Reading the spec's and geekbench scores coming in on the A10X chip. I can't help but feel this is the first serious chip that might make it's way into a MacBook (or is at least powerful enough) then maybe the Pro's later or at least it's making the case (pun intended) for Apple ARM A-Chips in their computers.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Is A10X first MacBook ready and capable chip?
- Thread starter Omega Mac
- Start date
- Sort by reaction score
Apple's chip designers are beyond compare. With TDP budgets around 10w they have been able to ramp up processor performance by an insane 500x over the last 5-6 years. Intel over the same duration has only managed a pedestrian 20 - 30% performance increase, year on year, even with much higher TDP budgets. I will not be surprised if the A10X runs circles around an i9 or a Xeon in performance, with only a 10w TDP.
Eh... no.
Mobile benchmarks live in a very different environment compared to desktop benchmarks. On a mobile device, an app essentially gets >90% of the device's resources at its disposal. In contrast, on desktop, an app may only be able to access anywhere between 10 -> 70% of the device's resources, but not much more.
iOS is a much more lightweight version of MacOS as well. If the A10X was tasked with running 4x the background processes, do you think it'll still run that fast?
There's a reason Apple has not put an ARM chip in the MacBook yet, despite those benchmark scores. They know it's not fast enough yet.
If they could, they would have done it already.
At the current rate of growth, we are pretty far away from an ARM-only MacBook, though the MacBook Pro with Touch Bar already has an ARM SoC to take care of the Touch Bar.
Mobile benchmarks live in a very different environment compared to desktop benchmarks. On a mobile device, an app essentially gets >90% of the device's resources at its disposal. In contrast, on desktop, an app may only be able to access anywhere between 10 -> 70% of the device's resources, but not much more.
iOS is a much more lightweight version of MacOS as well. If the A10X was tasked with running 4x the background processes, do you think it'll still run that fast?
There's a reason Apple has not put an ARM chip in the MacBook yet, despite those benchmark scores. They know it's not fast enough yet.
If they could, they would have done it already.
At the current rate of growth, we are pretty far away from an ARM-only MacBook, though the MacBook Pro with Touch Bar already has an ARM SoC to take care of the Touch Bar.
Beware Touchbar nowadays only use the ARM SoC control the electric behavior of Touchbar hardware (such as backlight control and power monitor), it just like Apple use ARM chip for BIOS or ARM chip for LCD display, the actually rendering process take care from Intel CPU and GPU, it just like another small display devices.
You ramblings is from a person who doesn't understand anything about CPU of OS design. One look at activity manager shows that an intensive app like when encoding can take 100% load of a CPU on a MacBook not his nonsense 10-70% made up figure.
A10x can handle the load fine. Only reason why they haven't moved is application compatibility and the exclusion of running boot camp.
Running boot camp may not be a priority if ever for those who want a MacBook... think about it.
Eh... did you know an actual intensive app can take 200% or even 300%? Or that the 100% may actually be 100% of "non-boosted clockspeed" because the performance of the CPU varies in order to save power?
A10X can't do it, no matter how much you're wishing it could.
The real reason is plain and simple: it just can't.
It's not bootcamp or anything of the sort. In fact, Apple could care less about Bootcamp if you judge the state of Bootcamp drivers now. They're only providing Bootcamp support because their customers are complaining about it, not because they want to.
Backward compatibility? No, not that either. Remember Apple cut off all PowerPC apps from their OS a long time ago?
Apple isn't shy of doing crazy things like taking away the 3.5mm port or making their computers all USB-C, or pushing a useless Touch Bar on users. They have proven that they can do it. That is, if it makes sense.
Again, A10X is not as powerful as people are wishing for it to be. An ARM-only MacBook is still pretty far away. If it's 2x its current performance, then maybe. But considering the current rate of growth, it would be years before that happens.
@bill-p - I am thus assuming from your comments that geek bench scores between devices are not to be compred...the latest A10X chip is not comparable to a geek bench score of any particular i5/i7 macbook or even better the mobile intel chips in the Macbook are more powerful than the mobile Apple A10X chip?
I get they main be desinged/tuned to a different demand environment. Ok but I'm running a Macbook '10 on a core 2Duo and it's it doesn't even touch the scores coming out of the iPhone let alone the IPad Pro.
