• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."
  • Community Question: What makes a good motherboard?

Apple A12 & A12X *** Now A12Z as well *** Now in a Mac mini

Page 12 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Eug

Lifer
Mar 11, 2000
23,015
536
126
I would love to see the innards of the A12Z Mac mini, but Apple strictly forbids disassembling these units.
 

Doug S

Senior member
Feb 8, 2020
466
674
96
I would love to see the innards of the A12Z Mac mini, but Apple strictly forbids disassembling these units.
It is probably not all that different than the teardown of the latest iPad just with some more RAM and flash. Apple has no reason to spend any effort on producing a good "product" for something that will be used for porting and the devs won't even be able to keep.
 

Eug

Lifer
Mar 11, 2000
23,015
536
126
Interestingly, Apple now supports VP9 video decoding in iOS/iPadOS 14, finally allowing for 4K YouTube.

Presumably this is hardware support, and presumably that hardware support has been there in the hardware for 5 years.
 

Doug S

Senior member
Feb 8, 2020
466
674
96
Interestingly, Apple now supports VP9 video decoding in iOS/iPadOS 14, finally allowing for 4K YouTube.

Presumably this is hardware support, and presumably that hardware support has been there in the hardware for 5 years.
What's the point of supporting 4K YouTube on a device that doesn't have a 4K display?
 

gdansk

Senior member
Feb 8, 2011
549
227
116
Interestingly, Apple now supports VP9 video decoding in iOS/iPadOS 14, finally allowing for 4K YouTube.

Presumably this is hardware support, and presumably that hardware support has been there in the hardware for 5 years.
They also added WebP support in Safari. I guess they gave up that fight.
 

Eug

Lifer
Mar 11, 2000
23,015
536
126
Aircast it to your 4k TV
AppleTV 4K (A10X) also supports 4K YouTube now (or at least when tvOS 14 is released). It didn't before.

Anyhow, the point is I assume the support is hardware decode, and who knows how long it's been buried there in the SoC. A10 is from 2016, and A10X is from 2017. I'm not sure about the earlier chips.
 

Eug

Lifer
Mar 11, 2000
23,015
536
126
Now that I think of it, it's not 100% clear that it's actually VP9 support. It could be AV1 I suppose.

Either way though, if true, that means Apple had implemented the hardware support many years ago, and just hadn't bothered to turn it on until they were ready.

The other option is that YouTube is serving up h.265, but I doubt it. The reason I was favouring VP9 is because AV1 wasn't actually out until 2018.
 

IntelUser2000

Elite Member
Oct 14, 2003
7,250
1,839
136

Man, I don't think people are going to complain about pitfalls regarding binary translation when that becomes true. If you can run same thing you do on your Mac desktop as you do on a phone. Sure, its perceived in a less excited manner than if Windows could do that but still.

-Say you are moving to your own ARM chip
-Then announce iPad/iPhones that can run Mac OS
-Few years when the developers move to ARM64 binaries then people will run Mac productivity applications on iPads! Consider that lots of graphics machines are on Mac. Then it'll compete against PCs without needing x86.

If this doesn't light a fire under Intel/AMD then they're in real trouble. AMD's doing ok now but Intel needs to get their act together fast. No more Netburst or Bulldozer shenanigans.
 
  • Like
Reactions: lightmanek

Doug S

Senior member
Feb 8, 2020
466
674
96
If this doesn't light a fire under Intel/AMD then they're in real trouble. AMD's doing ok now but Intel needs to get their act together fast. No more Netburst or Bulldozer shenanigans.
If Apple was successful with this what could Intel/AMD do? They couldn't do anything. They'd need to first make an x86 phone SoC for Android (since Windows Phone is toast) then convince Android OEMs to adopt it, then convince Android customers to buy it, then convince Microsoft to produce a Windows "app" for Android that let people run Windows software on it when connected to a big display, keyboard and mouse.

That ship sailed, Intel is out of the phone SoC business, the Android market is all-in on ARM, and Microsoft has no incentive to produce this app for Android - because trading one Windows license for another means they don't make any more money but have a huge new support headache with all the random hardware configs of the Android world.
 

beginner99

Diamond Member
Jun 2, 2009
4,705
1,104
136
What can they do? They can continue to execute so in the PC market it'll not create room for WoA efforts to succeed.
Exactly. The advantage with Apple (vertical integration) will now also become a huge issue of vendor-lock in. Your mac device is now vendor-locked. All you can do is run a Linux ARM VM. boot-camp is probably dead as well as Hackintosh.

