Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

iPadDad

macrumors 6502
Original poster
Sep 8, 2014
313
156
I thought this interesting, i was looking at the Geekbench 3 bench mark results for the iPad Air 2 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=A8x

This is where it gets interesting;

From what i could see the multicore score averages looked around;

iPad Air 2 - 4570 - http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=A8x
Macbook Air - 4740 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=2014+Macbook+air
iMac -4820 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=2014+imac

Anyway thought this was interesting,
 

kelon111

macrumors 6502
Mar 16, 2013
303
4
I thought this interesting, i was looking at the Geekbench 3 bench mark results for the iPad Air 2 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=A8x

This is where it gets interesting;

From what i could see the multicore score averages looked around;

iPad Air 2 - 4570 - http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=A8x
Macbook Air - 4740 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=2014+Macbook+air
iMac -4820 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=2014+imac

Anyway thought this was interesting,

Benchmarks can be misleading.
An Intel Core i5 has many more instruction sets than an Apple A8x has , meaning in some cases the Core i5 would easily blow the A8x out of the water in certain tasks that use those instructions.
 

Boyd01

Moderator
Staff member
Feb 21, 2012
7,635
4,528
New Jersey Pine Barrens
My Volkswagen can go just as fast as a tractor trailer truck. If I tried to haul a few tons of freight, I would probably discover that it isn't "as powerful" though. :D

I suspect we would get similar results comparing my MBA and your iPad. Try running a bunch of programs at the same time, or transferring data at 100MB/sec over a network, or putting a 300GB iTunes library on it.

The processor benchmarks may look the same, but I don't think that equates to "as powerful".
 

iPadDad

macrumors 6502
Original poster
Sep 8, 2014
313
156
Sure, but nevertheless that's pretty impressive for a tablet
 

motrek

macrumors 68030
Sep 14, 2012
2,613
305
I thought this interesting, i was looking at the Geekbench 3 bench mark results for the iPad Air 2 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=A8x

This is where it gets interesting;

From what i could see the multicore score averages looked around;

iPad Air 2 - 4570 - http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=A8x
Macbook Air - 4740 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=2014+Macbook+air
iMac -4820 http://browser.primatelabs.com/geekbench3/search?utf8=✓&q=2014+imac

Anyway thought this was interesting,

It is interesting and impressive.

The problem is that you're looking at the multi-core benchmarks. Multi-core performance is basically irrelevant for day-to-day use. You really want high single-core performance.

The multi-core numbers are skewed towards the iPad because it has 3 cores vs. 2 cores for the other computers.
 

Woochoo

macrumors 6502a
Oct 12, 2014
544
502
It is interesting and impressive.

The problem is that you're looking at the multi-core benchmarks. Multi-core performance is basically irrelevant for day-to-day use. You really want high single-core performance.

The multi-core numbers are skewed towards the iPad because it has 3 cores vs. 2 cores for the other computers.

This. Single cores of ARM perform almost the half of the Intel ones. Plus they haven't same instruction set. The day those 2 things are similar (with similar or even less power consumption), ARM will be really competing with Intel, not before.
 

motrek

macrumors 68030
Sep 14, 2012
2,613
305
This. Single cores of ARM perform almost the half of the Intel ones. Plus they haven't same instruction set. The day those 2 things are similar (with similar or even less power consumption), ARM will be really competing with Intel, not before.

Instruction set is irrelevant. Adding two numbers is adding two numbers in any language.
 

motrek

macrumors 68030
Sep 14, 2012
2,613
305
Not every instruction involves a direct addition.
Some involve dot products , matrix multiplication , etc.

Right, no, not every instruction is ADD. I was just trying to make a point.

Processors all have basically the same instructions, just formatted differently.

Sometimes a processor will have some instruction that makes something faster. For example, some processors have an instruction that counts the number of bits that are set in a word (generally referred to as POPCNT). This makes certain cryptography operations faster for some reason I don't understand.

If a processor doesn't have a particular instruction, it's because the engineers didn't think the performance increase was worth the extra complexity and die area to implement the instruction. Not because they didn't know such an instruction didn't exist or because they couldn't figure out how to implement it or something.

In other words, Intel processors aren't faster than ARM because they have a bunch of super-secret amazing instructions that only Intel knows how to make.

