Apple A8 Benchmarked

[tviceman]

Are Nvidia's TK1 canned benchmarks showing faster than this???

 

[dave_the_nerd]

I knew they wouldn't be able to deliver 2x CPU performance every year forever, but anything less than 50% is kinda disappointing. I wouldn't hold my breath for huge GPU gains either, but I am a bit more optimistic here. 1GB is a bummer considering these are late 2014 64-bit devices, there's no other way to look at it. I'll stick with my A7 devices and wait for 2015 16nm A9.

Disappointing? I think we may be getting spoiled.

That's Core 2 / Phenom II X2 level performance. It's more powerful than most netbooks and/or Atom/Kabini systems. That's insane.

http://browser.primatelabs.com/processor-benchmarks
In a CELL PHONE.

(And yes, I realize there are competitive ARM parts from other manufacturers, many of which are even quad-core. It's not just Apple working freaking miracles.)

 

[liahos1]

this is true although given the gains from a5 to a6 to a7 I'm sure people were expecting more magic fairy dust from apple. it appears to be a solid chip irrespective.

 

[Space69]

Seems like a 16% bump. Meh. I guess they are focused more on their NFC stuff for this iteration.

Macrumors specifically chose a low score - over 70% of all the iPhone 5S entries in the Geekbench have a single-core score of 1400 or above. With that in mind - it's only a 8% improvement over A7.

 

[Homeles]

Realistically as an Apple product the specs won't even matter. There will be droves lining up to buy one weeks before they are even available :sneaky: A good number of consumers loyal to Apple wouldn't even understand the specs if you told them. Number one metric of import is: does it have a white apple with a bite out of it on the product followed by does someone else have one and now I must also have one.

What is the relevance of this? The hardware doesn't matter to the vast majority of consumers, for any brand. This applies to other industries too, like cars.

Since we're here for the hardware, why not discuss that?

 

[Sweepr]

Disappointing? I think we may be getting spoiled.

That's Core 2 / Phenom II X2 level performance. It's more powerful than most netbooks and/or Atom/Kabini systems. That's insane.

http://browser.primatelabs.com/processor-benchmarks
In a CELL PHONE.

(And yes, I realize there are competitive ARM parts from other manufacturers, many of which are even quad-core. It's not just Apple working freaking miracles.)

Considering they delivered close to 2x the performance over the past few years, I'd say a 15-20% bump (10% at same clocks) and no RAM increase is disappointing. That's not to say A7/A8 performance is disappointing, but Tegra K1 Denver, Core M, Exynos 5433 and a bunch of 2015 mobile ARM SoCs should be able to outclass it.

 

[Homeles]

2GB is definitely a puzzling choice, but since Apple packages it in their A-series SoCs, there likely just isn't room for more without increasing the package size.

 

[Grooveriding]

What is the relevance of this? The hardware doesn't matter to the vast majority of consumers, for any brand. This applies to other industries too, like cars.

Since we're here for the hardware, why not discuss that?

I agree with you. The hardware matters to us here. I was just highlighting that it will not negatively impact the success of the new iPhone in a significant way. Consider my comment your standard rabble rabble apple products suck and are all marketing and aesthetics, rabble rabble.

I think specs are more important and have a greater effect on a product's success in the Android market.

 

[Mopetar]

next one will have 1.5GB probably

That seems like an odd choice.

2GB is definitely a puzzling choice, but since Apple packages it in their A-series SoCs, there likely just isn't room for more without increasing the package size.

That could be, but haven't people said the A8 was going to be on a 20nm process, which would give them some extra space?

Another reason I've heard that they don't use as much RAM is that it reduces the power consumption. If they did move to a newer node and aren't aggressively targeting performance, it may mean that they're really favoring power efficiency.

Moving to a more pixel-dense screen for the supposed 5.5" phone is going to require more power to drive the display's backlight, so it may make sense to pair that with a CPU that draws less power in order to compensate.

 

[Khato]

That's Core 2 / Phenom II X2 level performance. It's more powerful than most netbooks and/or Atom/Kabini systems. That's insane.

Though keep in mind that is according to Geekbench. I can't help but think back a decade or so to Apple's comparisons of their G4 and G5 powered systems against PCs where the general lack of good cross-OS benchmarks meant that Apple could happily make their gross exaggerations of performance superiority.

One good thing to come from those NVIDIA K1 performance comparison numbers was a nice clear chart of Baytrail vs Krait vs Cyclone vs Haswell vs Denver. Note how in DMIPS (lowest level, most worthless in terms of representing actual performance) even Krait is almost on par with Haswell? Whereas soon as you switch to SPECInt 2K (low level, but meant to reflect a variety of actual integer algorithms and hence somewhat representative of actual performance) the Intel performance relative to ARM almost doubles. This despite the fact that they're both purely measuring integer performance. You can easily see the same thing happening in that chart with a comparison between Geekbench 3 and Google Octane v2.0. While there are plenty of flaws in javascript benchmarks I don't understand why Geekbench is any better - at least half of its integer section are typically handled by dedicated hardware (with AES and SHA1 hardware acceleration affecting the score for Apple's SoC, then JPEG and likely Sobel filter would be handled by imaging DSP in actual usage.) Then there's the amusement of the floating point where Haswell DGEMM for example is nowhere close to where it should be - Geekbench shows an i7-4770k single thread DGEMM at under 6Gflops whereas assuming 3.9 GHz and 90% efficiency it should be more like 56 Gflops.

