Apple A13: Launch 2019-09-10 - Geekbench 4 & 5 scores available (GB4 multi-core 14070)

[Roland00Address]

Apple seems to skip A##x generations? Like there was no A11x, I think?

It probably has to do with the foundry generations. A12 and A13 are on the same 7nm [though TSMC does slight improvements.] Thus Apple can do a new chip generation for phones every year while the foundry generations take 2 to 3 years to happen.

In theory Apple could do new ipad A#x chips but it is a question is that really necessary for the iPad. Remember iPhone sales are 200+ million per year. The iPad sell roughly 40 million units per year [somewhere between 9 to 13 million per quarter] but remember most of those iPads are going to be using not the A#x cpus which you find in the iPad pros but instead the A# cpus either the most recent iPhone chip or a generation or two behind. For example the 6th gen 9.7" $329 ipad and the 7th gen 10.2" $329 ipad use the A10. Well the A10 is on TSMC 16nm foundries not the more recent 7nm foundries the A12 and A13 use.

 

[Glo.]

Errr... what voltage do you expect 50 amps to be delivered with??!!

Most extension sockets are only rated to 13 Amps max fuse.

Yeah, like Core i9-9900K is using 180W of power on 1.3v and thanks to 13 Amps delivered to CPU?

There is a very good reason why AM4 sockets has 200 Amp limits.

 

[soresu]

There is a very good reason why AM4 sockets has 200 Amp limits.

Where did this figure come from, is there an open design spec for AM4 in PDF form?

 

[Eug]

OK, that multi-score is better:

iPhone12,3 Geekbench 4 - 5472 / 13769

iPhone12,3 - Geekbench

Benchmark results for an iPhone12,3 with an ARM processor.

browser.geekbench.com

iPhone12,1 Geekbench 5 - 1307 / 2720

iPhone12,1 - Geekbench

Benchmark results for an iPhone12,1 with an ARM processor.

browser.geekbench.com

iPhone12,3 Geekbench 5 - 1324 / 3394

iPhone12,3 - Geekbench

Benchmark results for an iPhone12,3 with an ARM processor.

browser.geekbench.com

All have 4 GB RAM. AFAIK, iPhone12,1 is iPhone 11, and iPhone12,3 is iPhone 11 Pro.

 

[thunng8]

That single core score of 1324 on Geekbench 5 is insane. Almost the same as the top end Ryen 7 processor or i9.

 

[DrMrLordX]

It has the same ST score as my 3900x @ 4425 MHz. MT not so much.

 

[soresu]

Not sure how much Geekbench takes vectors into account either.

Even Zen 2 has 256 bit SIMD, and Intel has 512 bit SIMD types - whereas any non SVE ARM core is still stuck at 128 bit SIMD for now, though it wouldn't surprise me if Apple was the first to implement the new SVE2 as standard in a future core.

 

[Mopetar]

The limiting factor of the iPads and iPhones in the older models is the amount of ram that Apple is putting in the devices. That and the battery. Yes these new Apple SoCs are incredible but we are now reaching the point of diminishing returns.

I think that may have been why Apple was touting the improved efficiency of the SoC this time around more so than the raw power. Granted, unless you're doing something that really uses the full power of the chip, the screen is going to use more power, but coupled with improvements there, Apple was able to get another 4-5 hours of battery life for these models and I suspect that's a bigger improvement than almost anything else they could have done.

Assuming their figures are good, this should also mean a much longer lasting device. Batteries wear down as they go through an increased number of charge cycles, so extending the battery life means fewer charge cycles over the same period of time and a battery with a longer life ahead of it.

 

[Eug]

iPhone12,5 Geekbenchmarketing score up now too. Furthermore, a repeat iPhone12,1 (iPhone 11 N104) score is also up now, showing similar performance to its more expensive iPhone12,3 (iPhone 11 Pro D42) and iPhone12,5 (iPhone 11 Pro Max D43) cousins. BTW, all three models have 4 GB.

 

[Ajay]

I’d take the better perf/watt over any significant performance increase at this point. My iPhone 6S seems to run perfectly fine for what I do, but if I’m traveling using LTE I can be in low power mode (useless) in 4 hours. Make the damn phone 2mm thicker please and put a bigger battery in. I may switch to a Max version on my next upgrade, though since I carry in my front pocket, jeans are a problem.

 

[Eug]

I’d take the better perf/watt over any significant performance increase at this point. My iPhone 6S seems to run perfectly fine for what I do, but if I’m traveling using LTE I can be in low power mode (useless) in 4 hours. Make the damn phone 2mm thicker please and put a bigger battery in. I may switch to a Max version on my next upgrade, though since I carry in my front pocket, jeans are a problem.

