Hmm, think they'd introduce new iPads this week at WWDC with the A12 variant?

Or wait for iPhone to debut A12?
 
Hmm, think they'd introduce new iPads this week at WWDC with the A12 variant?

Or wait for iPhone to debut A12?
C’mon, the conference is in less than 24h!
Patience, young grasshopper! (Rumors say no, but what good is rumors this close to WWDC?)
 
I bought my iPad Pro last year so it'll be a few more years.

I think what would be most interesting for an iPad, since Apple touts movie-watching, would be an OLED screen with HDR.

Performance should be well down in the list of priorities, though I guess they're going to keep pushing Apple SOC performance to rival laptop CPUs.

They'll probably put in new SOC and FaceID and call it a day.
 
I have the current model big iPad Pro, and I’d say that reducing the bezel would be a major thing. It’s just a tad too bulky and heavy. They could do it a la the 10.5”, or go full iPhone X in which case they might want OLED.

In terms of SoC performance, the A10x does a good job with RAW image editing and video editing, but this is also an area where demands will increase the next couple of years so more performance would be appreciated. IPads sell a lot more than MacBooks, so it competes with laptops of all brands. More processing grunt is better, and Apple is free of intels market stratification games on this platform.
 
Last edited:
I have Lightroom mobile on my iPad Pro.

I even have a camera connector so on a recent trip, I used it to load my RAW files from my DSLR.

But I used it only for storage, had no desire whatsoever to edit photos on an iPad. I may try it some day but I'm paying $120 a year for Adobe CC subscription so I'm going to use it on my iMac, with the bigger screen and trackpad for pointer control. I don't do much with masks but I still want that pointer rather than tapping parts of an image to edit it.

The import of a couple of hundred RAW files took a long time and I can't imagine how long it would take to stitch say a dozen images for a panorama. Lightroom will peg my CPU during import, applying presets and stitching. I don't have the fastest CPU or GPU that I could have gotten on the 5K iMac but I loaded up the RAM and I can't imagine how much longer similar operations will take on the iPad, even with the fastest SOCs.

I use Lightroom Mobile mainly to view collections (or subsets) of my photos that I've sync'd from my desktop.

Even if there was processing power parity, I doubt I'd edit thousands of RAW files on an iPad.
 
I have Lightroom mobile on my iPad Pro.

I even have a camera connector so on a recent trip, I used it to load my RAW files from my DSLR.

But I used it only for storage, had no desire whatsoever to edit photos on an iPad. I may try it some day but I'm paying $120 a year for Adobe CC subscription so I'm going to use it on my iMac, with the bigger screen and trackpad for pointer control. I don't do much with masks but I still want that pointer rather than tapping parts of an image to edit it.

The import of a couple of hundred RAW files took a long time and I can't imagine how long it would take to stitch say a dozen images for a panorama. Lightroom will peg my CPU during import, applying presets and stitching. I don't have the fastest CPU or GPU that I could have gotten on the 5K iMac but I loaded up the RAM and I can't imagine how much longer similar operations will take on the iPad, even with the fastest SOCs.

I use Lightroom Mobile mainly to view collections (or subsets) of my photos that I've sync'd from my desktop.

Even if there was processing power parity, I doubt I'd edit thousands of RAW files on an iPad.
Nobody edits thousands of RAW files anywhere. Batch processing is where it’s at in those volumes. I don’t use my iPad as my primary editing station either, iMac here as well for that. But for doing stuff away from home, Affinity Photo and the iPad Pro is quite capable of fixing up some files for evaluation (and finishing, if I required it). I anticipate that world+dog will move to 8k video sometime before the 2020 tokyo olympics, and photo file sizes will follow suit. IPads are nice platforms (if not ideal, although file handling is said to keep progressing) for the tasks. With even the new mid-end Mali V76 GPU supporting 8k video encoding and decoding, upcoming Apple products are bound to do the same. Maybe the A12(x)?
 
