Apple A9X the new mobile SoC king

[ShintaiDK]

You still claim A9X is faster, yet provide no benchmarks. I am sure it will do great in memory bound benches. Since the A9X will have ~50GB/sec. But when it comes to real computational, it is going to be a different story.

Also I can assure you Apples Mac sales will be severely hit if they cant run Windows. Many people and more importantly enterprises simply buy it for the design, not for OSX.

Seems to be another case of the enemy of my enemy is my best friend.

 

[Edrick]

A modified version of their custom ARMv8 core which can scale in frequency to 3.5+ Ghz and power their workstation, desktop and high end notebook requirements. Nothing stops Apple from doing that.

Nothing stops Apple other than this:

What was in question was whether they would do it without equally reaching x86 class prices, die sizes, transistor budgets, and power consumption.

 

[MikeA65]

At this point all we have is Apple's claim of 1.8x A8X for A9X and Intel's claim of 2x A8X for the m7-6Y75. We will have to wait until real benchmarks (something better than Geekbench) are run to see how it turns out.

 

[Edrick]

You still claim A9X is faster, yet provide no benchmarks. I am sure it will do great in memory bound benches. Since the A9X will have ~50GB/sec. But when it comes to real computational, it is going to be a different story.

Again, the same people who praise Apple for doubling their memory throughput are the same ones who downplayed Intel when they released CPUs with quad channel memory. And I am fairly certain that the argument was "Show me real world performance, and not memory specific benchmarks" back then too.

 

[ShintaiDK]

Again, the same people who praise Apple for doubling their memory throughput are the same ones who downplayed Intel when they released CPUs with quad channel memory. And I am fairly certain that the argument was "Show me real world performance, and not memory specific benchmarks" back then too.

So true.

Long way to reach Core M performance.

 

[Stef123]

Intel's tick tock model is being bulldozed by Apple's model.

BTW the trend does not seem to be Apple specific. The whole of ARM world seem to be moving at the same pace. For example look at the new Samsung cores:

http://media.idownloadblog.com/wp-content/uploads/2015/08/Apple-A9-GeekBench-3.jpg

There is a good chance the leak is faked, but those numbers do not seem horribly off, those are ones we can certainly see in 2016.

The first to convert that type of power to desktop and server uses would be the next billionaire (unless it is one that already is billionaire ... see Apple )...

 

[Stef123]

So true.

Long way to reach Core M performance.

There is something very much off in this picture. For one a8x less than 20% faster than A7 ?!?!

And another is Nvidia Shield tablet with its A15 cores handily beating the Denver cores of Nexus 9 ...

 

[ShintaiDK]

The A7 to A8X is often severely overrated. Not really a surprise.

 

[Nothingness]

There is something very much off in this picture. For one a8x less than 20% faster than A7 ?!?!

And another is Nvidia Shield tablet with its A15 cores handily beating the Denver cores of Nexus 9 ...

There's obviously an issue with that benchmark (no matter what the company who wrote it said): when a single benchmark shows something vastly different from all others, it should be questioned. Anyone remember AnTuTu cheat to show some Intel chip in a better light? :biggrin:

 

[sm625]

This is the one that I'm watching. If A9X can even just match an i3-4020Y in Kraken, that would be huge.

 

[kimmel]

There's obviously an issue with that benchmark (no matter what the company who wrote it said): when a single benchmark shows something vastly different from all others, it should be questioned. Anyone remember AnTuTu cheat to show some Intel chip in a better light? :biggrin:

Remember the custom per benchmark overclocking switches to show Arm chips in a better light?

The best part is how posting benchmarks vs powerpoint triggers some global warming conspiracy reactions in people.

 

[witeken]

Apple sold 74.3M iPhone 6 in 2014 Q4, and there is no doubt they will sell at least as many iPhone 6S in the same period this year. My point is not only Apple has caught up with 14nm finfets, they are also going to easily outship Intel's 14nm chips in volume this year too.

LOL!!!

Do you know any company that has sold 1 billion FinFET chips? Do you know any other company that has made more than 100 billion USD from selling FinFETs?

 

[Nothingness]

Remember the custom per benchmark overclocking switches to show Arm chips in a better light?

Everyone is cheating, we know that. But how does that dismiss my point? Intel is no better, no matter where your preferences go.

The best part is how posting benchmarks vs powerpoint triggers some global warming conspiracy reactions in people.

We're talking about real benchmarks as reviewed, not powerpoint stuff. Intel cheating has been demonstrated, so there's no conspiracy theory here, just some blind people refusing to open their eyes; let me say it again: everyone is cheating.

 

[kimmel]

Everyone is cheating, we know that. But how does that dismiss my point? Intel is no better, no matter where your preferences go.

We're talking about real benchmarks as reviewed, not powerpoint stuff. Intel cheating has been demonstrated, so there's no conspiracy theory here, just some blind people refusing to open their eyes; let me say it again: everyone is cheating.

No, we're in a thread about how Apple has surpassed Intel's performance based on a powerpoint slide.

 

[StrangerGuy]

LOL!!!

Do you know any company that has sold 1 billion FinFET chips? Do you know any other company that has made more than 100 billion USD from selling FinFETs?

I'm so sorry for your bad reading comprehension.

Oh BTW do you know a company that supplies chips to an ever shrinking Windows PC market whose sales are now overtaken by iOS devices alone, much less Android?

http://9to5mac.com/2015/07/21/ios-sales-outpace-windows-sales/

 

[Stef123]

The A7 to A8X is often severely overrated. Not really a surprise.

I don't think so. The mere fact alone that it has one more core implies at least 40-50% better performance in multi-threaded applications.