BTW -- no processor has a matrix multiply instruction. That's an operation that's a million times too complicated to have as one instruction, and there would be no benefit to having it as one instruction vs. several.
 

2IS

macrumors 68030
Jan 9, 2011
2,938
433
Pretty sure some instructions [SETS] are licensed, so it's not simply a matter of a CPU maker not using them because they thought it wasn't important or that there was no benefit. Cost is the main concern. Bottom line is no, ARM is nowhere near as powerful as a modern i5
 
Last edited:

motrek

macrumors 68030
Sep 14, 2012
2,613
305
Pretty sure some instructions are licensed, so it's not simply a matter of a CPU maker not using them because they thought it wasn't important or that there was no benefit. Cost is the main concern. Bottom line is no, ARM is nowhere near as powerful as a modern i5

No, you can't have intellectual property rights on an instruction. That makes as much sense as getting a patent for addition.

You can get rights on an instruction SET, which is how instructions are encoded.

So you can't make an ARM-compatible chip without a license from ARM but you could make a "MRA" chip with all the same instructions and just a different encoding. It wouldn't be able to run ARM software without recompiling but there's nothing that would prevent it from performing just as well.

The reason Intel chips are faster than ARM is because Intel has been focused on making the fastest processors possible for ~30 years with little concern about power consumption and die size. They have hired the best engineers in the world to accomplish this and their architecture is designed from the ground up for high performance.

On the other hand ARM has always been focused on making small, low-power chips. Of course they have always tried to maximize performance given their power and die size budgets but that's a different ball game than designing chips for supercomputers.

Now both companies are headed in each others' directions, with Intel doing a massive push to reduce power consumption and ARM pushing to increase performance. But just as Intel isn't getting its Core chips down into the 2W range, ARM hasn't caught up to Intel's 30 year head start on high-performance computing.
 

2IS

macrumors 68030
Jan 9, 2011
2,938
433
No, you can't have intellectual property rights on an instruction. That makes as much sense as getting a patent for addition.

You can get rights on an instruction SET, which is how instructions are encoded.

So you can't make an ARM-compatible chip without a license from ARM but you could make a "MRA" chip with all the same instructions and just a different encoding. It wouldn't be able to run ARM software without recompiling but there's nothing that would prevent it from performing just as well.

The reason Intel chips are faster than ARM is because Intel has been focused on making the fastest processors possible for ~30 years with little concern about power consumption and die size. They have hired the best engineers in the world to accomplish this and their architecture is designed from the ground up for high performance.

On the other hand ARM has always been focused on making small, low-power chips. Of course they have always tried to maximize performance given their power and die size budgets but that's a different ball game than designing chips for supercomputers.

Now both companies are headed in each others' directions, with Intel doing a massive push to reduce power consumption and ARM pushing to increase performance. But just as Intel isn't getting its Core chips down into the 2W range, ARM hasn't caught up to Intel's 30 year head start on high-performance computing.

Pretty sure you know exactly what I was referring to when I said instructions. I'll go ahead and edit my post and add the word "sets" to appease you though, my point stands.

Sure you can make a processor crunch whatever instruction you want, but without using certain instruction sets, it could potentially be a whole lot less efficient in doing so.
 

iPadDad

macrumors 6502
Original poster
Sep 8, 2014
313
156
doubt IOS apps can take advantage of multicores like OS X apps

Dont forget you get iPhone Apps and iPad Apps. the newer iPad Apps especialy are completely redesigned to utilise the screen size and the processing capability of the newer iPads. A typical example of this is Pixelmator. Pixelmator is available for iPad only because of the exact reasons above.

I suspect we are going to be seeing the bridge between iPhone software (apps) and iPad software widening more and more.
 

motrek

macrumors 68030
Sep 14, 2012
2,613
305
Pretty sure you know exactly what I was referring to when I said instructions. I'll go ahead and edit my post and add the word "sets" to appease you though, my point stands.

Sure you can make a processor crunch whatever instruction you want, but without using certain instruction sets, it could potentially be a whole lot less efficient in doing so.

No, I didn't know you were referring to instruction sets, and you don't seem to know what that term means, so I don't think you meant instruction sets either.

A processor only supports one instruction set (with rare and irrelevant exceptions) and nobody can use multiple instruction sets to make anything more efficient.

http://en.wikipedia.org/wiki/Instruction_set

----------

doubt IOS apps can take advantage of multicores like OS X apps

They can. iOS has all the same multithreading support and programming interfaces as OS X. In fact they are literally the same.
 

