All iPads 4k encoding speed: Ipad Air 3 vs. 10.5 pro or 11 pro?

[Samut]

Are there any benchmarks to be found how the new Ipad Air 3 would stack up against Ipad pro 10.5 or Ipad pro 11 in 4k encoding?

I am considering on purchasing either the old pro or the new Air but havent been able to find benchmarks on this.

As a sidenote: is Lumafusion good enough for casual video editing and grading on the go?

 

[EugW]

Are there any benchmarks to be found how the new Ipad Air 3 would stack up against Ipad pro 10.5 or Ipd pro 11 in 4k encoding?

I am considering on purchasing either the old pro or the new Air but havent been able to find benchmarks on this.

Software encoding would be faster on the 11” of course but for hardware encoding it may not necessarily be significantly faster. I recall from the various reviews I’ve seen the 11” is usually moderately faster to over twice as fast as the Pro for various encoding tasks (although the tests were not necessarily 4K), but very occasionally it’s about the same. YMMV.

The Air would be at best only mildly faster than the 10.5” Pro.

IOW, if you really need performance, you should get the 11”.

 

[tps3443]

.

 

[tps3443]

Are there any benchmarks to be found how the new Ipad Air 3 would stack up against Ipad pro 10.5 or Ipad pro 11 in 4k encoding?

I am considering on purchasing either the old pro or the new Air but havent been able to find benchmarks on this.

As a sidenote: is Lumafusion good enough for casual video editing and grading on the go?

Ok, if you seriously want to have a smooth 4K editing experience then I would just jump straight in for a 11” 1TB iPad Pro with the 6GB of ram, if you can manage it. I kinda learned the hard way, I originally had a iPad Pro 11” 64GB with 4GB of ram. Well, the 4GB of ram goes quick when editing 4K video, performance goes down hill, and your timeline preview begins to sputter along, as iOS is starving for memory.

What happens is, you open up your editing software, and your getting in to a groove of the direction your going to go with your video, and as you work for about 10-20 minutes, cutting, deleting, adding more files, and effects it uses more and more and more ram.

It is manageable, but the stuttering is annoying, and it hinders what your trying to do.

I put a 512GB iPad Pro 11” 4GB ram, right beside a 1TB iPad Pro 11” with the 6GB of ram, specifically to only test ram usage in various applications over the period of a week. And everything was about the same, accept for editing 4K video. It is the sad truth, but you need more ram for a smooth experience. My experience so far with my iPad has been that I actually forget I’m encoding on a iPad lol.

(The iPad Pro 11” 512gb/256gb/64GB) only has 3.69GB of ram before it is even turned on. And once iOS is on and running you’ve got around 2.8GB with nothing even open. After using Lumafusion, iMovie, Adobe Rush CC these applications start using up all of the free ram, then they work on eating your inactive ram next. So, any quick jumping around in the applications timeline causes stuttering that’s pretty annoying at times. These are some impressive applications, with a lot of features like desktop software would have. So, unless more optimization can be done; the only way around this for seriously smooth editing is editing at a lower resolution..A laptop, or desktop lol, Or with a 1TB iPad Pro with 6GB of ram, you have 5.6GB of useable ram, and it’s always butter smooth during heavy encoding.

 

[EugW]

Ok, if you seriously want to have a smooth 4K editing experience then I would just jump straight in for a 11” 1TB iPad Pro with the 6GB of ram, if you can manage it. I kinda learned the hard way, I originally had a iPad Pro 11” 64GB with 4GB of ram. Well, the 4GB of ram goes quick when editing 4K video, performance goes down hill, and your timeline preview begins to sputter along, as iOS is starving for memory.

What happens is, you open up your editing software, and your getting in to a groove of the direction your going to go with your video, and as you work for about 10-20 minutes, cutting, deleting, adding more files, and effects it uses more and more and more ram.

It is manageable, but the stuttering is annoying, and it hinders what your trying to do.

I put a 512GB iPad Pro 11” 4GB ram, right beside a 1TB iPad Pro 11” with the 6GB of ram, specifically to only test ram usage in various applications over the period of a week. And everything was about the same, accept for editing 4K video. It is the sad truth, but you need more ram for a smooth experience. My experience so far with my iPad has been that I actually forget I’m encoding on a iPad lol.

