Apple A8x

[asendra]

Look like Scam . 2d and 3d performance only half of a8.

But it is faster that the results for an original iPad Air (~+100% in 2D, ~+30% in 3D)

 

[Eug]

Look like Scam . 2d and 3d performance only half of a8.

Agreed. This doesn't pass the smell test. However, how do device names get entered in the database?

What other device would it be though? Disguised AMD desktop?

Edit:

I downloaded the Passmark app for iOS and there is no way to manually edit the device description.

CPU score is easily explainable if you consider the 40% increase to mean single core performance, I think. Original Ipad Air score (37k) * 1.4 (40%) ~ 52k. If it scaled perfectly, that would mean a score of 78k for a 3 core cpu. Considering it almost never scales perfectly, a 73k could be posible.

IIRC, Apple always advertises multi-core speed improvements, for obvious reasons.

 

[asendra]

Agreed. This doesn't pass the smell test. However, how do device names get entered in the database?

What other device would it be though? Disguised AMD desktop?

Edit:

I downloaded the Passmark app for iOS and there is no way to manually edit the device description.


IIRC, Apple always advertises multi-core speed improvements, for obvious reasons.

I don't know, I'm relatively confident, reviewers have had the device for a few days already, anyone could have slipped and uploaded the result, but I guess we'll know it soon enough. (Although we still don't know which day the iPad is officially released, my order shows delivery 22-24, but there's no date for in store release)


Edit: Holy shit at that Disk performance improvement if true, LOL. Almost +250%. That bothers me more, like if someone had used a Pc with an SSD to upload fake results :S

 

[Eug]

The disk test doesn't make sense either. The iPad5,4 is way, way faster than anything else on the list. That suggests to me that it could be something like a fast SSD.

Edit

Beat me to it.

 

[Arachnotronic]

Schiller did say the A8X would be up to twice as fast in CPU related tasks in some cases.

Higher per-clock performance, higher clocks, and a third core would do it.

 

[Arachnotronic]

So I would estimate the Geekbench 3 AArch64 results to be ~ 2000 for single-core and ~ 5400 for multi-core!

Getting pretty close to MacBook Air levels of performance.

 

[Rakehellion]

Getting pretty close to MacBook Air levels of performance.

Could be an app developer who accidentally uploaded the results.

 

[witeken]

a third core would do it.

Where does this tricore rumor come from?

Apple's claiming a 40% increase in CPU performance versus A7. A third core is 50%, and that's without a boost in clock speed and the IPC increase.

BTW, A8(X) might be one of the lastest ARM SoCs to be used by Apple. Intel inside iPhone is 'confirmed'. https://twitter.com/FPiednoel/status/523920464515575808

 

[asendra]

Where does this tricore rumor come from?

Apple's claiming a 40% increase in CPU performance versus A7. A third core is 50%, and that's without a boost in clock speed and the IPC increase.

BTW, A8(X) might be one of the lastest ARM SoCs to be used by Apple. Intel inside iPhone is 'confirmed'. https://twitter.com/FPiednoel/status/523920464515575808

No disrespect, but LOL if that qualifies as 'confirmed'.

 

[Rakehellion]

Where does this tricore rumor come from?

Apple's claiming a 40% increase in CPU performance versus A7. A third core is 50%, and that's without a boost in clock speed and the IPC increase.

BTW, A8(X) might be one of the lastest ARM SoCs to be used by Apple. Intel inside iPhone is 'confirmed'. https://twitter.com/FPiednoel/status/523920464515575808

Performance doesn't scale linearly by adding cores.

 

[Arachnotronic]

Schiller also said that in some cases, CPU is 2x A7. Don't forget this tidbit.

 

[witeken]

No disrespect, but LOL if that qualifies as 'confirmed'.

When in history has Intel ever set a goal for its processors that it was not able to accomplish? Does it say nothing that Intel has such a major market share in every market it plays in? History Suggests Intel Will Succeed In Mobile.

It took Intel from 1981 to 1984 to reach 15% of the PC market; from 91 to 94 to reach 15% of the data center market; and, from 2000 to 2003 to reach 15% of the High Performance Computing market. This year it expects to have 15% of the tablet market.

Schiller also said that in some cases, CPU is 2x A7. Don't forget this tidbit.

I am not impressed. From the iPhone 6 review (AT):

Given that these are all integer benchmarks, it may very well be that MCF benefits from the integer multiplication improvements the most, as its performance comes very close to tracking the 2X increase in multiplication throughput.

 

[Nothingness]

When in history has Intel ever set a goal for its processors that it was not able to accomplish? Does it say nothing that Intel has such a major market share in every market it plays in? History Suggests Intel Will Succeed In Mobile.

Indeed Intel never failed. iAPX432 was a great success. Itanium was also a huge success (HP of course can be partly blamed for that). Larrabee took over the GPU market.

And even if Intel had always been successful I fail to see how it proves it always will be. ARM has always been successful, so should we assume it will take over desktop and server? Citing an article from a pseudo financial analyst with a vested interest in Intel success won't change anything

 

[liahos1]

there are a couple of big question marks I continue to have with regard to intel.

