Apple A9X the new mobile SoC king

[jfpoole]

Will GB4 include any SIMD processing?
Since GB4 will use Apple Xcode 7 for IOS, will desktop include the latest Intel Compiler?

Geekbench 3 takes advantage of SIMD instructions for some of the workloads.

We build Geekbench with the de-facto standard compiler for each platform since this how most software is built. I'm not aware of any platforms where the Intel compiler is used for a non-trivial amount of software.

Is the weighing between the 30 tests different when you compare mobile and desktop workloads?

The weights are the same. Again, the only difference is the data sets -- everything else is identical.

 

[Eug]

Another useless browser JavaScript benchmark:

Jetstream 1.1:

152.48: Core i7 870 iMac 2.93 GHz running Safari 8.0.8 and Mac OS X 10.10.5
146.85: Core i7 870 iMac 2.93 GHz running Firefox 41 and Mac OS X 10.10.5
144.82: Core i7 870 iMac 2.93 GHz running Chrome 45 and Mac OS X 10.10.5

84.211: Athlon II X3 435 2.9 GHz running Firefox 41 and Windows 7 SP1

Error: Core 2 Duo T8300 MacBook 2.4 GHz running Safari 6.1.6 and Mac OS X 10.7.5
83.148: Core 2 Duo T8300 MacBook 2.4 GHz running Firefox 41 and Mac OS X 10.7.5
84.655: Core 2 Duo T8300 MacBook 2.4 GHz running Chrome 45 and Mac OS X 10.7.5

75.968: Core 2 Duo P8400 MacBook Pro 2.26 GHz running Safari 8.0.8 and Mac OS X 10.10.5
81.175: Core 2 Duo P8400 MacBook Pro 2.26 GHz running Firefox 41 and Mac OS X 10.10.5
81.403: Core 2 Duo P8400 MacBook Pro 2.26 GHz running Chrome 45 and Mac OS X 10.10.5

50.080: Core Duo T2500 iMac 2.0 GHz running Chrome 34 and Mac OS X 10.6.8
Script timeout: Core Duo T2500 iMac 2.0 GHz running Firefox 41 and Mac OS X 10.6.8

119.47: iPhone 6s with iOS 9.1 Public Beta 2

84.256: iPad Air 2 with iOS 9.1 Public Beta 2

57.389: iPhone 5s with iOS 9.1 Public Beta 2

21.140: iPhone 5 with iOS 9.1 Public Beta 2

Crash: iPad 2 with iOS 9.1 Public Beta 2

For the iPhone 6s, it probably should be a little higher than 120 because I didn't turn off screen auto-lock, so I had to keep tapping the screen to keep it awake.

tl;dr:

Out of all the bazillion computers and iDevices I have in my house, the iPhone 6s handily beats everything except my Core i7 870 in (the largely single-threaded?) JetStream 1.1.

I redid the JetStream test for the Core i7 iMac with Safari 9.0 in Mac OS X 10.11 El Capitan GM Release Candidate. Updated score is 160.97.

As for the iPhone 6s, turning off screen auto-lock didn't give it a boost, so the above score of ~120 stands. That's over twice as fast as my iPhone 5s, and over 40% faster than the iPad Air 2.

 

[defferoo]

I redid the JetStream test for the Core i7 iMac with Safari 9.0 in Mac OS X 10.11 El Capitan GM Release Candidate. Updated score is 160.97.

As for the iPhone 6s, turning off screen auto-lock didn't give it a boost, so the above score of ~120 stands. That's over twice as fast as my iPhone 5s, and over 40% faster than the iPad Air 2.

since safari 9.0 on El Capitan and mobile safari on iOS 9 are supposed to use the same JS engine, comparing those two are probably the closest we're going to get to a cross-platform test on the browser side.

 

[Bubbleawsome]

Dang, I can't wait to see it in the iPad pro soon. I wonder what kind of apps we'll see taking advantage of that power.