(The iPad Pro 11” 512gb/256gb/64GB) only has 3.69GB of ram before it is even turned on. And once iOS is on and running you’ve got around 2.8GB with nothing even open. After using Lumafusion, iMovie, Adobe Rush CC these applications start using up all of the free ram, then they work on eating your inactive ram next. So, any quick jumping around in the applications timeline causes stuttering that’s pretty annoying at times. These are some impressive applications, with a lot of features like desktop software would have. So, unless more optimization can be done; the only way around this for seriously smooth editing is editing at a lower resolution..A laptop, or desktop lol, Or with a 1TB iPad Pro with 6GB of ram, you have 5.6GB of useable ram, and it’s always butter smooth during heavy encoding.

My next iPad will have a MINIMUM of 6 GB. Luckily I won't be buying for a few years, unless my current 4 GB RAM iPad Pro breaks.

It's interesting that Apple in 2019 now considers 3 GB to be the baseline. 3 GB is OK for 2019, but I suspect the baseline in say 2022 will be 4 GB, with the Pros starting at 6 GB, and the highest end Pros at 8 GB. I mention 2022 because I figure there will be new iPad Pros in 2020, and then yet another generation of iPad Pros in 2022, which will be about a good time for me to upgrade from my existing 2017 iPad Pro.

 

[Samut]

Big thanks for the comprehensive replies. It was extremely interesting to hear actual experiences with the 4gb vs. 6gb models.

The memory issue was the reason why I initially decided not to upgrade to 11 pro even though many people on the forums said it would be enough.

The primary reason I would be upgrading is that I currently use my Ipad Air 2 to edit raw photos on the go. Even though it works it is starting to show its age. Now that I have been lately slowly learning on how to shoot videos I would much prefer being able to edit them on the go. I travel quite a bit and never have enough time for that at home and had an Idea I could do also that on Ipad.

As far as I understood from the comments I might be better off with purchasing a cheaper Ipad Pro 10.5 with more memory and keep it for the next couple of years until Apple comes up with a pro with more memory as I would not be happy to pay for the 1tb version.

Edit: one of the options would be to get the Air 3 and use it meanwhile I replace it with the next pro (if it has more ram) as my parents will be needing a Ipad soon. Just worried if the 3gb of ram will be enough even for smoother Lightroom.

 

[KimHansenDK]

iPad Pro 3rd Gen with 4GB ram is perfectly fine for editing 4K video.

 

[Act3]

Not sure , but the Faster GPU in the 10.5 ( compared to air ) may help with editing.

 

[tps3443]

Big thanks for the comprehensive replies. It was extremely interesting to hear actual experiences with the 4gb vs. 6gb models.

The memory issue was the reason why I initially decided not to upgrade to 11 pro even though many people on the forums said it would be enough.

The primary reason I would be upgrading is that I currently use my Ipad Air 2 to edit raw photos on the go. Even though it works it is starting to show its age. Now that I have been lately slowly learning on how to shoot videos I would much prefer being able to edit them on the go. I travel quite a bit and never have enough time for that at home and had an Idea I could do also that on Ipad.

As far as I understood from the comments I might be better off with purchasing a cheaper Ipad Pro 10.5 with more memory and keep it for the next couple of years until Apple comes up with a pro with more memory as I would not be happy to pay for the 1tb version.

Edit: one of the options would be to get the Air 3 and use it meanwhile I replace it with the next pro (if it has more ram) as my parents will be needing a Ipad soon. Just worried if the 3gb of ram will be enough even for smoother Lightroom.

Yep, I’d honestly find yourself a good deal on a 10.5” iPad Pro 2nd gen. Sure, the CPU, and GPU may not be as fast as a iPad Pro 11” but it has 4GB of ram. One of the top user complaints for using iMovie is “lagging performance” the fix for this is closing unnecessary applications that are open, and or the device does not have enough memory to run the application smoothly, and to try using your application slower by doing only 1 thing at a time lol.

 

[EugW]

Yep, I’d honestly find yourself a good deal on a 10.5” iPad Pro 2nd gen. Sure, the CPU, and GPU may not be as fast as a iPad Pro 11” but it has 4GB of ram. One of the top user complaints for using iMovie is “lagging performance” the fix for this is closing unnecessary applications that are open, and or the device does not have enough memory to run the application smoothly, and to try using your application slower by doing only 1 thing at a time lol.

Yes, one of the reasons —- beside the cost of course —- that I didn’t buy the 11” was because it also only has 4 GB RAM. I figured that it wasn’t a good way to future proof a tablet purchase by getting a faster SoC and USB-C without more RAM to match.

If it had been A12X + USB-C and 6 GB RAM, I’d probably be typing on an 11” iPad right now.

I’m happy to wait until 2022 to get a new machine.

 

[tps3443]

Yes, one of the reasons —- beside the cost of course —- that I didn’t buy the 11” was because it also only has 4 GB RAM. I figured that it wasn’t a good way to future proof a tablet purchase by getting a faster SoC and USB-C without more RAM to match.

If it had been A12X + USB-C and 6 GB RAM, I’d probably be typing on an 11” iPad right now.

I’m happy to wait until 2022 to get a new machine.

The funny thing is, if the 4GB ram models just had a tiny bit more ram. Like possibly having the full 4GB available. There we never be a problem.

 

[EugW]

Software encoding would be faster on the 11” of course but for hardware encoding it may not necessarily be significantly faster. I recall from the various reviews I’ve seen the 11” is usually moderately faster to over twice as fast as the Pro for various encoding tasks (although the tests were not necessarily 4K), but very occasionally it’s about the same. YMMV.

Here's zollotech comparing the the iPad 9.7" 2018 vs. iPad Pro 10.5" 2017 vs. iPad Pro 12.9" 2018:

Video export test is at 16:08. All are exported as 4K from iMovie (although he provides few details)

9:12: A10 iPad 9.7" 2018
7:40: A10X iPad Pro 10.5" 2017
7:34: A12X iPad Pro 12.9" 2018

I'm rather impressed by the iPad Pro 2017 here, as it's basically the same as the iPad Pro 2019. I guess it must be a mainly hardware based export.

Given these results, with this particular test I wonder how well the A12 iPad Air 10.5" 2019 would do.

---

LumaFusion test at 4:22, for a 1080p h.264 export. Surprisingly, the 10.5" Pro beat the 11" Pro, albeit just barely.

Furthermore, in iMovie, the 10.5" Pro beat the 11" Pro again, although again it was just barely.

OTOH, I've seen a test in Adobe Rush where the 11" Pro was twice as fast as the 10.5" Pro.

 

[AutomaticApple]

.

Lonely period?

 

[smbu2000]

Lonely period?

Double post?

 

[tps3443]

Here's zollotech comparing the the iPad 9.7" 2018 vs. iPad Pro 10.5" 2017 vs. iPad Pro 12.9" 2018:

Video export test is at 16:08. All are exported as 4K from iMovie (although he provides few details)

View attachment 829500

9:12: A10 iPad 9.7" 2018
7:40: A10X iPad Pro 10.5" 2017
7:34: A12X iPad Pro 12.9" 2018

I'm rather impressed by the iPad Pro 2017 here, as it's basically the same as the iPad Pro 2019. I guess it must be a mainly hardware based export.

Given these results, with this particular test I wonder how well the A12 iPad Air 10.5" 2019 would do.

---

LumaFusion test at 4:22, for a 1080p h.264 export. Surprisingly, the 10.5" Pro beat the 11" Pro, albeit just barely.

Furthermore, in iMovie, the 10.5" Pro beat the 11" Pro again, although again it was just barely.

OTOH, I've seen a test in Adobe Rush where the 11" Pro was twice as fast as the 10.5" Pro.

In applications like this, there is a massively noticeable difference in performance. The A12X can literally encode a video over twice as fast sometimes, and usually nearly double. All of those videos were before the iPad even launched, or right around that time. Minimal optimization for the A12X in applications back in November.


This is a more up to date comparison.

4K 1 minute, video encoded 11” vs 10.5”


physical and chemical properties of ammonia

 

[EugW]

In applications like this, there is a massively noticeable difference in performance. The A12X can literally encode a video over twice as fast sometimes, and usually nearly double. All of those videos were before the iPad even launched, or right around that time. Minimal optimization for the A12X in applications back in November.


This is a more up to date comparison.

4K 1 minute, video encoded 11” vs 10.5”


physical and chemical properties of ammonia

I’m not convinced it’s due to optimization. I suspect in these situations where the 11” did twice as well it was due to software encoding.

In contrast, if it’s hardware encoding, even a “slow” SoC can do it very fast. That’s similar to why I can decode a demo 4K p60 10-bit HEVC file from Sony on my 2017 Core m3 12” MacBook with just 25% CPU usage, whereas a 2015 Core i7-6700K iMac cannot do it even at 100% CPU utilization with the fans blaring at maximum. Obviously, the top-of-the-line 2015 iMac is usually way, way, way faster than my entry-level 2017 MacBook, but for certain actions, particularly related to video, it isn’t.

My point earlier was that perhaps some of these applications depend heavily on hardware encoding so at least with those settings in those apps, it’s quite possible the 11” wouldn’t offer significantly faster rendering times. But if it’s pure software encoding, one should expect the 11” to be much faster.

For this reason, I waited until 2017 to buy my MacBook and iMac, and partially also for this reason, I was happy to upgrade to an older iPad Pro as long as it had at least A10X. Note I say A10X and not A10 just because A10 is too close to A9 and A8X in general CPU performance, and I already had an A8X iPad Air 2. The A10X provides a much bigger jump in general CPU performance, along with the more complete hardware video decoding/encoding support.

P.S. You have a link about ammonia in your post.

 

[tps3443]

P.S. You have a link about ammonia in your post.

The website used to create IMG links adds in there own random advertising, anytime I post a photo lol.
[doublepost=1554001662][/doublepost]2 more cores, 35% faster ipc, double cache size, and a faster base frequency.

So your saying that all of this results in a slower cpu when using its hardware encoding? I don’t know about that.

I wish I had a iPad Pro 2nd gen, I’d love to test them head to head.

Ive ran a comparison on my iPad 6th gen and the difference was more like 18 minutes vs 3 minutes. Not a minimal difference as in some of the videos above.

How do we know if Adobe Rush, Lumafusion, or iMovie is using software encoding, or just raw hardware encoding? When running on one device to the next.

 

[plucky duck]

Does both the iPad Pro 11" and 12.9" 1TB models have 6gb ram, or just the larger 12.9" model? Top of the line 12.9" 1TB w/cellular comes to a whopping $2400, damn!

 

[EugW]

So your saying that all of this results in a slower cpu when using its hardware encoding? I don’t know about that.

No, I’m saying hardware encoding is very fast and getting a faster CPU doesn’t necessarily help as the the design of the hardware encoder block might be similar.

Software decoding is much slower but is dependent upon on how fast the CPU is.

eg. Slow CPU might software encode in 4 minutes whereas fast CPU can do it in 2 minutes, but a hardware encoder might do it in 1 minute (even on that first slow CPU).

Apple’s SoCs have had hardware h.264 encoders for a very long time now and h.265 8-bit hardware encoders since A10. (Actually earlier, but they didn't allow anyone to use it.) There is no h.265 10-bit hardware encoder though so if you had to encode that then it would be all software based and slow.

For example, I found this benchmark that compares a number of Intel and AMD CPUs with and without hardware encoding, in Adobe Premiere.

Of interest is the inclusion of the i9-7980XE and the i7-8700K. The i9 has 18 cores (!) at 2.6 GHz with Turbo Boost to 4.4 GHz. The i7 has 6 cores at 3.7 GHz with Turbo Boost to 4.7 GHz. One key difference though is that the i9 has no hardware video encoder at all, whereas the i7 has a robust hardware video encoder. In general CPU usage, the i9 is much, much faster than the i7... but the same is not necessarily true with video encoding:

i9-7980XE software encoder: 30 minutes
i7-8700K hardware encoder: 33 minutes
i7-8700K software encoder: 46 minutes

The other benefit is that hardware decoding and encoding uses much, much less power, so better battery life.

