Apple A12 benchmarks

[Space Tyrant]

You have to understand that the OS is irrelevant for Geekbench. Only the compiler has impact. Of course compiler choice correlates with the OS but the fallacy is to equate performance differences to the OS.
What we do know is, that gcc produces faster code than msvc, therefore we are seeing higher scores under Linux than under Windows.

Assuming GB makes no system calls, that still leaves any OS background tasks and the way the scheduler handles them. Granted, those may not be significant in any particular benchmark run.

 

[JoeRambo]

One just has to love how this thread about mobile device SoC does not mention its direct competition from CPU stone age vendors. Samsung/QC really need to step up their game, i know they are used to being 2x slower in ST, but this is new level of embarrassing.

 

[FIVR]

I think most of Samsung/QC's problems arise from Android. Hard to get good performance when your OS is such a dumpster fire. Certainly they are inferior chip designers to Apple but they are at least improving (unlike intel).

 

[Thala]

Assuming GB makes no system calls, that still leaves any OS background tasks and the way the scheduler handles them. Granted, those may not be significant in any particular benchmark run.

When programming a benchmark, you will use a timer, which measures the time which was spend in the user process and not in kernel services.

 

[Thala]

I think most of Samsung/QC's problems arise from Android. Hard to get good performance when your OS is such a dumpster fire. Certainly they are inferior chip designers to Apple but they are at least improving (unlike intel).

I am pretty sure, when ARM would develop a 6(7?)-wide core it would be very comparable to Apples. Currently the Cortex A75 is only 3-wide. So i would not be so fast calling out inferior chip designers.

 

[SAAA]

Apparently, since Ryzen scores very similarly to intel, as shown in the post above, they have the same "flaw". Really it is an apples to oranges comparison. Maybe we should ask to see benchmarks of the A12 running BF1 multiplayer.

Wish I could see that, just a six core A12 stripped of all the extras on SoC and running at 3.5GHz with a desktop class GPU attached... wait until 2020 then who knows, maybe Apple will stick with Nvidia and release ARM powered laptops so some sort of comparison will be possible.

All the tests i have done as far with using Visual Studio 2017(MSVC) and compiling Win32 desktop apps for both ARM64 and X64 giving indication to the fact, that Geekbench is pretty reliable. However i am not blaming engineering talent.
My hypothesis is that x86/x64 will you only get that far independent how many engineering talent you throw at it. Its just an antiquated architecture, which relies on variable instruction length, microcode caches and predecode stages and less registers than any competing RISC architecture. No-one with a sane mind would come up with such an instruction set architecture today.

It's both enlightening and disheartening to hear that.

The good part is that seemingly IPC can go up another 40% from where we are using novel ideas, so if max clocks can be kept in the 4-5GHz range there's still a great deal of single thread increases for years to come.
Maybe it is time to move on from x86 but I don't see the big push toward something much newer and better... will it happen when that VISC architecture or a derivative sees the light?

 

[Abwx]

Intel 8700k @ 5Ghz uses 100 Watts.

It use 115W at stock on FP+AVX2 workloads and roughly 75W on Cinebench, dunno how it could be at 100W@5Ghz even on CB, guess that it must be with 1% voltage margin and light loadings.

https://www.computerbase.de/2017-11/intel-core-i7-8700-i5-8600k-test-auto-oc-ddr4/3/

To get back OT i would be surprised if those tablet/phone CPUs would do as well as x86 ones in similar softwares.

 

[Nothingness]

I come to two conclusions, mutually exclusive:

-geekbench is crap and these scores mean nothing in the real world, like an Ax series desktop chip wouldn't run games and apps nearly as fast as the numbers suggest.

-scores are somehow valid thus all the genius CPU architects must have gone to Apple and are working on fantastic tech that unfortunately will never power our PCs.

And the third one: these scores are just fake.

 

[Eug]

https://m.weibo.cn/status/4236380060065313

~25% improvement in both Single thread and Multiple thread benchmarks.

Given the uniform improvement in both, this lead me to believe A12 may likely only be ~25% higher clock speed ( Max Clock 3Ghz ) 7nm A11. We don't know if there are any changes in other parts of SoC such as GPU.

I have long wonder when will Apple stop moving from node to node and IPC improvement YoY.

So, why should we believe this post? Is this person a known reliable poster, or just some random forum person?

BTW, if this is real (and that's a big if), and if this Google Translation is accurate, that last part about higher than expected power utilization for 7 nm is concerning.

"A12 related, current GB4 scores 5200 13000 up and down, increase the number of branch forecasting units, is currently solving large nuclear power problems (average power consumption is still 23% higher than expected in the case of 7nm)"

BTW, those numbers are closer to ~21% higher than A11. If just a node shrink with no IPC change, then that would translate to a 2.9 GHz A12.