 

[Space69]

Geekbench 3 takes advantage of SIMD instructions for some of the workloads.

Which workloads use NEON/SSE SIMD instructions? We do talk about SIMD instructions and not scalar instructions, right?

We build Geekbench with the de-facto standard compiler for each platform since this how most software is built. I'm not aware of any platforms where the Intel compiler is used for a non-trivial amount of software.

So Win version of GB4 will use VS2015?

 

[Thala]

On topic, I really would like to see a few more people run geekbench on their A9 devices because I have seen two different sets of scores: 2300/4000 and 2550/4450. I think one of them is TSMC 16 FF and the other is SS 14nm FF (not sure which is which).

For the technology to show impact it must be either:
1) They use different designs (on RTL level)
2) They run at different clock-speeds

I consider both very unlikely.

 

[raghu78]

Unfortunately because of the split contract people will have to acquire multiple samples before they can be sure to TSMC's and Samsung's A9 for comparison purposes.

You were vocal in stating that A9 is all TSMC because of superior yields and that Samsung is out of the picture due to very bad yields. Now you say Apple has split the contract between Samsung and TSMC and thus we have to source multiple samples to confirm if there are multiple A9 chips. Anyway we are getting a clue that A9 for iPhone 6S has gone to Samsung and A9 for iPhone 6S Plus has gone to TSMC.

Chipworks iPhone 6s teardown

http://www.chipworks.com/about-chipworks/overview/blog/inside-the-iphone-6s

"First off - Apple’s new A9 processor is the APL0898, and it’s ~8.7 x 10.7 mm, or ~94 mm2, which agrees with our 80% shrink guesstimates. Maybe that’s a reflection of the 14/16 nm processes only shrinking the transistor dimensions, not the metallization, maybe it means that Apple have crammed more in there (probably both). And it seems to confirm our postulated 8MB of L3 cache."

"The APL0898 size fits with an 80% shrink on the A8, and seems to have 8 MB cache - and our first look leads us to believe that our sample is from Samsung. "

iFixit confirms iPhone 6S has APL0898 (likely Samsung) while iPhone 6S Plus has APL1022 (likely TSMC). Normally the smaller iPhone model sells twice the units of the larger Plus model. This would mean Samsung gets the bulk of the volume this generation.

https://www.ifixit.com/Teardown/iPhone+6s+Teardown/48170

Step 17 - Apple A9 APL0898 SoC + Samsung 2 GB LPDDR4 RAM (as denoted by the markings K3RG1G10BM-BGCH)

https://www.ifixit.com/Teardown/iPhone+6s+Plus+Teardown/48171

Step 12 - Apple A9 APL1022 SoC + SK Hynix LPDDR4 RAM as denoted by the markings H9HKNNNBTUMUMR-NLH

So from first appearances it looks like Samsung and TSMC got the A9 with Samsung getting the bulk of the volume at 65% given the sales split.

With this split atleast within the iPhone 6S there will be no variances in efficiency and battery life. It does not make sense to have 2 different suppliers with different processes to supply chips for the same device. Apple has made a logically consistent choice.

 

[coercitiv]

You were vocal in stating that A9 is all TSMC because of superior yields and that Samsung is out of the picture due to very bad yields. Now you say Apple has split the contract between Samsung and TSMC and thus we have to source multiple samples to confirm if there are multiple A9 chips. Anyway we are getting a clue that A9 for iPhone 6S has gone to Samsung and A9 for iPhone 6S Plus has gone to TSMC.

In the future, could you please leave your original post content in place and mark additional (edited) content accordingly?

Your original post made it sound like Idontcare was wrong about the split, when in fact your newly posted data more or less confirms it.

 

[Thanatosis]

In the future, could you please leave your original post content in place and mark additional (edited) content accordingly?

Your original post made it sound like Idontcare was wrong about the split, when in fact your newly posted data more or less confirms it.

He was completely wrong about the split. IDC claimed that "all" prodcution of A9 would be TSMC due to "Samsung's low yields". As we can see here, Samsung has excellent yields to the point of now 13 million iphones (a record) being sold, most of which being Samsung 14nm 6s. It is unfortunate that IDC chose to contribute misinformation but as I said earlier, he's out of the industry. He doesn't know what a good process in 2015 is, maybe he did in 1999. Things have changed.


I suspected that the lower scoring iphones were on TSMC 16nm FF and it appears the 6s plus scores ~ 100 less in single core and about 150 less in multicore than the 6s. I wonder how scaling compares between the two?

 

[Nothingness]

iPhone 6s and iPhone 6s Plus Preliminary Results posted on Anandtech. I guess we'll get SPEC results later.

 

[raghu78]

In the future, could you please leave your original post content in place and mark additional (edited) content accordingly?

Your original post made it sound like Idontcare was wrong about the split, when in fact your newly posted data more or less confirms it.

Idc's claims were that A9 was slated to be 80% Samsung early in the year and as it got nearer to production, the Samsung share kept coming down due to bad yields and finally when production started it was all TSMC due to much superior yields compared to Samsung.

Now we see from the teardowns that A9 in iPhone 6S has gone to Samsung and 6S Plus has gone to TSMC. idc's claims on Samsung A9 supply volume share and yields are wrong. Samsung gets 65% of the A9 sales as the smaller 4.7 inch models sell twice the units of the larger 5.5 inch Plus models.

He was completely wrong about the split. IDC claimed that "all" prodcution of A9 would be TSMC due to "Samsung's low yields". As we can see here, Samsung has excellent yields to the point of now 13 million iphones (a record) being sold, most of which being Samsung 14nm 6s. It is unfortunate that IDC chose to contribute misinformation but as I said earlier, he's out of the industry. He doesn't know what a good process in 2015 is, maybe he did in 1999. Things have changed.

exactly. idc was horribly wrong with his claims. As we know 13 million units sold in the first 3 days and given that the smaller iPhone sells twice the unit volume of the larger Plus model its clear that Samsung (with Glofo maybe) was able to supply close to 9 million A9 chips.

 

[NTMBK]

Idc's claims were that A9 was slated to be 80% Samsung early in the year and as it got nearer to production, the Samsung share kept coming down due to bad yields and finally when production started it was all TSMC due to much superior yields compared to Samsung.

Now we see from the teardowns that A9 in iPhone 6S has gone to Samsung and 6S Plus has gone to TSMC. idc's claims on Samsung A9 supply volume share and yields are wrong. Samsung gets 65% of the A9 sales as the smaller 4.7 inch models sell twice the units of the larger 5.5 inch Plus models.



exactly. idc was horribly wrong with his claims. As we know 13 million units sold in the first 3 days and given that the smaller iPhone sells twice the unit volume of the larger Plus model its clear that Samsung (with Glofo maybe) was able to supply close to 9 million A9 chips.

Just because one 6S Plus is TSMC and one 6S is Samsung, does not mean that they all are. Small sample size.

 

[coercitiv]

He was completely wrong about the split. IDC claimed that "all" prodcution of A9 would be TSMC due to "Samsung's low yields".

Idc's claims were that A9 was *snip* all TSMC due to much superior yields compared to Samsung.

But this was not the subject of raghu's post, but the rather his revised opinion/speculation which was even acknowledged in the post.

Now you (idc) say Apple has split the contract between Samsung and TSMC and thus we have to source multiple samples to confirm if there are multiple A9 chips.

How is this not close to what other info sources brought us?

Now we see from the teardowns that A9 in iPhone 6S has gone to Samsung and 6S Plus has gone to TSMC.

As I stated earlier, it would be quite helpful to not replace the contents of a post when evaluating the statements of someone else, but rather add additional findings as you see fit.