 

[Ajay]

As mentioned, the #1 problematic use memory wise is a core app written by Apple - Safari.

It seems esp. bad with 64-bit iDevices, presumably because 64-bit apps and OS require more memory.

Current IOS and apps are 32 bit. If Apple doesn't put 2 GB in the next iPad models, that would really be a bummer - having tabs swapped out to flash storage is lame, IMO.

 

[witeken]

Considering they delivered close to 2x the performance over the past few years, I'd say a 15-20% bump (10% at same clocks) and no RAM increase is disappointing. That's not to say A7/A8 performance is disappointing, but Tegra K1 Denver, Core M, Exynos 5433 and a bunch of 2015 mobile ARM SoCs should be able to outclass it.

You can't increase performance if you can't lower the power consumption. You can't increase CPU performance per clock if you can't extract more ILP. Both things are hardly possible anymore. If Intel can't double performance every year, why could Apple?

 

[Arachnotronic]

You can't increase performance if you can't lower the power consumption. You can't increase CPU performance per clock if you can't extract more ILP. Both things are hardly possible anymore. If Intel can't double performance every year, why could Apple?

The mobile CPU market has been a lot like the PC market in its early days -- big performance leaps enabled by better process tech, larger area budgets, and increasing power envelopes.

The "slowdown" in Intel's CPU performance increases only came about once Intel had to deal with trying to keep power consumption the same/drive it lower. Oh, and keep in mmd that die sizes are limited by the per-unit margin targets that these chip vendors aim for.

Further, not all of the thermal/transistor headroom could be spent on just the CPU in an increasingly integrated world. Everything from PC processors to server chips increasingly requires more sophisticated non-CPU blocks.

In short: welcome to the maturation of the smartphone/tablet markets 🙂

 

[Batmeat]

Apple fans will pull excuses from everywhere about why it's the optimal amount of RAM, but the same people would laud Apple's foresight and brilliance if three months later they release iphone 6.1 with 8GB of RAM.

Not this apple fan. 1gb is pathetic. I do like apple products for the ease of use, but they lag behind the competition in nearly every aspect. Might be time I switched back to android.

 

[witeken]

You don't have to welcome me, Intel17, you have to welcome all the folks who expected 50% or higher performance increase. I'm not surprised. I actually expected Apple to focus on power, I was right.

 

[witeken]

Apple claims 25% CPU, 50% GPU. So A8 won't be the fastest SoC this time around.

Also, really annoying that AT isn't doing a live blog.

 

[FatherMurphy]

Yeah, I am a little surprised that there wasn't a live blog from either/both the Apple Event and IDF.

 

[witeken]

Also, A8 has 2B transistors. I wonder what the die size is.

Yeah, I am a little surprised that there wasn't a live blog from either/both the Apple Event and IDF.

Even more annoying is Apple announcing its things at the same time as IDF, quite immature of them.

 

[Eug]

Current IOS and apps are 32 bit. If Apple doesn't put 2 GB in the next iPad models, that would really be a bummer - having tabs swapped out to flash storage is lame, IMO.

Nope. 64-bit.

 

[NTMBK]

Apple claims 25% CPU, 50% GPU. So A8 won't be the fastest SoC this time around.

Also, really annoying that AT isn't doing a live blog.

This is still A8 in a phone, and they are deliberately keeping TDP down to prolong battery life and maintain sustainable performance:

An iPad version will probably be faster with a higher TDP.

 

[Ajay]

Nope. 64-bit.

My bad, IOS 7 brought 64b to the table.

 

[Sweepr]

You can't increase performance if you can't lower the power consumption. You can't increase CPU performance per clock if you can't extract more ILP. Both things are hardly possible anymore. If Intel can't double performance every year, why could Apple?

Yet some people though they would somehow keep doubling performance every year and beat Intel's low-power x86 chips in a few years.
50% increase in iGPU performance, not too shabby but not terribly impressive either. Probably a bit faster than Snapdragon's S805 yet slower than Tegra K1. Let me guess, PowerVR GX6650 @ ~450MHz? A8 is 13% smaller than A7 (102mm²).

About the iPad variant, it probably runs at slightly higher clockspeeds (100MHz anyone?) with the same iGPU performance. Let's not forget iPad Air doesn't usa PoP RAM like iPhone / iPad mini Retina so their larger tablets might come with 2GB RAM.

Overall nothing revolutionary but pretty solid performance for 2014-2015 devices.

 

[know of fence]

So is it a 20nm chip, or what?

 

[IntelUser2000]

Yet some people though they would somehow keep doubling performance every year and beat Intel's low-power x86 chips in a few years.

It does, and it does quite a bit. Low power x86: Atom

 

[liahos1]

how is it they doubled the number of transistors but the performance hasnt gone up by all that much?

also how is it that they have higher density in it that core m???

1.3 / 82mm^2 vs 2 / 87 mm^2?