 

[JoeRambo]

I am pretty sure, when ARM would develop a 6(7?)-wide core it would be very comparable to Apples. Currently the Cortex A75 is only 3-wide. So i would not be so fast calling out inferior chip designers.

Aren't Samsung / QQ / ARM in competitive? If they could design and release such core in reasonable power envelope, users would flock to buy phones with it.

 

[french toast]

Glad you brought some perspective to this thread instead of just the intel bashing. I was just getting ready to post a question of whether AMD would score any better. Apparently the answer, as I expected, is no.

Intel bashing?? People are using intel processors because they have the best ST performance, if Ryzen was better we would be comparing that, chill out Lancelot.

 

[Arachnotronic]

Did I ever say anything about Ryzen being any better?


The difference is that intel continues to arrogantly claim to have a process lead when all evidence suggest they are (at best) equal to the competition.


It is also obvious that Apple should be able to out-design AMD given that Apple's quarterly service revenue is equal to AMD's yearly revenue. It is far more embarrassing for a 200B company like intel to fail to compete so drastically.

Apple's quarterly revenue is roughly equal to Intel's annual revenue (Apple's best quarter in a year is greater than Intel's entire annual revenue).

AMD's annual revenue is about a tenth of Apple's quarterly revenue, worst case.

So, really, what's the problem here? Apple is bigger and richer than Intel and puts out better processor cores.

 

[dark zero]

So that explains that... higher power consumption...

 

[Arachnotronic]

Aren't Samsung / QQ / ARM in competitive? If they could design and release such core in reasonable power envelope, users would flock to buy phones with it.

Most smartphone buyers have no idea what a CPU core is or what a Geekbench is.

 

[Arachnotronic]

You can make all the excuses you want but in the end if the test was "stupid" or had no relation to performance intel would be able to game it and score just as well. Unless you are suggesting that Apple can "game" these test but intel is simply "too honest" to do the same.


The simple explanation is that something about intel's architecture is inherently less efficient than apple's.

Apple probably has a more capable CPU architecture team than Intel or anybody else does. This is a pretty well known fact within the industry and the trend is probably going to further shift in Apple's favor as Apple is pretty ruthless about poaching the best talent in the industry, which is why they've grabbed so much Intel talent over the last five years. Now that AMD is getting interesting too, don't be surprised to see some high profile poaching going on from there, too

 

[Eug]

So that explains that... higher power consumption...

Well, the quote is (if Google Translate is right), the power consumption is higher than expected for 7 nm. It didn't actually say it was higher than A11 though.

Apple probably has a more capable CPU architecture team than Intel or anybody else does. This is a pretty well known fact within the industry and the trend is probably going to further shift in Apple's favor as Apple is pretty ruthless about poaching the best talent in the industry, which is why they've grabbed so much Intel talent over the last five years. Now that AMD is getting interesting too, don't be surprised to see some high profile poaching going on from there, too

Since when is this "fact"? Apple has great people, sure, but that's not the same thing as "pretty well known fact within the industry" "Apple probably has a more capable CPU architecture than than Intel or anybody else does".

 

[Thala]

Aren't Samsung / QQ / ARM in competitive? If they could design and release such core in reasonable power envelope, users would flock to buy phones with it.

Within Android everyone is using stock ARM cores except Samsung. So far there was no incentive for ARM to go much wider. Besides the announcement of the next Cortex A is not far away.

 

[Lodix]

If ARM follows the same cadence this year, they will release the next core later this May.

 

[FIVR]

Apple's quarterly revenue is roughly equal to Intel's annual revenue (Apple's best quarter in a year is greater than Intel's entire annual revenue).

AMD's annual revenue is about a tenth of Apple's quarterly revenue, worst case.

So, really, what's the problem here? Apple is bigger and richer than Intel and puts out better processor cores.

If you compare R&D budgets it looks even worse for intel. They spend over 20x what AMD spends and they probably spend close to what Apple spends on chip development.


Apple did 9 billion in revenue from services alone last quarter. I think 2016 AMD revenue was like 10B?


You can also compare the number of employees if you would like to make intel look worse further still. The idea that AMD should be able to compete with Apple "if intel can" is laughable on its face if you know anything about business.

 

[ksec]

Since when is this "fact"? Apple has great people, sure, but that's not the same thing as "pretty well known fact within the industry" "Apple probably has a more capable CPU architecture than than Intel or anybody else does".

It is a fact when you are inside the "industry". Intel loses talent when their roadmap and product is no longer interesting and matters little years ago. The best talents often aren't just about money, they want challenges and make a difference. And there is no better place then Apple at the time. ( Now it may be differenent when Intel is trying make some changes ) And I don't think you need to be industry insider to know how poor Intel product management is, and it is not hard to find ex-Intel employees who agrees with that.