How are we to read these geek bench scores? What is the point of geek bench other than to sell apps (maybe I answered that question). I thought the whole value is it creates a comparable score sheet that work off a baseline.
I'm also noting the comment on stage that the A10X had four times the GPU power of an i7 chip (or some such sentiments), it was an almost throw away comment by the Affinity deme guy and maybe Apple don't want to stand over it as it's a general marketing point. I don't know but they demonstrated so much you could see everyone was racing more than normal.
As per other threads on the issue, what stops Apple releasing a beefier version of a AX Chips... which is what they do for the iPad Pro's they get extra cores. Put that in a macbook with more space and more battery the potential increases, you could have dual and quad A-Chip MacBooks. The thing is passively cooled. If that's an issue then in laptop it becomes less so with fans on hand to pitch in.
I get they main be desinged/tuned to a different demand environment. Ok but I'm running a Macbook '10 on a core 2Duo and it's it doesn't even touch the scores coming out of the iPhone let alone the IPad Pro.
How are we to read these geek bench scores? What is the point of geek bench other than to sell apps (maybe I answered that question). I thought the whole value is it creates a comparable score sheet that work off a baseline.
I'm also noting the comment on stage that the A10X had four times the GPU power of an i7 chip (or some such sentiments), it was an almost throw away comment by the Affinity deme guy and maybe Apple don't want to stand over it as it's a general marketing point. I don't know but they demonstrated so much you could see everyone was racing more than normal.
As per other threads on the issue, what stops Apple releasing a beefier version of a AX Chips... which is what they do for the iPad Pro's they get extra cores. Put that in a macbook with more space and more battery the potential increases, you could have dual and quad A-Chip MacBooks. The thing is passively cooled. If that's an issue then in laptop it becomes less so with fans on hand to pitch in.
Eh... did you know an actual intensive app can take 200% or even 300%? Or that the 100% may actually be 100% of "non-boosted clockspeed" because the performance of the CPU varies in order to save power?
A10X can't do it, no matter how much you're wishing it could.
The real reason is plain and simple: it just can't.
It's not bootcamp or anything of the sort. In fact, Apple could care less about Bootcamp if you judge the state of Bootcamp drivers now. They're only providing Bootcamp support because their customers are complaining about it, not because they want to.
Backward compatibility? No, not that either. Remember Apple cut off all PowerPC apps from their OS a long time ago?
Apple isn't shy of doing crazy things like taking away the 3.5mm port or making their computers all USB-C, or pushing a useless Touch Bar on users. They have proven that they can do it. That is, if it makes sense.
Again, A10X is not as powerful as people are wishing for it to be. An ARM-only MacBook is still pretty far away. If it's 2x its current performance, then maybe. But considering the current rate of growth, it would be years before that happens.
Do you even know what 300% means? The MacBook has 4 threads total for 400% total so that 300% means that task is loading 75% of the capacity of all the CPUs in the MacBook. A far cry from your 10-70% assertion before. I've seen some app that can load close to maximum of all available resources..ie close to 400% in case of the MacBook.
[doublepost=1496951144][/doublepost]
I run it and I know quite a lot of my colleagues who do it as well. Just a small sample. Not sure where you get your data from.
75% is a "far cry" from 70%? I see...
Did you also know that even when pegged at 400% full speed, the processor may actually NOT be at its full speed because Apple intentionally limits the thermal profile of the processor? And that the processor varies its speed depending on whether it needs more?
Did you also notice that in Activity Monitor, there's a split between User % and System % to illustrate how much of those resources system processes are taking up when an app does anything?
Again, if Apple thought it was ready, they would have either done it already (saves them a lot of margin because they have to pay Intel a hell of a lot for those Core M chips), or added even more desktop-like features to iOS 11.
Last but not least, I have to remind you again that we are basing all of this off comparisons between a mobile version of a benchmark against its desktop version. Definitely not the same thing.
Some persons thing that replaced Intel CPU to ARM CPU in MacBook can will reduced the cost of the system and already ask when is the MAC system go for ARM.
One of the very bad thing is Apple and developer need test and tweak two type of CPU, and it dynamic increase the cost of Apple and should reflected on product price.