On top of that I never really got the idea of having only 1 device being all that cool. Especially now with cloud services isn't unneeded. You can read your mails and calendar on the phone and read documents stored on the cloud (not like I want to edit power point on my phone). Much more efficent to have 2 or more devices best at doing their specific job over a jack of all trades master of none.

Plus let's not forget the more smartphones and other ARM-based devices are out there, the more servers are needed. Servers which mostly run x86 (for now) and have far higher margins. Since the mobile ship has sailed, AMD rightly so bet their cards on server.
And often the real complexity of software is server-side. X86 software which you simply don't want to waste millions or billions to migrate for maybe some minor power usage gains at best. And it's not only the cost but capcity of IT departments. If porting software was easy and cheap, no one would be using and paying for Oracle at this point.
 

blckgrffn

Diamond Member
May 1, 2003
7,344
628
126
www.teamjuchems.com
Finally - an explanation for why Oracle makes money.
This is true of basically all ERP systems or other big data systems. I left one a few years back and they were still installing "new" Itanium systems to support software maintenance contracts well into this decade. Lot's of Power based clusters as well with full AIX development ongoing as well.

Enterprise software is probably also going to one of the slowest ships to "change course" to anything new, so this a is a hugely profitable, very long term market.

@beginner99 has it really right - eyes on the prize (margins) and with the rise of cloud computing crazy scale as well.
 
Last edited:
  • Like
Reactions: Tlh97 and Gideon

Eug

Lifer
Mar 11, 2000
23,015
536
126
Despite the dev kit agreement stating benchmarketing is not allowed, Rosetta 2 benchmarks are now up.

VirtualApple A12Z 2.4 GHz

~840 single core
~2950 multi-core

Note that this only utilizes four cores. Also, the clock speed is slightly lower.

iPad Pro A12Z 2.49 GHz

~1125 single core
~4700 multi-core

However, this is utilizing all 8 cores.

If we just compare single core speed, and ignore the 3.75% clock speed discrepancy, then the Rosetta 2 speed is about 75% of native. I would assume this represents translation at install.

BTW, coincidentally, the single core speed here is basically identical to the single core speed of my A10X iPad Pro from 2017, and multi-core of the dev kit under Rosetta is much faster than my iPad Pro. My iPad Pro is my daily mobile driver, and I have no qualms about its speed for general usage. The times when I've noticed it slow down is when doing video editing.
 
Last edited:

gdansk

Senior member
Feb 8, 2011
549
227
116
Despite the dev kit agreement stating benchmarketing is not allowed, Rosetta 2 benchmarks are now up.

VirtualApple A12Z 2.4 GHz

~840 single core
~2950 multi-core

Note that this only utilizes four cores. Also, the clock speed is slightly lower.

iPad Pro A12Z 2.49 GHz

~1125 single core
~4700 multi-core

However, this is utilizing all 8 cores.

If we just compare single core speed, and ignore the 3.75% clock speed discrepancy, then the Rosetta 2 speed is about 75% of native. I would assume this represents translation at install.

BTW, coincidentally, the single core speed here is basically identical to the single core speed of my A10X iPad Pro from 2017, and multi-core of the dev kit under Rosetta is much faster than my iPad Pro. My iPad Pro is my daily mobile driver, and I have no qualms about its speed for general usage. The times when I've noticed it slow down is when doing video editing.
If you look at the JSON for the two tests, the clock speed is actually basically the same:
1593451099272.png1593451120214.png
 

Eug

Lifer
Mar 11, 2000
23,015
536
126
Take this as you will:


Some tests we did on the Apple Transition kit with Apple Silicone.

Encoding the free animation movie bbb_sunflower_1080p_30fps to a MP4 format with HEVC and AAC with a video bitrate of 6000k and audio bitrate of 256k
Encoding with FFmpeg 4.3.1 compiled versions for ARM and Intel (with x265 library)

Both system running Mac Big Sur Beta 5

Software encoding for video and audio:
Intel Core i9 2.3Ghz 8core 5:09
Intel Core i3 2.8Ghz. 23:18
Apple Silicone A12Z 36:43

Hardware encoding with Apple Videotoolbox (*) for video and audio:
Intel Core i9 2:58
Apple Silicone A12Z 10:51
Intel Core i3 N/A (does not have HEVC hardware accelerated encoding)

Although the Apple transition kit is not using the most powerful ARM atm it does clearly indicate they are much slower in doing hard crunching numbers. With the A14 it may be a bit closer to Intel Core i3
The hardware accelerated encoding was much slower too (which was a surprise).
Basically a Core i5, i7 and i9 will be much faster for the foreseeable future.