BTW, this picture illustrates why I waited until 2017 to update my MacBook and iMac. While I didn't do it for VP9, it applies to various aspects of h.265 HEVC as well.

Software decoding with this chip uses 6 Watts of CPU+GPU power with 70-90% CPU utilization. Hardware decoding on the chip that replaces it uses only 0.8 Watts of CPU+GPU power with 5-25% CPU utilization, even though general CPU performance of these two chips is otherwise relatively similar.

I mentioned before that my 2017 Core m3 MacBook (Geekbench score 7000) can decode one example 4K 10-bit HEVC video with 25% CPU usage (hardware decode), whereas a 2015 Core i7-6700K (Geekbench score 18000) can't do it cleanly even with 100% CPU usage (software decode). IOW, with this very specific test, my Core m3 MacBook beats all 2015 iMacs and all 2016 MacBook Pros, because my little MacBook has a hardware decoder to handle the job whereas all the pre-2017 models do not.

 

[tps3443]

No, I’m saying hardware encoding is very fast and getting a faster CPU doesn’t necessarily help as the the design of the hardware encoder block might be similar.

Software decoding is much slower but is dependent upon on how fast the CPU is.

eg. Slow CPU might software encode in 4 minutes whereas fast CPU can do it in 2 minutes, but a hardware encoder might do it in 1 minute (even on that first slow CPU).

Apple’s SoCs have had hardware h.264 encoders for a very long time now and h.265 8-bit hardware encoders since A10. (Actually earlier, but they didn't allow anyone to use it.) There is no h.265 10-bit hardware encoder though so if you had to encode that then it would be all software based and slow.

For example, I found this benchmark that compares a number of Intel and AMD CPUs with and without hardware encoding, in Adobe Premiere.

Of interest is the inclusion of the i9-7980XE and the i7-8700K. The i9 has 18 cores (!) at 2.6 GHz with Turbo Boost to 4.4 GHz. The i7 has 6 cores at 3.7 GHz with Turbo Boost to 4.7 GHz. One key difference though is that the i9 has no hardware video encoder at all, whereas the i7 has a robust hardware video encoder. In general CPU usage, the i9 is much, much faster than the i7... but the same is not necessarily true with video encoding:

i9-7980XE software encoder: 30 minutes
i7-8700K hardware encoder: 33 minutes
i7-8700K software encoder: 46 minutes

The other benefit is that hardware decoding and encoding uses much, much less power, so better battery life.

BTW, this picture illustrates why I waited until 2017 to update my MacBook and iMac. While I didn't do it for VP9, it applies to various aspects of h.265 HEVC as well.

Software decoding with this chip uses 6 Watts of CPU+GPU power with 70-90% CPU utilization. Hardware decoding on the chip that replaces it uses only 0.8 Watts of CPU+GPU power with 5-25% CPU utilization, even though general CPU performance of these two chips is otherwise relatively similar.

I mentioned before that my 2017 Core m3 MacBook (Geekbench score 7000) can decode one example 4K 10-bit HEVC video with 25% CPU usage (hardware decode), whereas a 2015 Core i7-6700K (Geekbench score 18000) can't do it cleanly even with 100% CPU usage (software decode). IOW, with this very specific test, my Core m3 MacBook beats all 2015 iMacs and all 2016 MacBook Pros, because my little MacBook has a hardware decoder to handle the job whereas all the pre-2017 models do not.

Well that’s amazing how that 8700k could almost keep up with that 7980xe using its built in hardware encoder. Very nice write up.

Any software that can take advantage of 18 core and 36 threads, and those 7980XE’s will run at a pretty low all core boost on all cores. So it looks like clock speed is king for the 8700K even though it only has 12 threads. I imagine anyone who actually owns a 7980XE is running it heavily overclocked on all cores, and that 30 minute time frame while using a software encoder would probably be cut in half or more. Although, your pc might be drawing like 800-1000watts at just the cpu socket lol.

Anyways, the cpu market has been turned on its head. It is amazing where we’re headed with desktop processors since Ryzen came out.

So most software we are using on iOS is using a hardware encoder, and there’s the same support for that across both platforms from 2nd gen to 3rd gen? But if both are using software encoding, obviously the 3rd gen iPad Pro would prevail when using only cpu power.

 

[EugW]

Well that’s amazing how that 8700k could almost keep up with that 7980xe using its built in hardware encoder. Very nice write up.

Any software that can take advantage of 18 core and 36 threads, and those 7980XE’s will run at a pretty low all core boost on all cores. So it looks like clock speed is king for the 8700K even though it only has 12 threads. I imagine anyone who actually owns a 7980XE is running it heavily overclocked on all cores, and that 30 minute time frame while using a software encoder would probably be cut in half or more. Although, your pc might be drawing like 800-1000watts at just the cpu socket lol.

Anyways, the cpu market has been turned on its head. It is amazing where we’re headed with desktop processors since Ryzen came out.

I am not sure you completely understand. I’m no expert either but here goes: The hardware encoder is a very specific purpose built design, effectively separate from the rest of the chip. And it’s not really directly dependent on the clock speed of the main chip either. Theoretically you can have a 2.5 GHz CPU and a 4 GHz CPU with similar hardware encoders that perform relatively similarly.

My whole point in this discussion is that I think some of the video encoder benchmarks out there are testing different things, so the performance results are going to be very dependent upon what software is used and what specific settings are used too.

 

[tps3443]

I am not sure you completely understand. I’m no expert either but here goes: The hardware encoder is a very specific purpose built design, effectively separate from the rest of the chip. And it’s not really directly dependent on the clock speed of the main chip either. Theoretically you can have a 2.5 GHz CPU and a 4 GHz CPU with similar hardware encoders that perform relatively similarly.

My whole point in this discussion is that I think some of the video encoder benchmarks out there are testing different things, so the performance results are going to be very dependent upon what software is used and what specific settings are used too.

I understood what you were saying. But, the 7980XE was on software encoding so it is just raw cpu power. Yes, the 8700K was nearly just as fast using its hardware encoder, showing impressive results especially being an absolutely measly cpu in comparison to a 7980XE.

I get exactly what your saying.

It’s like comparing DX11 benchmarks to DX12 in the same games, or Vulkan or something like that. Performance variables.

 

[tps3443]

What are your thought on this.

How about when your not encoding the actual file, and you’re just editing and skimming through say; a very long timeline. Selecting endpoints, startpoints, and removing large bits and sections of 4K footage.

This is more cpu and gpu intensive, and both iPads are utilizing just raw power to run the software smoothly.

And the hardware encoder would only be useful when actually encoding the video at the end, which in that case, finish, encode, walk away. But, I spend most of my time actually editing the video for hours sometimes.

 

[EugW]

What are your thought on this.

How about when your not encoding the actual file, and you’re just editing and skimming through say; a very long timeline. Selecting endpoints, startpoints, and removing large bits and sections of 4K footage.

This is more cpu and gpu intensive, and both iPads are utilizing just raw power to run the software smoothly.

And the hardware encoder would only be useful when actually encoding the video at the end, which in that case, finish, encode, walk away. But, I spend most of my time actually editing the video for hours sometimes.

For sure. I don't do much video editing, but I would imagine the 11" would be significantly smoother for complex stuff. I do note some of the reviews said that that for video editing the interface is less prone to stutters, etc. on the 11" as opposed to the 10.5".

I do also note that looking at the 10.5" and 11" side by side can highlight the differences. For OS actions and surfing, the 10.5" seems near instantaneous for most stuff. OTOH, the 11" IS instantaneous. Both are excellent, but the 11" is even more excellent, if that makes sense.

Furthermore, the 11" has a wider screen in landscape mode. It's actually fairly noticeable.

 

[Samut]

Interesting information about the hardware and software encoding. I recall reading that hardware encoding might have worse image quality compared to software encoding why the later would be preferred. Dont know if there is any scientific background to back that up though.

I ended up buying the new Air regardless of the lesser ram as there were no good local deals on 10.5 pro and I believe the A12 might be supported for a longer period. I was still positively surprised how well it works with Lumafusion when editing 4k video.