators are getting increasingly keen on battery life because, while consumers are not using data and not downloading files they are not using the operator’s network. So now, one of the first questions carriers ask is, “how long does the battery last?”’
​

 

[Yuriman]

When exactly did Google acquire Motorola?

Wonder if there's any relation.

 

[ShintaiDK]

When exactly did Google acquire Motorola?

Wonder if there's any relation.

http://money.cnn.com/2012/05/22/technology/google-motorola/index.htm

 

[Roland00Address]

phenom x6 will be 5 years old when this atom show up man...
it's not insane to think that atoms will reach that, specially when intel don't talk about number of cores, specific tdp, with HT or not

if you pay atention in the slides....These atom will have similar igp performance of last year bobcat, kinda patetic no?

I am not saying it is impossible but that it is marketing slide of an unreleased product that has even been made yet. Marketing slides are often promise the world but then the info turns out to be misleading for it isn't like that performance in real world scenarios, or the product is delayed, or the slide is flat out wrong.

--------------------------------------------

Lets put it this way I am very much looking forward to the Intel 2013 Silvermont Atom Cpus. We know they are going to be up to quad core, out of order, using the low power version of 22nm with 3d transistors/finet technology (current 22nm are high power versions), and use ivybridge graphics (a variant of intel hd2500 but with 4 execution units instead of 6). All this in the next 12 months.

I can't trust anything that is going to come out 3 years from now with marketing slides for it is 3 years in the future on a process that has been made with a cpu architecture that has not been taped out.

 

[Khato]

The huge jump in 2013 is almost definitely facilitated by a 22nm Silvermont based SoC that has 4 cores. If you divide the score by 2 to negate the core scaling you get individual cores that are about 62.5% faster (based on the size of the bar). This huge boost will be partially due to better IPC from a new uarch, but most likely also due to a clock speed increase.

Possibly. Or the IPC increase could be even more than 62.5% with a reduced clock speed to keep power usage to a minimum. There's no evidence to support either conclusion.

 

[Roland00Address]

Possibly. Or the IPC increase could be even more than 62.5% with a reduced clock speed to keep power usage to a minimum. There's no evidence to support either conclusion.

The same slide that talked about valleyview (valleyview is the soc, silvermont is the cpu used in valleyview) that stated 4 ivybridge EUs also listed, up to 4 silvermont cores, and ghz speed from 1.2 to 2.4 ghz.

http://www.anandtech.com/show/6203/...soc-emerge-22nm-atom-with-ivy-bridge-graphics

 

[Exophase]

Possibly. Or the IPC increase could be even more than 62.5% with a reduced clock speed to keep power usage to a minimum. There's no evidence to support either conclusion.

Except for S|A's report that it'll still be two-way (which makes sense, given Jaguar is for instance - three x86 decoders is not as cheap as some make it sound) and have relatively mild OoO capabilities resulting in a 20-30% increase.

Not saying S|A is anything close to reliable, they are laughably wrong a lot of the time, but any chance that this information comes from anywhere legitimate at all makes it the most likely possibility so far, unless someone has contradictory information.

> 62.5% IPC increase is really huge... this capability has to come from somewhere, Atom cores aren't tiny so unless they flubbed the design hardcore I think you'd be looking at some pretty big cores in a market where Intel doesn't want really big cores.

 

[Fjodor2001]

Another question that should be raised: Do most people need more CPU performance in their mobile phones? Aren't we reaching a level where it is good enough for most people, i.e. same as for desktop CPUs?

I'd rather have better battery life than better CPU performance in my mobile phone. Remember the pre-Android and pre-iPhone days when you only had to charge your phone every two weeks? :thumbsup:

And assuming CPU performance is good enough and the focus instead is on battery life, doesn't it looks like ARM will win that battle after all?