The Geekbench is totally useless and designed for misleading consumer. (Which why manufacturer like this )
Firstly it test every sub-test multi thread by sequence (which the ideal of multi-process CPU),
so the CPU only use a small part in every sub-part,
also most of the sub-test can use DSP/GPU core for improve speed dynamic (for example basic image process )
which good for single task multi thread,
and the test of every sub-test is too short, it can not reflect the normal activities of devices which many threads and different task work together and some task need single thread and work it first
(Actually many User activity can not be multithread).
Also the score is not mean everything, if the memory speed increase a little, the score will increase rapidly,
but for real case this can not help a lot of speed especially nowadays the memory speed is so fast enough for CPU use.
One of the very bad thing is Apple and developer need test and tweak two type of CPU, and it dynamic increase the cost of Apple and should reflected on product price.
The Geekbench is totally useless and designed for misleading consumer. (Which why manufacturer like this )
Firstly it test every sub-test multi thread by sequence (which the ideal of multi-process CPU),
so the CPU only use a small part in every sub-part,
also most of the sub-test can use DSP/GPU core for improve speed dynamic (for example basic image process )
which good for single task multi thread,
and the test of every sub-test is too short, it can not reflect the normal activities of devices which many threads and different task work together and some task need single thread and work it first
(Actually many User activity can not be multithread).
Also the score is not mean everything, if the memory speed increase a little, the score will increase rapidly,
but for real case this can not help a lot of speed especially nowadays the memory speed is so fast enough for CPU use.
With the benchmarks that appear to show A10X performance, I would like to see how it would perform under macOS. Although I would suspect they would maybe use a dual core variant, and the single core processor hasn't improved that much so the A10X wouldn't be ready.
It is not 75% vs 70%. It is close to 100% vs your Nonsense 10-70%
Since geekbench 4 mobile and desktop versions run exactly the same benchmark. MacOS while running the benchmark uses all available resources so the benchmarks are comparable. I have no idea why you think they are different.
If you get a third party resources program for iOS it will also show user/system resources etc. There is no difference. If macOS was running geekbench on a10x it would get the same or within very small margin of error than it would on iOS.
There many other consideration than just speed. Biggest is software compatibility.
Eh... just... no.
Here's my MacBook m7 with GeekBench 4 ran just this morning:
https://browser.geekbench.com/v4/cpu/3066613
And here's the same exact m7 this afternoon, same conditions:
https://browser.geekbench.com/v4/cpu/3068957
What's going on? My scores improved by a whopping 500 points. In fact, my scores for the m7 are significantly higher than the average as well. Do I have a magical m7 MacBook? One that decides to speed up during the afternoon when ambient temperature is even HIGHER than in the morning??
Seriously, it's just the overhead of the OS, good sir, and it does change from time to time. Those of us who have used a desktop OS for years know this. And this is why a mobile OS like iOS is stripped down so as to optimize for the performance of the device it's running on.
If you still don't believe that the OS has that much of an effect, then I do implore you to get a MacBook and try running MacOS. You'll see immediately why the A10X is not ready yet.
I'm not ruling out the possibility that Apple may move to ARM at some point in the future, of course. But again, they are not working magic here. A10X running on iOS is barely able to match Intel's slowest Core processors means it's a far cry from being able to squeeze itself onto the Pro line. And if Apple is going to jump ship to a new architecture for their Mac computers, it's gotta be on everything.
Say, you are in fact correct, and that A10X will in fact replace the equivalent chips in the MacBook. Will it be able to drive a 5K display on an iMac 27" while handling VR contents as well? Because that's where Apple is heading next.
Here's my MacBook m7 with GeekBench 4 ran just this morning:
https://browser.geekbench.com/v4/cpu/3066613
And here's the same exact m7 this afternoon, same conditions:
https://browser.geekbench.com/v4/cpu/3068957
What's going on? My scores improved by a whopping 500 points. In fact, my scores for the m7 are significantly higher than the average as well. Do I have a magical m7 MacBook? One that decides to speed up during the afternoon when ambient temperature is even HIGHER than in the morning??