Obviously I think the company has the assets and scale in place to succeed long term but there are definitely a few things that I'd love to pick BK's brain on with regard to recent performance.

1) Historically intel's manufactured atoms at n-1, with baytrail they got an updated uarch and 22nm. But then the company needed contra revenue support to gain share. Why? They say it was because the platform costs were higher but why were they higher and why couldnt intel anticipate platform costs would be uncompetitive?

2) Why does intel need rockchip, spreadtrum and RDA to gain share. Other than distribution what do these guys bring to the table from a knowledge base? It makes little sense. Are they better at spinning out SOC's at a faster than intel??? If it's purely for distribution - inte's been in china forever, how come they dont have the assets in place for distribution w/o help

3) Why is intel producing its 7260 baseband at TSM? Why has it taken so long to get internal production for either a discrete or internal baseband? Are finfets more difficult to product analog circuits with? Seems like the timeline keeps slipping back.

4) Cherrytrail seems delayed. What happened to broxton? I just dont understand their atom roadmap anymore and it doesnt seem like CORE M is going to grab a lot of share in the android space likely due to price. what is their android/tablet/fone strategy

 

[Dribble]

When in history has Intel ever set a goal for its processors that it was not able to accomplish? Does it say nothing that Intel has such a major market share in every market it plays in?:

Thing is it succeeds by being the only player in the market and then charging lots of money for it's cpu's - that's how x86 works. They can't do that with ARM, they can't patent lock it down. Hence while they can produce cpu's they can't make money on them because they can't control pricing. Intel basically expects 50%+ margins, they simply can't get that when competing with ARM.

Apple requires a 0% margin on their A8 because they make all their money on the end product. ARM charge Apple say $1 to license the chip and are happy, Intel can't compete with that. It doesn't help that Intel is forcing itself to use the dated x86 architecture which is less efficient, hence the great problems they have producing anything power efficient even with their process node advantage.

They are of course still using x86 because they if they switch people to that then they regain their lock down, but big companies (e.g. Apple) aren't stupid - they aren't going to let Intel take over and eat into their profits. Hence they develop their own cpu's, and don't care that they aren't always as good as Intel's because they know that freedom is keeping the money rolling in.

 

[witeken]

iAPX432 was a great success.

x86 is. You only need 1 ISA.

Itanium was also a huge success (HP of course can be partly blamed for that).

I haven't invested too much into researching Itanium, but again, who needs Itanium when there is a 64-bit version of x86 that everyone is using?

Larrabee took over the GPU market.

Intel's IGPs haven taken over the IGP market, although not yet in performance across the stack, but that's irrelevant. Intel changed plans and is not interested in entering the dGPU market.

ARM has always been successful, so should we assume it will take over desktop and server?

What is ARM's revenue? Does ARM have a long standing track record of taking over the markets it enters? Does ARM have full control from silicon to retail (EDM)? Has ARM a track record of decades of innovation and successfully securing their markets from opponent's strategies of (re)gaining market share? Has ARM a 50B dollar revenue stream to subsidize innovation? And finally, does ARM have technology that is 2-4 years ahead of its competitions, like Intel has with its semiconductor technology?

ARM has not been successful because of its microarchitecture, but because of its licensable ISA, which is cheap and convenient and had room for differentiation. But not anymore. Not in this maturing market, and you already see it. Qualcomm's reigning, and it accidentally happens to be using ARM's ISA. All low-end SoCs use cheap and slow in-order A7s.

It takes the right company to do something like what you propose, and ARM is not the right company, nor the company that Intel has to fear. Intel must face competition from other powerful companies like Qualcomm that have a holistic approach (Nvidia is only interested in the high-end, for example).

So, does Qualcomm have what it takes to defend Intel's attack. They don't have their own bleeding edge fabs, for one. Two years after Swift, they don't have 1 high-IPC SoC. They're just squeezing more and more MHz out of their ever increasing core counts. And it seems like we'll have to wait another year for such a core, although Krait will already be retired early next year. They're also very closed, unlike AMD, Nvidia and Intel, so maybe their vision is buried somewhere deep inside the company, far away from PR.

Indeed Intel never failed.

From a low-level perspective, sure, every companies is not perfect, but Intel is still the most successful semiconductor company, almost half a century since it was founded.

[...] And even if Intel had always been successful I fail to see how it proves it always will be.

Because it doesn't prove anything..

->

Citing an article from a pseudo financial analyst with a vested interest in Intel success won't change anything

Credentials are not what is important.

Its contents tell a relevant story.

 

[Homeles]

Why do you insist on quoting him? :\

 

[witeken]

1) Historically intel's manufactured atoms at n-1, with baytrail they got an updated uarch and 22nm. But then the company needed contra revenue support to gain share. Why? They say it was because the platform costs were higher but why were they higher and why couldnt intel anticipate platform costs would be uncompetitive?

You know that Intel changed CEO half a year before Bay Trail's launch? BK's predecessor is often cited for his lack of engineering background, focusing too much on margins. Brian Krzanich did away with all those ideas right from the start.