 

[Eug]

Idc's claims were that A9 was slated to be 80% Samsung early in the year and as it got nearer to production, the Samsung share kept coming down due to bad yields and finally when production started it was all TSMC due to much superior yields compared to Samsung.

Now we see from the teardowns that A9 in iPhone 6S has gone to Samsung and 6S Plus has gone to TSMC. idc's claims on Samsung A9 supply volume share and yields are wrong. Samsung gets 65% of the A9 sales as the smaller 4.7 inch models sell twice the units of the larger 5.5 inch Plus models.

So are we 100% sure that ALL 6s units are Samsung and ALL 6s+ units are TSMC? Possible, but not guaranteed.

exactly. idc was horribly wrong with his claims. As we know 13 million units sold in the first 3 days and given that the smaller iPhone sells twice the unit volume of the larger Plus model its clear that Samsung (with Glofo maybe) was able to supply close to 9 million A9 chips.

Even without these teardowns, we knew at least some of idc's claims were wrong. The claim was that Apple would have released the iPhone 6s/6s+ earlier had Samsung's yields have been better. But that doesn't make much sense since the iPhone since the 4s has always been released in September/October.

 

[tempestglen]

iPhone 6s and iPhone 6s Plus Preliminary Results posted on Anandtech. I guess we'll get SPEC results later.

Since iPhone6S has 2G ram, SPEC06 is more suitable for comparing A9 with intel desktop class product.

After all SPEC06 is nearly 10 years old.

 

[kimmel]

The weights are the same. Again, the only difference is the data sets -- everything else is identical.

Why would the dataset be different? Makes comparing mostly useless.

 

[Arachnotronic]

Just because one 6S Plus is TSMC and one 6S is Samsung, does not mean that they all are. Small sample size.

This!

 

[Eug]

Why would the dataset be different? Makes comparing mostly useless.

I guess you missed his explanation for this earlier. Summary:

The results are the same, or least within about 1%, but the mobile ones are smaller to reduce the time to take to complete the test.

However, with Geekbench 4, the data sets are the same, mainly because mobile performance has advanced so much that the execution times are no longer such a big annoyance on mobile.

 

[R0H1T]

Why would the dataset be different? Makes comparing mostly useless.

Because a large dataset would likely result in overheating & throttling on the iphone while a desktop or notebook would be unaffected. Also worth noting is that (IIRC) smartphones throttle based on the chassis or shell temperature & even if the SoC is not running hot & say the display is, it would still throttle in a given benchmark. So unless you're adding an active heatsink to the smartphone as well, hoping to level the field just based on datasets or anything else is kind of pointless since the test environment is heavily skewed towards the computing platform with active cooling.

 

[kimmel]

Because a large dataset would likely result in overheating & throttling on the iphone while a desktop or notebook would be unaffected. Also worth noting is that (IIRC) smartphones throttle based on the chassis or shell temperature & even if the SoC is not running hot & say the display is, it would still throttle in a given benchmark. So unless you're adding an active heatsink to the smartphone as well, hoping to level the field just based on datasets or anything else is kind of pointless since the test environment is heavily skewed towards the computing platform with active cooling.

Thanks for perfectly explaining why comparing them is mostly pointless. You did forget about different datasets causing cache residency changes as well which makes it even worse.

 

[witeken]

Where does AnandTech have the Twister name from?

 

[Arachnotronic]

Where does AnandTech have the Twister name from?

Probably a "little birdie."

 

[witeken]

Probably a "little birdie."

I didn't know Typhoon either.

Pretentious names, BTW.

 

[Arachnotronic]

Pretentious names, BTW.

Why so?

 

[Eug]

Thanks for perfectly explaining why comparing them is mostly pointless. You did forget about different datasets causing cache residency changes as well which makes it even worse.

They already said the score results on mobile are within 1% whether they use the bigger data sets or not.

But, for Geekbench 4 it will be moot argument, because the data set sizes are the same. That should quiet the complaining to a certain extent.