Seriously, it's just the overhead of the OS, good sir, and it does change from time to time. Those of us who have used a desktop OS for years know this. And this is why a mobile OS like iOS is stripped down so as to optimize for the performance of the device it's running on.
If you still don't believe that the OS has that much of an effect, then I do implore you to get a MacBook and try running MacOS. You'll see immediately why the A10X is not ready yet.
I'm not ruling out the possibility that Apple may move to ARM at some point in the future, of course. But again, they are not working magic here. A10X running on iOS is barely able to match Intel's slowest Core processors means it's a far cry from being able to squeeze itself onto the Pro line. And if Apple is going to jump ship to a new architecture for their Mac computers, it's gotta be on everything.
Say, you are in fact correct, and that A10X will in fact replace the equivalent chips in the MacBook. Will it be able to drive a 5K display on an iMac 27" while handling VR contents as well? Because that's where Apple is heading next.
Err.. just no. Same thing can happen in iOS if you are running things in the background. Obviously you know nothing about scheduling of tasks. The scores would tank if I was encoding an imovie video while running geekbench. They are just OS with the same limitations.
I can get the same variation on my iPad pro. Just have a look at the all the results. It varies greatly. That's why geekbench is best when it is the only active application running.
[doublepost=1497062967][/doublepost]
Others considerations:
- a10X only has 4GB RAM
- connectivity .i.e thunderbolt, USB-C, external monitor support on GPU etc.
If Apple really wanted to target laptop users they would probably need another build with more RAM and connectivity but then there's the software compatibility to consider.
Performance is not an issue. The a10X is faster than the macbook's processor even the high end built to order options. The GPU speed also kills the macbook's.
[doublepost=1497063254][/doublepost]
Do you anything about geekbench?
- The memory weighting is 20%.
- only the crypto and a few of the floating point benchmarks are tuned to each architectures special instructions sets. The vast majority are not.
- the mobile and desktop workloads are the same.
Here is the benchmark description:
http://geekbench.com/doc/geekbench4-cpu-workloads.pdf
Last edited:
Wait, what? You said...
So which is it? Are the background tasks able to influence the results or not?
And also, in both of the above cases, GeekBench was the only thing running. I did say "same condition" for both. If there are more background tasks in the morning, I certainly do not see it in the log since both are run from a fresh restart of the MacBook to make sure.
So which is it? Are the background tasks able to influence the results or not?
And also, in both of the above cases, GeekBench was the only thing running. I did say "same condition" for both. If there are more background tasks in the morning, I certainly do not see it in the log since both are run from a fresh restart of the MacBook to make sure.
I never said background tasks didn't affect the results. I said geekbench will use all available resources. Your macbook might've been doing some background indexing for example. BTW iOS can also do this so the benchmarks are comparable
There's no much use arguing with a brick wall. I'm out of here. Just read up on OS scheduling of tasks and the geekbench benchmarks descriptions. iOS and MacOS share the same basic kernel that does the scheduling of tasks. There's nothing magical about iOS that would make it run Geekbench faster when there's are no other tasks running.
https://9to5mac.com/2016/09/15/iphone-7-geekbench-benchmark-scores/
"This would once have been a completely meaningless comparison: it used to be the case that there was no direct comparison between Geekbench scores for desktop and mobile devices. Geekbench has been working hard of late, though, to bring the two more in line, so this is now a reasonable comparison to make – at least in theory."
You're the brick wall, good sir. A lot of evidence has been presented to the contrary and you stubbornly refuse them.
On a MacBook, the performance of the processor varies depending on whether the OS "thinks" the task needs more, or not. And then there's thermal throttling as well. This is independent of the app. The app has no control over this. The OS has the final say. There's a reason why some developers keep calling Apple's system a "walled garden", though that's a different issue.
Please don't bring things you don't truly understand into the discussion as if they actually mean something. "OS scheduling of tasks" is meaningless in the face of a mobile OS like iOS because it doesn't actually let any task run in the background at all, or at least not for long. As soon as a foreground task says it wants some resources, guess what? The other task is shut down. Mercilessly. Unless it's a system task (you have to request permission to latch your own process to one of these system tasks), of which there are very very very few. Contrast this to a desktop OS where tasks continually run even when in the background. There's one difference between MacOS and iOS.