So, why should we believe this post? Is this person a known reliable poster, or just some random forum person?

I should have stated it earlier. This person is well known for any SoC information, mostly from Snapdragon and Apple A Chip. He was spot on for A10 and A11. I was actually surprise the information has no mention of Eight Core chip that were rumoured earlier. And it had more core, the MT results should be much higher.

BTW, those numbers are closer to ~21% higher than A11. If just a node shrink with no IPC change, then that would translate to a 2.9 GHz A12.

He actually mentions it somewhere in the comment it is closer to 25%. Making this 25% would be 3Ghz. The latter is what I calculate to be around 3Ghz, although he did mention performance coming from better branch prediction, and no mention of clock speed.

 

[ksec]

If you compare R&D budgets it looks even worse for intel. They spend over 20x what AMD spends and they probably spend close to what Apple spends on chip development.

Apple's R&D is doing chip design, you need to add TSMC Capex and R&D spending, as well as many other foundry partners in the ecosystem to compare to Intel. Which of coz is much larger then what Intel is currently spending on.

 

[Roland00Address]

Intel has figured out how to get these chips to turbo up to 5Ghz for momentary benchmarks but when they run actual applications it turns into a disaster. Go buy a 12" macbook and you will quickly find out why "fanless" and "intel" should never have been combined in the same product.

I know of people who have the 3rd generation macbook 12" (2017 model, 7th gen intel) who also had the 1st generation macbook 12" (2015 model, 5th gen intel) and the CPU performance is much better for it is much more consistent. They keyboard is also better.

Now I do not own either products and I probably would be very picky but my point for bringing this stuff up is incremental change can happen enough between generations that your first impressions of a product with the 1st generation can be relatively fixed by the 3rd or 4th generation given enough incremental improvement.

That said I am all for ARM macbooks instead of intel, and I am all for a brand new keyboard. (Also Touch macOS but that is a dream but ARM for Mac and a new keyboard I believe is possible in the near reality not as an engineering feat but very high in the likely to occur feat.)

 

[dark zero]

If Samsung is smart, they should start with the following things:
- Talk with Microsoft in order to make their cores compatibles with Windows in order to make their own laptops. Their own cores works perfectly on Laptops due the size and power consumption.
- Make a custom small core in order to be power efficient and having better performance.
- Talk with AMD in order to get a decent GPU with low consumption. If not, they would ask Power VR to do that.

That would force ARM and Qualcomm to finally make an advance

 

[Eug]

It might not be 10x but it's at least 3-5x more heat produced by these Core M chips than the iPad. I own both (core M macbook and iPad pro) and the iPad is both faster and runs cool, which is very unlike the macbook. The macbook can become intolerably hot from 30 minutes of light microsoft excel work.

Intel has figured out how to get these chips to turbo up to 5Ghz for momentary benchmarks but when they run actual applications it turns into a disaster. Go buy a 12" macbook and you will quickly find out why "fanless" and "intel" should never have been combined in the same product.

The 2015 Core M MacBook is not great. Its CPU and keyboard are why I didn't buy it. The CPU is at best mediocre and the keyboard is horrible. The 2017 Core m3 MacBook is quite a bit nicer however, on both counts, and it adds hardware 10-bit 4K HEVC decode too. It also does pretty well with sustained workloads. We actually tested this, by repeating Cinebench R15 for extended periods, over 30 minutes. Remember also this is not with a 4.5 W TDP, since Apple uses TDP-up.

"m3 (wood)" is my Core m3 on a pine table, and "m3 (granite)" is my Core m3 on a granite countertop, the latter helping to conduct heat away from the MacBook. You'll see that even with the insulating wood tabletop, the benchmark score only drops after 10 runs from 265 to 246, which is only a 7.2% drop. IMO, that's pretty decent for a fanless laptop.

On the flip side, I couldn't easily get through a couple of HEVC videos I had that I was trying to play back using software playback on my iPhone 7 Plus. It'd play perfectly fine for about 10-15 minutes, and then stutter like mad. I'd let the phone cool for 5 minutes, and then it'd play fine again, so obviously it was overheating. The good news is nowadays hardware HEVC decoding should be available to third party players, but the point is it didn't do great when forced to run for extended periods under sustained load. These chips are great, but they're not magic.

Fortunately, it probably does better in a 13" iPad Pro though, and the same can be said about A12 in a 13" MacBook.

BTW, what are the comparative die sizes? A11 on TSMC 10 nm is 88 mm2. A10X is 96 mm2 on 10 nm, and A10 is 125 mm2 on 16 nm. It will be interesting to see what's the size of A11X, and of course, A12 too.

 

[StinkyPinky]

So their single core is matching the 8700K? I struggle to believe that. I think this just shows the flaws of Geekbench as a benchmark software.

As for the macbook, the one port makes it useless as far as I'm concerned.