They took out 3D Touch which some people estimated added 0.5 mm to the phone thickness. It seems they didn't let that extra space go to waste, since this year, all the batteries are higher capacity.

At this point I'm just wondering if we'll get an A13X, and if so, if it will be sooner rather than later. Some believe there will be no A13X though. In truth, I don't need the extra CPU or GPU performance. I just want to get a 6 GB iPad, and don't want to pay through the nose for the one year-old A12X 1 TB model which already has 6 GB. SoC speed boosts are great and all, but the <1 TB iPad Pros have been at 4 GB for 4 years now, time for a memory upgrade.

In my personal experience, I've been limited by amount of memory much more than CPU speed for device usability and device longevity.

BTW, this chip scores literally twice as fast in GB4 as my Core m3-7Y32, yet I have 16 GB in that laptop. I know you can't compare memory like that directly, but still...

 

[Panino Manino]

"I truly believe Apple chips are silly powerful, but for the last four years, Apple really hasn’t let us touch that power. I shared my rendering experiences again, comparing the iPhone XS against the iPhone SE. In iMove, the iPhone SE continues to render video faster than the XS."

Sorry Apple, iPhones aren’t for pro video! -OR- What I learned arguing with people on Twitter!

The main takeaway, which became resoundingly clear over the last 48 hours? Marketing works. Tuesday we all sat through a round up of new product announcements from Apple, with the stock boiler plat…

somegadgetguy.com

Intriguing.
Why the iPhones are so slow here?

 

[Roland00Address]

I think that may have been why Apple was touting the improved efficiency of the SoC this time around more so than the raw power. Granted, unless you're doing something that really uses the full power of the chip, the screen is going to use more power, but coupled with improvements there, Apple was able to get another 4-5 hours of battery life for these models and I suspect that's a bigger improvement than almost anything else they could have done.

Assuming their figures are good, this should also mean a much longer lasting device. Batteries wear down as they go through an increased number of charge cycles, so extending the battery life means fewer charge cycles over the same period of time and a battery with a longer life ahead of it.

I hear you, but I am also skeptical, even if I believe this is true.

It depends on what workload for many specialized tasks can get specialized DSP hardware (DSP as in Digital Signal Processor, a specialized microprocessor that is only used for a certain tasks and is completely useless if the task is not needed, or the format of the data is not formatted in the right way for the DSP or the DSP is not called by the software.)

Take for example video / movie playback. That is run on a DSP and not the cpu or the gpu of the soc. DSP can be order of magnitude more efficient than any cpu or gpu. But unless the video is formatted in the right way the DSP can't do its job and there is a fallback onto the CPU and GPU. Well video playback is kind of a best case scenario for dedicated hardware for there is only a few different ways to format video for video works on standards and if you do not follow the standards your video does not spread to others and it is unwatchable.

But software itself is dynamic in a way that video is not.

-----

So the question is under what workloads the increase battery life possible? I have no doubt there are battery life improvements in the SoC for the stars aligned of what is possible with the apple phone platform, they control the OS, they kind of control what software, apis, etc that is run adjacent to the OS, they control the hardware at a very specific level, etc. If anyone is able to eek out battery life improvements on a phone platform it is Apple in a way that Android and Android Phone OEM makers can not, and it is even harder for eeking out battery life on Windows with its Laptop and Mobile platforms.

"I truly believe Apple chips are silly powerful, but for the last four years, Apple really hasn’t let us touch that power. I shared my rendering experiences again, comparing the iPhone XS against the iPhone SE. In iMove, the iPhone SE continues to render video faster than the XS."

iPhone SE has less pixels and less points it draws the screen. I wonder if there is any differences on XS vs XR.

The XS Max and XR share the same user interface elements as in "points" accessible to the software developer and user. Think of points as element real estate not in inches but a ratio of your screen. At the same time the XS Max takes those points and renders them 3x ratio while the XR renders at 2x ratio. Technically 3x should be a little harder on the XS Max hardware since they use the same SoC but fixed function hardware DSP may make this difference trival to perform [and the XS Max also has more ram]

The XS non max has more total pixels but has less points available than the XR.

And the iPhone SE has the least amount of points that are used by any current iPhone that Apple is still supporting with OS updates. The iPhone XS non max has 67% more points, and the XR and XS max 104% more points.

This is a nice table [I will link to the website below] that compares the Points vs Pixels in the various iPhones.