Don't believe me? Then you should be reading up:
https://developer.apple.com/library.../BackgroundExecution/BackgroundExecution.html
iOS and MacOS share the same kernel? True.
iOS and MacOS does the same scheduling of tasks? False. Very very very very very false.
That's the magical part about iOS.
And that's also why benchmarks ran on iOS compared to MacOS may never be comparable. You just admitted as much that my MacBook may be indexing in the background. Guess what? Geekbench has no control over that. It can't possibly shut down the indexing and ask for all of the resources of the computer to run the comprehensive test. On iOS, it does get close to that by default because the OS allows it to.
Please see torana355's excellent post below for the same discussion when the iPhone 7 came out.
Actually, let me leave you with something else as well:
https://twitter.com/MattMariska/status/776495257311584261
The iPhone 7 is faster than the latest Mac Pro. How's that? Heck, my fanless MacBook with Core m7 scored 3600 in Geekbench, so that means I have a computer that's even faster than the latest Mac Pro, right?
Well the number of cores is the reason for that result, the single core performance on the Apple chip is great however on modern OS's that are optimised for multicore the Apple chip would struggle. Its way more efficient to have 12 cores with slightly less power per core then to have 2 really fast cores as the load can be distributed and run things in parallel if the software is optimised for it. Intel could easily create a dual core chip that would destroy the A10 in Single core performance, its just not suited to a full desktop OS which is why we keep seeing more cores in the latest intel chips. Hell even the Samsung S8 has 8 cores.
Last edited:
1. The memory weight is not too high but it can help a lot, such as Memory Copy from 12GB to 16GB can help get few hundred. (which can help the sub-score of memory from 3500 to 4600)
But this nothing mean for anything.
2. Most of the testing is actually helped by hardware DSP in ARM cpu,
such as jpeg depression/compression, html5, camera graphic process....,
modern CPU use special circuit design for compression (LZMA)
,also the sqlite test is high affected by cache size,
Geekbench said that "Quickly Generating Billion-Record Synthetic Databases" but actually we rarely use sqlite and sqlite not optimum for this task, instead sqlite is design for light database for storage small information but frequency update and search for future use.
LLVM also useless, the code too small and no linking for reflect the real cause.
Also, i don't why Dijkstra is test for Integer, it all mislead by Geekbench.
May be Geekbench thing Floating Point = 3D application relatively, and other they said this is Integer.
Just read the PDF of the description of the tests. Most are not accelerated by dsps. Not JPEG, not html5 or any of the other tests you mentioned.
If the jpeg or camera tests was accelerated by DSPs, the a10 would be orders of magnitude faster
[doublepost=1497107898][/doublepost]
Let's keep it simple. For a Geekbench run if no other tasks are using significant resources - then the scores are comparable. This applies to both iOS and Mac.
iOS systems tasks like indexing can also run indefinitely. So can a lot of other tasks. How do you think running apps keep track of where you have been or message applications work or music in the background. Furthermore, a newer ipad can keep up to 4 apps (ipads with 4GB RAM) running indefinitely as well.
No - your macbook or iphone 7 can run single threads faster. It is obvious from the benchmark description.
[doublepost=1497108149][/doublepost]
What is said is totally correct. The scores are comparable when only considering CPU performance.
There are also other aspects like RAM capacity (a10X having only 4GB RAM) and SSD speed (I believe the SSD speed of iPad pros are comparable to the 2016 Macbooks but not macbook pros) which affects the overall experience.
Last edited:
Uh... no, please read the article that I linked above. It gives the answers to everything else you just asked and also showed you what I just said.
iOS doesn't run anything aside from its own system tasks in the background. Period. Even if the device has more memory, the system simply does not want to keep anything in the background at all.
If you need more to read:
https://apple.stackexchange.com/questions/252239/do-ios-apps-run-in-the-background
iOS indexing doesn't run indefinitely either:
https://developer.apple.com/library.../General/Conceptual/AppSearch/AppContent.html
It's very different from MacOS even though both run the same kernel. This is like saying Android and Linux are the same thing, but they aren't.
Again, Geekbench scores are NOT comparable. Even if Geekbench's developers would like you to believe otherwise.
