First A14 Benchmarks from iPad Air 4 Appear Online

[EugW]

No one is talking about the other geekbench Benchmark.
In Metal compute, the A14 gets a 70% improvement over the A13. It’s even better than the iPad Pro

iPad13,2 - Geekbench

Benchmark results for an iPad13,2 with an ARM processor.

browser.geekbench.com

And it does it with the same number of cores as A13, or half the cores of A12Z.

 

[centauratlas]

“3.66GB of memory “

that seems like a very strange figure

 

[PPietra]

“3.66GB of memory “

that seems like a very strange figure

it is a strange number, but we see the same on the iPhone 11 Pro

 

[mr.steevo]

Personally, I consider 3 GB RAM the new entry level for RAM, meaning us geeks may want more.

<iPhone 6s and SE, iPod Touch and iPad Mini looking over the fence>

“Hey man, we’re still here.”

 

[PPietra]

And it does it with the same number of cores as A13, or half the cores of A12Z.

Metal compute benchmark evaluates the GPU cores mainly, supposedly.

 

[satchmo]

No one is talking about the other geekbench Benchmark.
In Metal compute, the A14 gets a 70% improvement over the A13. It’s even better than the iPad Pro

iPad13,2 - Geekbench

Benchmark results for an iPad13,2 with an ARM processor.

browser.geekbench.com

Can you translate this into a real-world use case? Sample app?

 

[EugW]

Metal compute benchmark evaluates the GPU cores mainly, supposedly.

Yes. I'm talking about GPU cores of course.

(A14 has the same number CPU cores as A13, so my post wouldn't make sense if I was talking about CPU cores.)

<iPhone 6s and SE, iPod Touch and iPad Mini looking over the fence>

“Hey man, we’re still here.”

Those are all quite old now.

I have both a 6s and orig. SE BTW. They work fine in iOS 14, but they're phones, and for me they're just backup phones. Regarding the 3 GB, I was talking more about the iPads, since it's easier to multitask and run out of memory on the iPads.

That said, I have been starting to notice occasional pauses with my 3 GB iPhone 7 Plus. I look forward to a 6 GB iPhone 12 Pro/Max in 2020 and a 6 GB iPad Pro 11" in 2021.

 

[PPietra]

Can you translate this into a real-world use case? Sample app?

It’s the geekbench GPU compute benchmark... It can be "used" in any number of scenarios, from video editing to games, from augmented reality to AI

 

[EugW]

“3.66GB of memory “

that seems like a very strange figure

It's not really. The total is actually 4 GB of physical memory on-board, but not all is visible to Geekbench.

 

[cmaier]

The first turbo switch I remember were on some Samsung PC clones with an 80286. Probably around 1988, it was 6MHz normal speed and 8MHz turbo.

The 8MHz turbo would sometimes cause crashes with certain programs—looking at you MS Flight Simulator 2—so we’d have to switch it to 6 sometimes. I don’t think it could be switched on the fly; iirc, you had to reboot it.

(We ran Digital Research’s GEM OS at the time, a Mac-like GUI that ran on top of MS-DOS. Mostly we used GEM Draw and Xerox’s Ventura Publisher, a surprisingly powerful WYSIWYG desktop publishing program.)

the machine i bought when i started my undergrad was, I believe, a 386 with a turbo button, made by “AST.” I had forgotten about it until you mentioned this. It was nice not to have to use the computer labs, though I recall i had to wire up some sort of connector myself to get it to connect to the school’s mainframe (which i needed maybe twice a semester).

I wish i could remember what happened to that machine. No recollection whatsoever. And, weirdly, I can’t remember how long I had it, or what machine i used in grad school (by that point i was using Unix SPARC s and rs-6000s most of the day, but i must have had some machine on which i wrote my dissertation. It had to have supported a Zip drive, because i still have the drive and the discs, but i cannot remember the machine, and this is now going to drive me crazy. Maybe the same one? Zip drive with rs-232?)

 

[Piggie]

Do we assume that the A14 fitted in the iPhone 12 will have to run slower than the A14 in the iPad.
Simply due to a MUCH larger battery in the iPad and vastly more space for cooling?

 

[cmaier]

Do we assume that the A14 fitted in the iPhone 12 will have to run slower than the A14 in the iPad.
Simply due to a MUCH larger battery in the iPad and vastly more space for cooling?

yes.

 

[EugW]

Do we assume that the A14 fitted in the iPhone 12 will have to run slower than the A14 in the iPad.
Simply due to a MUCH larger battery in the iPad and vastly more space for cooling?

No

 

[PPietra]

Do we assume that the A14 fitted in the iPhone 12 will have to run slower than the A14 in the iPad.
Simply due to a MUCH larger battery in the iPad and vastly more space for cooling?

we should expect it to run a bit slower in multicore. With the A12 it was 12% slower, though with A14 might not be as much if the processor efficiency improved a lot.

 

[EugW]

we should expect it to run a bit slower in multicore. With the A12 it was 12% slower, though with A14 might not be as much if the processor efficiency improved a lot.

iPhone A12 is not 12% slower than iPad A12.

 

[cwwilson]

The Pro still owns this thing but the Air is closing the gap. But as we all know the 2021 Pro is set to be a big upgrade with the new LED displays, better cameras and A14X chips. They will be insane.

At that point the Air will settle in nicely as a fairly feature packed iPad for those who don’t need the extra ‘umph’ from the Pro but want something nicer than the base iPad.

 

[PPietra]

iPhone A12 is not 12% slower than iPad A12.

The iPad Air with A12 has a 12% higher multicore score in geekbench 5 than the iPhone XS with A12

 

[chucker23n1]

iPhone A12 is not 12% slower than iPad A12.

True. 2832 vs. 2497 are actually 13%. The Internet regrets the error.

 

[PPietra]

True. 2832 vs. 2497 are actually 13%. The Internet regrets the error.

😀
well... that actually gives that the iPhone "A12" is 12% slower than iPad "A12": 12% slower, 13% faster 😀

 

[Unregistered 4U]

And it does it with the same number of cores as A13, or half the cores of A12Z.

I think the GPU performance is going to surprise a lot of people.

Do we assume that the A14 fitted in the iPhone 12 will have to run slower than the A14 in the iPad.
Simply due to a MUCH larger battery in the iPad and vastly more space for cooling?

It has been true historically, so you can expect it. Generally not a huge amount, though.

 

[chucker23n1]

😀
well... that actually gives that the iPhone "A12" is 12% slower than iPad "A12": 12% slower, 13% faster 😀

I was agreeing with you. Just saying that it's 13%, not 12%.

 

[EugW]

True. 2832 vs. 2497 are actually 13%. The Internet regrets the error.

The iPad Air with A12 has a 12% higher multicore score in geekbench 5 than the iPhone XS with A12

iPhone XS - Geekbench

Benchmark results for an iPhone XS with an Apple A12 Bionic processor.

browser.geekbench.com

iPad Air (3rd generation) - Geekbench

Benchmark results for an iPad Air (3rd generation) with an Apple A12 Bionic processor.

browser.geekbench.com

 

[PPietra]

View attachment 962906

iPhone XS - Geekbench

Benchmark results for an iPhone XS with an Apple A12 Bionic processor.

browser.geekbench.com

View attachment 962907

iPad Air (3rd generation) - Geekbench

Benchmark results for an iPad Air (3rd generation) with an Apple A12 Bionic processor.

browser.geekbench.com

You don’t pick some random test which you don’t have any idea how it was done, under unknown thermal conditions. You pick the average score for the device. Those aren’t the average scores for the devices.

 

[PPietra]

I was agreeing with you. Just saying that it's 13%, not 12%.

I was just making a joke, pointing out that it would be 13% faster means 12% slower

 

[EugW]

You don’t pick some random test which you don’t have any idea how it was done, under unknown thermal conditions. You pick the average score for the device. Those aren’t the average scores for the devices.

No, you should never pick the average scores listed on Geekbench's list. That's exactly why your claim is false.

Ideally, you should make sure nothing else is running, after a fresh reboot, and let the benchmark run. When you do that, you are usually able to get scores within about 5% of the peak scores posted on Geekbench, whether it is an iPad or an iPhone (unless you have a giant thick case on the phone in a 27C room or something).