2) Why does intel need rockchip, spreadtrum and RDA to gain share. Other than distribution what do these guys bring to the table from a knowledge base? It makes little sense. Are they better at spinning out SOC's at a faster than intel??? If it's purely for distribution - inte's been in china forever, how come they dont have the assets in place for distribution w/o help

Because Intel doesn't have all the best employees in the world. It's better to have other companies at your side. I don't doubt that Intel can get a foothold in the market, but I never said it would be easy.

3) Why is intel producing its 7260 baseband at TSM? Why has it taken so long to get internal production for either a discrete or internal baseband? Are finfets more difficult to product analog circuits with? Seems like the timeline keeps slipping back.

Good question, I would like to know the answer of why it takes so long. Maybe because it simply does? I dunno. Idontcare?

4) Cherrytrail seems delayed. What happened to broxton? I just dont understand their atom roadmap anymore and it doesnt seem like CORE M is going to grab a lot of share in the android space likely due to price. what is their android/tablet/fone strategy

CT is delayed for the same reason BDW was. The delay of 14nm just came at a very unfortunate time. Broxton will converge Bay Trail and Moorefield. It will make Intel much more competitive from a TTM POV (and obviously in price, performance and power too, with Gen9, Goldmont and 14nm). I don't think anything happened to Broxton, but they simply haven't talked about 2015 anymore (IM is November 20).

Core M exists because Intel has to move its own market forward as well, and to serve the markets that demand more than Android and in-order quadcores.

 

[jfpoole]

For Example Geekbench SHA 1/2 operates on a single data buffer.

Geekbench 3 SHA-1 and SHA-2 workloads operate on a single data buffer because we don't believe cryptographic hashes operating on multiple data buffers is representative of real-world usage.

 

[witeken]

Why do you insist on quoting him? :\

What do you take issue with?

 

[asendra]

Geekbench 3 SHA-1 and SHA-2 workloads operate on a single data buffer because we don't believe cryptographic hashes operating on multiple data buffers is representative of real-world usage.

Geekbench dev? Common spill the beans, you guys must already have some A8X benchmarks on the database

 

[jfpoole]

Geekbench dev? Common spill the beans, you guys must already have some A8X benchmarks on the database

Nothing yet! This search should show the A8X results when finally arrive on the Geekbench Browser:

http://browser.primatelabs.com/geekbench3/search?q=ipad5

 

[Eug]

I still don't believe that Passmark iPad5,4 entry, but FWIW, in that graphics bench the iPad Air comes in lower than my lowly iPad 2. Same goes for the Retina iPad mini. Similarly, the iPad 3 comes in below the original iPad.

http://www.iphonebenchmark.net/g3dmark_chart.html

So, I think we can probably ignore the graphics benchmarks.

OTOH, iPad5,4 does come in faster than both the iPad Air and iPad 2.

No disrespect, but LOL if that qualifies as 'confirmed'.

Indeed, it was a frickin' twitter post with essentially no content, just wishful thinking. In all honesty that was the lamest "confirmed!!!" post I've come across in recent AnandTech history.

 

[witeken]

Thing is it succeeds by being the only player in the market and then charging lots of money for it's cpu's - that's how x86 works. They can't do that with ARM, they can't patent lock it down. Hence while they can produce cpu's they can't make money on them because they can't control pricing. Intel basically expects 50%+ margins, they simply can't get that when competing with ARM.

You logic does not make sense; Intel obviously knows that, yet they decided to enter the market anyway, which costs them billions per year.

You have a point, though. If Paul Otellini knew what would happen, Intel was now making tons of money. Now they'll make less because they have to compete with other companies that already participated in a race to the bottom. Qualcomm won and Qualcomm is doing absolutely fine. But that doesn't matter. Intel is losing a billion every quarter. They aren't going to give up now, they are simply going to take Qualcomm's place and become profitable. Once they have the position, who knows, maybe they'll be able to ask higher prices.

Apple requires a 0% margin on their A8 because they make all their money on the end product. ARM charge Apple say $1 to license the chip and are happy, Intel can't compete with that.

Intel can also compete. They can outcompete, until their fabs are so far ahead and their technology is so much better than Apple, then Apple won't have any other choice if they don't want to fall behind the leading edge of performance.

It doesn't help that Intel is forcing itself to use the dated x86 architecture which is less efficient,

x86 is not less efficient in any meaningful way. This was debunked by AnandTech almost 2 year ago and is now especially being proven by Core M.

hence the great problems they have producing anything power efficient even with their process node advantage.

They are of course still using x86 because they if they switch people to that then they regain their lock down, but big companies (e.g. Apple) aren't stupid - they aren't going to let Intel take over and eat into their profits. Hence they develop their own cpu's, and don't care that they aren't always as good as Intel's because they know that freedom is keeping the money rolling in.

Fear, uncertainty, doubt.

(Or do you have anything to back up your statements?)

 

[witeken]

Indeed, it was a frickin' twitter post with essentially no content, just wishful thinking. In all honesty that was the lamest "confirmed!!!" post I've come across in recent AnandTech history.

You take it way too literally. This is the first time that I've seen any statement like that; a statement of serious competition.

So far Intel's efforts have been in the tablet market. It's good to see their intentions for the smartphone space.