Also fanless Core m7 is comparable to single-thread performance of Intel Xeon in Mac Pro = laughable. The reason why the single-core results are so high in my Core m7 results are because:
1. Dedicated accelerators are engaged, that means the video accelerator. This is only useful for video processing and has zero benefit elsewhere. It's not a secret that the Mac Pro's video accelerator is outdated and super slow by today's standards.
2. My SSD may in fact be faster than whatever is in the 2013 Mac Pro, and that allows certain tasks that are not bound by integer performance (like Dijkstra) to run faster. Dijkstra specifically and in essence requires "faster lookup" rather than "faster calculation", so things like faster memory, less latency, and faster storage will help a lot here. CPU performance? Not so much as long as it has enough I/O to handle the algorithm's lookups. Yet Geekbench has it put under "Integer performance", which is cringe-worthy.
We have dissected this as far as possible, and it has been discussed every single time a new iOS device with a new chip comes out... (we had this exact same talk the last time the iPhone 7's A10 results came out)
For the last time, Geekbench results and any benchmark should be taken with a grain of salt. If you still don't believe that despite the staggering amount of technical info pointing to otherwise, then I honestly don't know what else to say. Believe what you will, I guess.
iOS doesn't run anything aside from its own system tasks in the background. Period. Even if the device has more memory, the system simply does not want to keep anything in the background at all.
If you need more to read:
https://apple.stackexchange.com/questions/252239/do-ios-apps-run-in-the-background
iOS indexing doesn't run indefinitely either:
https://developer.apple.com/library.../General/Conceptual/AppSearch/AppContent.html
It's very different from MacOS even though both run the same kernel. This is like saying Android and Linux are the same thing, but they aren't.
Again, Geekbench scores are NOT comparable. Even if Geekbench's developers would like you to believe otherwise.
Also fanless Core m7 is comparable to single-thread performance of Intel Xeon in Mac Pro = laughable. The reason why the single-core results are so high in my Core m7 results are because:
1. Dedicated accelerators are engaged, that means the video accelerator. This is only useful for video processing and has zero benefit elsewhere. It's not a secret that the Mac Pro's video accelerator is outdated and super slow by today's standards.
2. My SSD may in fact be faster than whatever is in the 2013 Mac Pro, and that allows certain tasks that are not bound by integer performance (like Dijkstra) to run faster. Dijkstra specifically and in essence requires "faster lookup" rather than "faster calculation", so things like faster memory, less latency, and faster storage will help a lot here. CPU performance? Not so much as long as it has enough I/O to handle the algorithm's lookups. Yet Geekbench has it put under "Integer performance", which is cringe-worthy.
We have dissected this as far as possible, and it has been discussed every single time a new iOS device with a new chip comes out... (we had this exact same talk the last time the iPhone 7's A10 results came out)
For the last time, Geekbench results and any benchmark should be taken with a grain of salt. If you still don't believe that despite the staggering amount of technical info pointing to otherwise, then I honestly don't know what else to say. Believe what you will, I guess.
Nothing in that links says indexing cannot run indefinitely. You might be right but in any case - indexing can run for a longer period than a geekbench run.
iOS can run certain third party background apps for up to 10 minutes - which again is more than a geekbench run. So background tasks in iOS can affect the results.
Read the benchmark description. No video accelerators are engaged in Geekbench. If you want to see how good the video accelerators are - there is now Geekbench compute benchmark that attempts to do this.
There are certain sub benchmarks that uses the chip's SIMD instructions. The crypto sub benchmarks uses the dedicated crypto units. This is a real life scenario. Any application that does crypto would definitely use the crypto unit. Same for applications that can be accelerated via SIMD.
In fact - one could argue that geekbench disadvatanges iOS devices by not using the imaging DSP in Ax chips. Normal iOS camera/photo applications would use this. Intel does not have an imaging DSP on chip.
I believe that's why the new Affinity photo application - in the keynote the developers said that certain operations on the iPad Pro are 4X faster than a high end i7 chip.
Geekbench does not test the speed of the SSD at all. Everything is done in memory.
This is all a red herring from you. The bottom line as I said is that if there are no significant background tasks running - the scores are comparable.
Last edited: