Anand: Apple's A7 Cyclone Microarchitecture Detailed (2014-03-31)

[NTMBK]

That's a fallacy. The wrong assumption you made is that Apple builds the best CPUs, but they don't. They're free to develop their own architecture, but why would they if they can't make the best? Silvermont has already proven to be vastly superior, so other companies can use it anyway while Apple sticks with their own inferior designs on inferior nodes.

Sorry, Silvermont has proven itself vastly superior? How so? It is in zero phones yet, and a mere handful of Android tablets (propped up by contra-revenue payments). Intel certainly has the superior fab, no denying that, but it will be very interesting to see how Silvermont does when ported to 28nm TSMC in Sofia.

And Apple is certainly the most rapidly improving CPU designer on the planet- they went from "never designed a CPU" to "people are comparing their designs to Intel" in the space of just two designs.

 

[Nothingness]

Silvermont has already proven to be vastly superior, so other companies can use it anyway while Apple sticks with their own inferior designs on inferior nodes.

Silvermont is so good it's been called "Conroe strikes back" by some Intel fanboys. Goldmont will be the one (c). Amusing :ninja:

 

[witeken]

Sorry, Silvermont has proven itself vastly superior? How so? It is in zero phones yet, and a mere handful of Android tablets (propped up by contra-revenue payments). Intel certainly has the superior fab, no denying that, but it will be very interesting to see how Silvermont does when ported to 28nm TSMC in Sofia.

Sales don't matter when you're comparing performance/watt, in which Silvermont indeed is vastly superior. I agree with your second sentence.

 

[Eug]

Sorry, Silvermont has proven itself vastly superior? How so? It is in zero phones yet, and a mere handful of Android tablets (propped up by contra-revenue payments). Intel certainly has the superior fab, no denying that, but it will be very interesting to see how Silvermont does when ported to 28nm TSMC in Sofia.

Yeah, even though it looks impressive, Silvermont hasn't really proven anything yet.

And Apple is certainly the most rapidly improving CPU designer on the planet- they went from "never designed a CPU" to "people are comparing their designs to Intel" in the space of just two designs.

Well, technically, Apple started with ARM 25 years ago.

 

[Rakehellion]

But the A7 probably can't scale to 3.4 GHz

That's why I said if.

But it isn't that far-fetched.

 

[jdubs03]

Sorry, Silvermont has proven itself vastly superior? How so? It is in zero phones yet, and a mere handful of Android tablets (propped up by contra-revenue payments). Intel certainly has the superior fab, no denying that, but it will be very interesting to see how Silvermont does when ported to 28nm TSMC in Sofia.

And Apple is certainly the most rapidly improving CPU designer on the planet- they went from "never designed a CPU" to "people are comparing their designs to Intel" in the space of just two designs.

Well said, and that leads us to the situation where it is going to be Broxton that will have to be the end all be all for Intel in the mobile space (in terms of entrenching upon the QCOM dominance).

Revisit this in 5 months and we'll have a better picture of where Intel vs. Apple is in terms of performance and of course perf/w. Apple has a chance here to really prove just how good their dual-core low-power solution is, and like I said before if it can compete with Broadwell-Y (4.5W TDP), its going to be very interesting.

 

[Rakehellion]

Sales don't matter when you're comparing performance/watt, in which Silvermont indeed is vastly superior. I agree with your second sentence.

It isn't a matter of sales, it's a matter of a ghost product with no real-world numbers. Intel probably has a 10GHz i9 in a laboratory somewhere, but that's no good for consumers if they can't buy it.

 

[witeken]

Revisit this in 5 months and we'll have a better picture of where Intel vs. Apple is in terms of performance and of course perf/w. Apple has a chance here to really prove just how good their dual-core low-power solution is, and like I said before if it can compete with Broadwell-Y (4.5W TDP), its going to be very interesting.

Don't expect too much from a CPU that's 1 node and many iterations behind, and lacks FinFET. Things get even worse if A8's made on a 28nm process. To compare: A7 consumes at least 1.5x as much power even though is has only half the amount of cores and 1.7x lower clock speed.

So A7 consumes 5x as much per clock per core, while it's only 2.5x faster per clock per core at most. 20nm won't change this 2x performance/watt deficiency because it competes against 14nm Tri-Gate, and when FinFET arrives Broxton will already have been launched.

It isn't a matter of sales, it's a matter of a ghost product with no real-world numbers. Intel probably has a 10GHz i9 in a laboratory somewhere, but that's no good for consumers if they can't buy it.

What do you mean with no real-world numbers?

 

[TreVader]

But if they can scale it to ~2GHz, that would be enough to replace Intel in MacBooks.

I think that's what they may be doing. I'd love to see apple designing it's own CPUs. Intel seems to have stagnated since it became the undefeated leader in high end CPUs.



If Apple takes intel out it's laptops, intel instantly loses a MASSIVE marketshare in notebooks. Aren't a majority of laptops currently sold MacBooks?



If Apple can make their first in house CPU compete with Haswell that easily, imagine what else they can do?

 

[Nothingness]

Don't expect too much from a CPU that's 1 node and many iterations behind, and lacks FinFET. Things get even worse if A8's made on a 28nm process. To compare: A7 consumes at least 1.5x as much power even though is has only half the amount of cores and 1.7x lower clock speed.

So A7 consumes 5x as much per clock per core, while it's only 2.5x faster per clock per core at most. 20nm won't change this 2x performance/watt deficiency because it competes against 14nm Tri-Gate, and when FinFET arrives Broxton will already have been launched.