The Ultimate Guide To iPhone Resolutions

Articles about PaintCode's new features, interviews, application development topics.

www.paintcodeapp.com

 

[Eug]

"I truly believe Apple chips are silly powerful, but for the last four years, Apple really hasn’t let us touch that power. I shared my rendering experiences again, comparing the iPhone XS against the iPhone SE. In iMove, the iPhone SE continues to render video faster than the XS."

Sorry Apple, iPhones aren’t for pro video! -OR- What I learned arguing with people on Twitter!

The main takeaway, which became resoundingly clear over the last 48 hours? Marketing works. Tuesday we all sat through a round up of new product announcements from Apple, with the stock boiler plat…

somegadgetguy.com

Intriguing.
Why the iPhones are so slow here?

iMovie on iDevices usually encodes in hardware, so the general purpose CPU is less important. Encoding speed would depend on the strength of the hardware encoder and the various settings.

I don’t use iMovie but have dabbled with LumaFusion on my iPad Pro 10.5” with A10X and it was interesting to note that my encodes to h.264 AVC and h.265 HEVC took exactly the same time, despite the fact that HEVC is much more computationally complex. Furthermore, I was checking out some reviews that demonstrated that encoding with the same software and the same settings on my 2nd generation iPad Pro sometimes took exactly the same time as on the 3rd generation iPad Pro with A12X. However, in other reviews with different settings it was faster on the 3rd generation. There have also been rare cases where the iPad Pro 10.5" (A10X) is slightly faster than the iPad Pro 11" (A12X).

So, video encoding on iOS isn’t necessarily a good way to test these iDevices. All iDevices from A13 in the iPhone 11 all the way down to A10 in the IPhone 7 have hardware HEVC encoders. And that's the whole point after all, since if you tried to encode this in software on the general purpose CPU, it would forever.

An important point here though is that the A9 in the iPhone SE does not have a hardware HEVC encoder so iMovie by default would encode In h.264. It’s quite possible iMovie is defaulting to h.264 on the SE (A9) and h.265 on the XS (A12). In that scenario I wouldn’t be completely shocked that the SE hardware encodes h.264 faster than the XS hardware encodes h.265 as it wholly dependent upon the specific encoder. I will tell you that in several reviews, hardware h.265 encodes on the iPad Pro (A10X) is faster than hardware h.265 encodes on the iPad Air (A12).

For example, here at the 302 second mark:

iPad Air 10.5 (A12) on left, iPad Pro 10.5 (A10X) on right

What this doesn't tell you is how well the CPU actually does with editing. I mentioned before that in some very specific tests the encodes on A10X took about the same time as A12X, despite the fact that A12X is a much, much more powerful processor. However, what is interesting to note is that the actual editing process was much smoother on A12X in those same comparison reviews. Layer on multiple 4K video tracks along with a bunch of effects, and the A10X becomes quite laggy during the editing process. In contrast, the A12X stays smooth. It's not memory, because both machines have 4 GB. The key here is editing like that requires CPU power, but final encoding may not need much.

 

[Panino Manino]

@Roland00Address

iPhone SE has less pixels and less points it draws the screen. I wonder if there is any differences on XS vs XR.

I don't think the resolution of the screen affects anything here.
The work was just convert a file.

@Eug

An important point here though is that the A9 in the iPhone SE does not have a hardware HEVC encoder so iMovie by default would encode In h.264. It’s quite possible iMovie is defaulting to h.264 on the SE (A9) and h.265 on the XS (A12).

I checked.
What was being converted was a video from a Panasonic camera, standard MP4 H264. Output was also a video encoded with H264.

His point is that the practical actual observed performance doesn't come nowhere near Apple's claims of absolute undisputed performance lead.
From where those graphs comes from?

How mysterious.

 

[Eug]

@Roland00Address


I don't think the resolution of the screen affects anything here.
The work was just convert a file.

@Eug


I checked.
What was being converted was a video from a Panasonic camera, standard MP4 H264. Output was also a video encoded with H264.

His point is that the practical actual observed performance doesn't come nowhere near Apple's claims of absolute undisputed performance lead.
From where those graphs comes from?

How mysterious.

Not mysterious at all. This is not software encoding. This is hardware encoding. This does not leverage the CPU advantage here. So basically, his whole argument is totally meaningless.

Put it this way, my 1.2 GHz dual-core low power Core m3-7Y32 MacBook encodes HEVC many, many times faster than a 4 GHz quad-core desktop Core i7-4790K iMac. It's no contest, the MacBook wins. In fact, I wouldn't be surprised if this fanless MacBook was 5X as fast or even faster.

How can a little 4.5W chip like this beat that 88 W desktop chip? Because it's not using software rendering. It's all being done in hardware on the MacBook, but it would have to be done in software on the CPU on the iMac.