2IS

macrumors 68030
Jan 9, 2011
2,938
433
No, I didn't know you were referring to instruction sets, and you don't seem to know what that term means, so I don't think you meant instruction sets either.

A processor only supports one instruction set (with rare and irrelevant exceptions) and nobody can use multiple instruction sets to make anything more efficient.

http://en.wikipedia.org/wiki/Instruction_set

----------



They can. iOS has all the same multithreading support and programming interfaces as OS X. In fact they are literally the same.

SSE, SSE2, SSE3, AVX, AVX2 and all these other extensions don't do anything then? Sounds to me like you're nit picking on terminology. Superiority complex much?

Call it what you want, i5 > ARM... You can provide reasons for why, but that doesn't matter. ARM is not as fast as an i5
 

motrek

macrumors 68030
Sep 14, 2012
2,613
305
SSE, SSE2, SSE3, AVX, AVX2 and all these other extensions don't do anything then? Sounds to me like you're nit picking on terminology. Superiority complex much?

Call it what you want, i5 > ARM... You can provide reasons for why, but that doesn't matter. ARM is not as fast as an i5

No, I don't have a superiority complex. I have a degree in electrical engineering and I studied processor design. So I know about this stuff. And from that perspective, the posts you're writing aren't making any sense. You might think I'm nit-picking and making irrelevant points about terminology but really it's like you're talking about anatomy with a doctor and you keep calling muscles bones and then you get upset at the doctor when he doesn't understand your "terminology."

Yes, you're right, in certain contexts you could refer to SSE (and all the others) as an instruction set. But in this context it makes no sense. SSE doesn't give Intel any advantage over ARM and there's no reason why ARM would want to implement SSE. It makes no sense. It would be like putting an outboard motor on the back of your television.

SSE, SSE2, etc. is simply Intel's encoding for vector instructions. ARM has their own vector instruction set called NEON that does all the same stuff, so who cares?

You're right, the A8 is not as fast as an i5. I explained the reason in broad strokes above. But it has nothing to do with instruction sets. In fact, the ARM instruction set is far superior to the i5's instruction set which is x86. If anything, the i5 is working with a fairly significant handicap by having to support x86.
 

Nee412

macrumors 6502
Jun 25, 2010
281
8
Sunny England!
The A8X isn't as fast as an i5, but Apple is getting closer with every A-series release. Eventually Apple will be on on par with Intel.

That's not to say that the A8X isn't an impressive chip. There's hardly any apps in the AppStore that truly push what it is capable of, and none that it can't run with impressive ease. For example when considering the average consumer Pixelmator on the iPad Air 2 is pretty much a capable replacement for Photoshop. Obviously there are some features that mean professional photographers will still use Adobe, but that doesn't make Pixelmator any less impressive.

Honestly I'm impressed at the difference between the iPad Air 2 and my 2 year old iPad Mini. Imagine what we will have in another two years...
 

iPadDad

macrumors 6502
Original poster
Sep 8, 2014
313
156
The A8X isn't as fast as an i5, but Apple is getting closer with every A-series release. Eventually Apple will be on on par with Intel.

That's not to say that the A8X isn't an impressive chip. There's hardly any apps in the AppStore that truly push what it is capable of, and none that it can't run with impressive ease. For example when considering the average consumer Pixelmator on the iPad Air 2 is pretty much a capable replacement for Photoshop. Obviously there are some features that mean professional photographers will still use Adobe, but that doesn't make Pixelmator any less impressive.

Honestly I'm impressed at the difference between the iPad Air 2 and my 2 year old iPad Mini. Imagine what we will have in another two years...

So true, the advancement in tablet chip design and especially performance is amazing, i am really looking forward to seeing what the A9 chip will bring,
 

2IS

macrumors 68030
Jan 9, 2011
2,938
433
Wasn't terribly difficult to predict that ARM would run into diminishing returns just as Intel has. A lot of people thought ARM was the second coming of christ especially with Apple designing their own chips, the apple fans got carried away. They both have to abide by the laws of physics, and Intel has more experience in this area than anyone. Pretty silly to think that ARM based chips were just going to blow past high end desktop processors in short amount of time.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.