A side note: the Apple kit got incredible warm and it may have did some throttling along the test.
 
  • Like
Reactions: Tarkin77

soresu

Golden Member
Dec 19, 2014
1,569
780
136
The hardware accelerated encoding was much slower too (which was a surprise).
Not a great surprise - this feature has been optimised a heck of a lot on Intel CPU's to keep them competitive with AMD and nVidia hardware encoding solutions.

Not just the encoder itself, but also the decoders that feed the video in and likely the copy process between the 2 to minimise latency and power usage in hardware mode.

I don't think you can really say that Apple were actually competing with Qualcomm on this particular feature for speed so much as quality.

When you are constantly competing with multiple Android vendors camera sensor/image processing solutions it is more important to make sure that the actual encoded video is of high quality per bit than encoded particularly fast - though as this use case becomes a priority they will likely switch gears there.
 
  • Like
Reactions: Eug

wlee15

Senior member
Jan 7, 2009
309
14
81
Take this as you will:


Some tests we did on the Apple Transition kit with Apple Silicone.

Encoding the free animation movie bbb_sunflower_1080p_30fps to a MP4 format with HEVC and AAC with a video bitrate of 6000k and audio bitrate of 256k
Encoding with FFmpeg 4.3.1 compiled versions for ARM and Intel (with x265 library)

Both system running Mac Big Sur Beta 5

Software encoding for video and audio:
Intel Core i9 2.3Ghz 8core 5:09
Intel Core i3 2.8Ghz. 23:18
Apple Silicone A12Z 36:43

Hardware encoding with Apple Videotoolbox (*) for video and audio:
Intel Core i9 2:58
Apple Silicone A12Z 10:51
Intel Core i3 N/A (does not have HEVC hardware accelerated encoding)

Although the Apple transition kit is not using the most powerful ARM atm it does clearly indicate they are much slower in doing hard crunching numbers. With the A14 it may be a bit closer to Intel Core i3
The hardware accelerated encoding was much slower too (which was a surprise).
Basically a Core i5, i7 and i9 will be much faster for the foreseeable future.

A side note: the Apple kit got incredible warm and it may have did some throttling along the test.
Not surprising at all as x264 and x265 have a ton of assembly code for x86.
 
  • Like
Reactions: Eug

Thala

Golden Member
Nov 12, 2014
1,135
446
136
Take this as you will:


Some tests we did on the Apple Transition kit with Apple Silicone.

Encoding the free animation movie bbb_sunflower_1080p_30fps to a MP4 format with HEVC and AAC with a video bitrate of 6000k and audio bitrate of 256k
Encoding with FFmpeg 4.3.1 compiled versions for ARM and Intel (with x265 library)

Both system running Mac Big Sur Beta 5

Software encoding for video and audio:
Intel Core i9 2.3Ghz 8core 5:09
Intel Core i3 2.8Ghz. 23:18
Apple Silicone A12Z 36:43

Hardware encoding with Apple Videotoolbox (*) for video and audio:
Intel Core i9 2:58
Apple Silicone A12Z 10:51
Intel Core i3 N/A (does not have HEVC hardware accelerated encoding)

Although the Apple transition kit is not using the most powerful ARM atm it does clearly indicate they are much slower in doing hard crunching numbers. With the A14 it may be a bit closer to Intel Core i3
The hardware accelerated encoding was much slower too (which was a surprise).
Basically a Core i5, i7 and i9 will be much faster for the foreseeable future.

A side note: the Apple kit got incredible warm and it may have did some throttling along the test.
The x265 library is currently lacking ARM SIMD optimizations. So you better use the X264 library, which is relatively good optimized for ARM64 or wait until X265 is also optimized - it is currently being worked on.
 
  • Like
Reactions: Eug and teejee

Doug S

Senior member
Feb 8, 2020
466
674
96
The x265 library is currently lacking ARM SIMD optimizations. So you better use the X264 library, which is relatively good optimized for ARM64 or wait until X265 is also optimized - it is currently being worked on.
That's the problem with people picking a single benchmark that exercises the CPU in only one way. Despite its limitations, at least Geekbench does a couple dozen different tests and won't suffer this type of issue with a particular library being optimized for one CPU over another.

It is also testing a two year old SoC, and not whatever the actual ARM Macs will end up with.
 

ASK THE COMMUNITY