A9 has a hardware h.264 encoder. So does A12, but it seems for this very specific hardware encode, A9 does better for whatever reason. However, A12 has the added advantage of being able to encode HEVC in hardware. If you tried doing this on A9 via software encoding, it'd literally take hours or for a longer video, maybe even days. In contrast, on A12 via hardware encoding, it takes just minutes.

 

[Eug]

Geekbench 5: 1338 / 3543

 

[Thala]

His point is that the practical actual observed performance doesn't come nowhere near Apple's claims of absolute undisputed performance lead.

Thats because he is making stupid assumptions, namely that the video encoding would be CPU limited. Apples claim is related to (single threaded) CPU performance - but the video export is limited by HW encoding speed and possibly write speed of the storage media.

 

[Panino Manino]

@Eug @Thala
"It's not done in the CPU", so what? The GPU and media blocks also changed and their higher performance should be showing.
An iPhone from 2018 is slower in that work than an iPhone from 2016. A work where brute should should play a role.
He also compares with some Androids. He frequently uses his LG V30 to produce his videos and it's also faster.
No matter what file or what program he uses to test the result is the same, the thing isn't as fast as Apple says.
Why and how?

 

[scannall]

@Eug @Thala
"It's not done in the CPU", so what? The GPU and media blocks also changed and their higher performance should be showing.
An iPhone from 2018 is slower in that work than an iPhone from 2016. A work where brute should should play a role.
He also compares with some Androids. He frequently uses his LG V30 to produce his videos and it's also faster.
No matter what file or what program he uses to test the result is the same, the thing isn't as fast as Apple says.
Why and how?

It would appear anyway they have cut down the hardware encoder somewhat. If so, most likely to give the cores a little more headroom. Encoding video isn't really the primary use of a smartphone anyway.

 

[Ajay]

Is this also a problem on iPad Pros? Seems like they would have a more pressing use-case scenario for video transcoding. Seems like the A12X should have much better hardware for this type of application.

 

[Thala]

@Eug @Thala
"It's not done in the CPU", so what? The GPU and media blocks also changed and their higher performance should be showing.

You wrote:
"His point is that the practical actual observed performance doesn't come nowhere near Apple's claims of absolute undisputed performance lead. "

This is just wrong as you compare totally different metrics.
If you would have said, that you consider the video encoding performance underwhelming compared to older devices i could certainly agree to - but thats not what Apple claimed. They specifically were referring to CPU performance.

 

[Eug]

Is this also a problem on iPad Pros? Seems like they would have a more pressing use-case scenario for video transcoding. Seems like the A12X should have much better hardware for this type of application.

It depends, but hardware video encoding on average for A12X is noticeably faster than A10X, and both are usually faster than A12.

Judging by reviews, performance for hardware video encoding on A12X ranges from very slightly slower than A10X (rarely), about the same as A10X (some of the time), a bit faster than A10X (some of the time), to up to 2.5X as fast as A10X (some of the time). I don’t remember the specific scenarios and settings to account for the differences.

However what impressed me was that on A10X, I could layer a couple of h.264 video tracks plus a 4K 8-bit HEVC video track over each other and add some effects and transitions and only then would scrubbing start to lag. Rendering of this to 4K HEVC was in near real time.

However, according to reviews, on A12X people can often layer multiple videos plus effects and still have a near stutter free experience during editing and scrubbing, probably because of both the higher CPU performance and better video decoding performance. Furthermore, people are saying the experience is smoother in most situations than trying to do the same on a recent Core i7 laptop.

Let’s just say that although people think of GPU gaming when they talk about the Apple X chips, it’s seems they are also using these to maximize hardware video decoding and encoding and editing performance. They effectively are almost purpose-built SoCs in that regard. As such, a subset of say wedding videographers and YouTube bloggers are starting to use A12X almost exclusively for their video editing needs. This should only expand with the release of iPadOS in two weeks, since that finally brings true external USB storage support.

Now we just need Apple to put 6 GB minimum into all the iPad Pros. Right now most are still at 4 GB, with only the $$$$ 1 TB models having 6 GB. The iPad Pro line is overdue for a memory upgrade, as they’ve been at 4 GB since 2015.

 

[Eug]

Antutu scores are showing up now for A13. I suspect these are from embargoed reviewers. The reviews should come out next week, mid-week.

 

[Panino Manino]

but thats not what Apple claimed. They specifically were referring to CPU performance.

Apple shows graphs and brags about both CPU and GPU.
They claim to have unparalleled hardware x software integration and the iPhone ships with iMovie.