Anyone showing less than that is suspect. Most often than not they are single/dual-threaded loads in a heavily multi-threaded world. That's why I call such benchmarks "suspect" (they only show single-threaded performance).

 

[ShintaiDK]

I don't think so. The mere fact alone that it has one more core implies at least 40-50% better performance in multi-threaded applications.

Anyone showing less than that is suspect. Most often than not they are single/dual-threaded loads in a heavily multi-threaded world. That's why I call such benchmarks "suspect" (they only show single-threaded performance).

We all know ST performance matters. To try and hide this is simply misleading. Else we are pretty much back to FX8xxx CPUs sales argument.

 

[Nothingness]

No, we're in a thread about how Apple has surpassed Intel's performance based on a powerpoint slide.

I certainly don't buy that, at least not until there's been some independent reviews that demonstrate that. I hope Anandtech will run SPEC 2000 (or even 2006) on the iPad Pro.

 

[dark zero]

The A7 to A8X is often severely overrated. Not really a surprise.

Short timers benchmarks are not serious... Someone put Prime95 to see how.much the Core M resist.

 

[Thanatosis]

A9X is in another league, A8X was already a match for base Broadwell - Y (core M). Shintai and the rest of the intel apologist squad can you please tell me which cross-platform benchmark you would like to use besides geekbench, since you get wrecked sooo badly every time they are cited? Haha


There isn't one. Geekbench is the best way to compare platforms between x86 and ARM at this point in time, and if intel doesn't like being made a fool then they should maybe add a core or something? Actually improve the uArch enough maybe to get 10% in a year? Is that so hard? Or I guess margins and investors are more important... but the investors are starting to see the writing on the wall....


https://www.google.com/search?q=int...j0j4&sourceid=chrome&es_sm=91&ie=UTF-8#q=INTC

I wonder if Microsoft will start using Apple ARM cores for their wintel books, or would Apple even be willing to give them a chip? They are probably dying to get rid of intel's horrible power consumption and Atom is killing surface 3. Every person I know who bought one returned it for an ipad. The ipad Pro just seals the deal.

 

[ComplexEntity]

I don't think so. The mere fact alone that it has one more core implies at least 40-50% better performance in multi-threaded applications.

Anyone showing less than that is suspect. Most often than not they are single/dual-threaded loads in a heavily multi-threaded world. That's why I call such benchmarks "suspect" (they only show single-threaded performance).

You are too late into this debate. I remembered when A7 was released almost everyone on the Apple camp cried fault on 3DMark 11 physics test for been "not meeting Apple's claim of performance increase therefore something must be fishy". You have to understand this Physics benchmark is a CPU benchmark that indicate how well a CPU utilize the Bullet Engine (which is multi-threaded). For the amount of applications that implement/utilize Bullet Engine, please refer to https://en.wikipedia.org/wiki/Bullet_engine. FutureMark even documented their response as why the result wasn't met someone's expectation http://www.futuremark.com/pressrele...results-from-the-apple-iphone-5s-and-ipad-air. Notice the 20% performance increase the A8 over A7 may very well been the IPC increases as A7 operates on similar Ghz vs A8 AFAIK (which is still very impressive). All I can say is that it just appears that Intel chips are more general-purpose designed than their Apple counter parts (which serves Apple product only).

 

[Sweepr]

Geekbench is crap for x86 vs ARM. If that represents real-world performance then Exynos 7420 matches the MT performance of a 15W Core i5 5200U and A9X will be faster than a 28W Core i7 Haswell-U.

 

[Stef123]

You are too late into this debate. I remembered when A7 was released almost everyone on the Apple camp cried fault on 3DMark 11 physics test for been "not meeting Apple's claim of performance increase therefore something must be fishy". You have to understand this Physics benchmark is a CPU benchmark that indicate how well a CPU utilize the Bullet Engine (which is multi-threaded). For the amount of applications that implement/utilize Bullet Engine, please refer to https://en.wikipedia.org/wiki/Bullet_engine. FutureMark even documented their response as why the result wasn't met someone's expectation http://www.futuremark.com/pressrele...results-from-the-apple-iphone-5s-and-ipad-air. Notice the 20% performance increase the A8 over A7 may very well been the IPC increases as A7 operates on similar Ghz vs A8 AFAIK (which is still very impressive). All I can say is that it just appears that Intel chips are more general-purpose designed than their Apple counter parts (which serves Apple product only).

The link that you give me is old and explains why the A7 SoC fell short, I don't see how it helps in proving my remark right or wrong.

A7 and A8x is the same architecture, A8x has a tiny IPC advantage as well as higher clock rate. That's enough to represent the 20% difference seen between the two.

But where's the extra-core? Do you need to tell me that a whole extra core makes absolutely no difference in performance while all the heavily multi-threaded tests show the exact opposite?

It's easy to pick and choose benchmarks to show a point. Typically the ones that do not account for as central things as extra cores are the ones of the worse to do just that (because they're easy not to be taken seriously).

Like I said any benchmark that shows less than 40% difference between CPUs of the same architecture yet different core count is suspect at best; in most cases it means that it's not a multi-threaded test.

Geekbench is crap for x86 vs ARM. If that represents real-world performance then Exynos 7420 matches the MT performance of a 15W Core i5 5200U and A9X will be faster than a 28W Core i7 Haswell-U.

It's close enough.
If you try to run Linux on ARM you'd see just that. Phoronix showed that the weaker (cpu-wise) Terga X1 almost matches i3 performance (apart from some tests).

http://www.phoronix.com/scan.php?page=article&item=nvidia-tegra-x1&num=2

 

[Stef123]

mods please delete

 

[soccerballtux]

double post (mods please delete)

wow. ARMH might still be a 'buy'