Who cares? What matters is final product battery life. Apple iPad lasts about as long as BT-based tablets. This is due to two factors: the SoC isn't the main consumer in a phone/tablet; Apple benefits from vertical integration something Intel lacks.

What do you mean with no real-world numbers?

For Silvermont phones, we have no data for obvious reasons.

 

[coercitiv]

If Apple takes intel out it's laptops, intel instantly loses a MASSIVE marketshare in notebooks. Aren't a majority of laptops currently sold MacBooks?

In the U.S they have about 14% market share.

 

[krumme]

I looks like Charlie was spot on with his story of this Apple development years back. At that time it seemed crazy and unbelievable in my ears. But if anand story is right it is just that. Damn incredible fast development. Its some fat cat there.

Sure Apple wants to get this into their own laptops and desktops. No doubt they want control of this critical component. They want to stay independant and protect their valuable portfolio. A 2ghz a8 just do that for the low end. A9 at 16nm fin fets surely gets the job done. Makes sense.

Compettition is fierce ! Wow.

 

[TuxDave]

If Apple can make their first in house CPU compete with Haswell that easily, imagine what else they can do?

I haven't looked at any benchmarks lately. But by "compete", what numbers have you seen?

 

[witeken]

Who cares? What matters is final product battery life. Apple iPad lasts about as long as BT-based tablets. This is due to two factors: the SoC isn't the main consumer in a phone/tablet; Apple benefits from vertical integration something Intel lacks.

Nice straw man, since when does performance not matter anymore?

 

[PPB]

I looks like Charlie was spot on with his story of this Apple development years back. At that time it seemed crazy and unbelievable in my ears. But if anand story is right it is just that. Damn incredible fast development. Its some fat cat there.

Sure Apple wants to get this into their own laptops and desktops. No doubt they want control of this critical component. They want to stay independant and protect their valuable portfolio. A 2ghz a8 just do that for the low end. A9 at 16nm fin fets surely gets the job done. Makes sense.

Compettition is fierce ! Wow.

Seems that the one to Conroe their competition was actually Apple with their Cyclone design. Heh, wonder under which rock will the Intel-Silvermont evangelists hide this time?

 

[witeken]

No doubt they want control of this critical component. They want to stay independant and protect their valuable portfolio. A 2ghz a8 just do that for the low end. A9 at 16nm fin fets surely gets the job done. Makes sense.

Compettition is fierce ! Wow.

Apple already has quite a lot of control on Intel's roadmap.

 

[TreVader]

I haven't looked at any benchmarks lately. But by "compete", what numbers have you seen?

Sunspider. Actually if you take A7 sun spider scores Apple is beating intel already on a clock/clock core/core basis.



And they're just getting started.

 

[TuxDave]

Sunspider. Actually if you take A7 sun spider scores Apple is beating intel already on a clock/clock core/core basis.

And they're just getting started.

I normally hate Youtube links... but this thing? (first thing that came up)
https://www.youtube.com/watch?v=uiKd1_WOpLw

 

[witeken]

Just look at the AnandTech review of the iPad Air.

 

[Homeles]

I normally hate Youtube links... but this thing? (first thing that came up)
https://www.youtube.com/watch?v=uiKd1_WOpLw

It's a browser benchmark. It's useless when doing cross platform comparisons.

 

[Eug]

http://www.anandtech.com/show/7335/the-iphone-5s-review/5

We start with SunSpider's latest iteration, measuring the performance of the browser's js engine as well as the underlying hardware. It's possible to get good performance gains by exploiting advantages in both hardware and software here. As of late SunSpider has turned into a bit of a serious optimization target for all browser and hardware vendors, but it can be a good measure of an improving memory subsystem assuming the software doesn't get in the way of the hardware.

---

http://www.anandtech.com/show/7460/apple-ipad-air-review/3

 

[TreVader]

It's a browser benchmark. It's useless when doing cross platform comparisons.

The only thing it is used for is cross-platform. iOS, Android, Windows.





In practical terms, Apple is already beating Intel in mobile processors. Maybe intel isn't all that amazing after all? Maybe the only thing it takes to design a CPU that beats Core is deep pockets (hence AMDs failure to beat intel).

 

[Homeles]

The only thing it is used for is cross-platform. iOS, Android, Windows.

And it does a terrible job at it.

 

[krumme]

Its not like an qq s800 lacks any power in a 4.4 phone imo and a damn fine total solution with eg excellent high end dsp and connectivity all over. And the sales numbers proves it.

But cyclone is hands down laptop class hardware. Period. Its a different thing and a different goal. Its build for raw power. Its like when the first Amd K7 arived with its brutal fat fpu. What was prior heavy server class design was brought to pc.

Now it looks like Apple just dumps a big apple each year. That is impressive. I like that.

Hopefully this will give us our octo core i5 in a year for desktop we dont want our expensive small desktop intel cpu to finance Intel futile venture into batling with apple samsung qualcomm and google. Naa. Give us some cheap octo cores now on a low level api!

 

[coercitiv]

Sunspider. Actually if you take A7 sun spider scores Apple is beating intel already on a clock/clock core/core basis.

Sunspider 1.0
2x Cyclone @ 1.4GHz - 384ms
4x Haswell i7 @ 800Mhz - 387ms (Win 8.1, IE 11) And guess what, most of the work happens on one thread. (on the Win machine at least)
2x Haswell i7 @ 1.4Ghz - 215ms (IE11 process affinity was restricted to CPU cores 0 & 2)
1x Haswell i7 @ 1.4Ghz - 270ms +/-10% (running the benchmark on one core only seems to affect confidence)