Supposed Geekbench run shows 128KB L1 caches for A12.

https://browser.geekbench.com/v4/cpu/8816787
The overall scores are somewhat disappointing if true, especially with the multi-core score, unless
  • the A12 focuses heavily on low power (especially given the iPhone shutdown and throttling controversy) and
  • the A12X has much higher clock speeds (similar to the A9 and A9X).
That being said, we've seen this level of improvement before with the A7 → A8. The difference this time is that the A11 itself had a small performance increase over the A10, while the A7 was a massive improvement over the A6.
Code:
SoC                A6      A7      A8      A9     A10     A11     A12 

Geekbench 3 ST    707    1394    1607    2532 
Increase            —      97%     15%     58% 
Geekbench 4 ST    754    1265    1411    2397    3437    4219    4673 
Increase            —      68%     12%     70%     43%     23%     11% 

Clock (MHz)      1300    1300    1400    1849    2340    2390    2490 
Increase            —       0%      8%     32%     27%      2%      4% 

GB3 ST / clock   0.54    1.07    1.15    1.37 
Increase            —      97%      7%     19% 
GB4 ST / clock   0.58    0.97    1.01    1.30    1.47    1.77    1.88 
Increase            —      68%      4%     29%     13%     20%      6%
(The A6 through A11 numbers are from the Geekbench website and archive, and the A12 number is from the claimed leak. For multiple products using the same chip, I used the highest iPhone value.)
 
The overall scores are somewhat disappointing if true, especially with the multi-core score, unless
  • the A12 focuses heavily on low power (especially given the iPhone shutdown and throttling controversy) and
  • the A12X has much higher clock speeds (similar to the A9 and A9X).
That being said, we've seen this level of improvement before with the A7 → A8. The difference this time is that the A11 itself had a small performance increase over the A10, while the A7 was a massive improvement over the A6.
Code:
SoC                A6      A7      A8      A9     A10     A11     A12

Geekbench 3 ST    707    1394    1607    2532
Increase            —      97%     15%     58%
Geekbench 4 ST    754    1265    1411    2397    3437    4219    4673
Increase            —      68%     12%     70%     43%     23%     11%

Clock (MHz)      1300    1300    1400    1849    2340    2390    2490
Increase            —       0%      8%     32%     27%      2%      4%

GB3 ST / clock   0.54    1.07    1.15    1.37
Increase            —      97%      7%     19%
GB4 ST / clock   0.58    0.97    1.01    1.30    1.47    1.77    1.88
Increase            —      68%      4%     29%     13%     20%      6%
(The A6 through A11 numbers are from the Geekbench website and archive, and the A12 number is from the claimed leak. For multiple products using the same chip, I used the highest iPhone value.)

A8 was focused on sustained performance and not throttling just as everyone was beginning to optimize to the benchmarks. I believe that has been their philosophy from A8, so a similar focus on efficiency theoretically isn’t needed.
 
A8 was focused on sustained performance and not throttling just as everyone was beginning to optimize to the benchmarks. I believe that has been their philosophy from A8, so a similar focus on efficiency theoretically isn’t needed.
They've thrown sustained performance out of the window in the recent generations. The small improvements could very well be due to power constraints.
 
They've thrown sustained performance out of the window in the recent generations. The small improvements could very well be due to power constraints.
That would be even more concerning. These results of reflective of a direct copy of architecture, minus cache increase, and a meager 4% clock bump. Doesn’t feel like a ‘Plan A’ design.
 
They've thrown sustained performance out of the window in the recent generations. The small improvements could very well be due to power constraints.
Seems like a reasonable hypothesis. If so, lets hope sustained performance is correspondingly higher and/or battery life longer. Glass isn’t a wonderful material when it comes to dissipating heat but I haven’t seen A11 SoC temps measured. Any problems with heat dissipation would be best dealt with by simply reducing heat generation from the SoC.
 
Do you folks think this speculation at Real World Tech forums is reasonable?
Maynard Handley said:
Perhaps, but I'd suggest another possibility -- this is basically EXACTLY the same CPU as an A11, only "hardened". They took the A11 design, added whatever was believed necessary to protect against every speculation attack they could imagine (presumably things like tagging various structure field by PID or protection level) and the "real" A12 was delayed for a year to get the same treatment. […]
Elsewhere in the discussion, Andrei F notes that the cache numbers that Geekbench displays may not be accurate.
 
Is it only Apple SOCs which has seen incremental improvements?

Would that mean SnapDragon and ARM's own designs have been gaining ground or passing the A1x SOCs?
 
Do you folks think this speculation at Real World Tech forums is reasonable?
Elsewhere in the discussion, Andrei F notes that the cache numbers that Geekbench displays may not be accurate.
Difficult to say without knowing how long the design cycles are. The general idea, that hardening the cores against malicious code will have a price in performance, seems sound.
It will be interesting to see that hypothesis tested, but IF we assume that this is a legitimate result that also accurately describes the finished product (big if) it seems more straightforward to simply assume that Apple has reduced the allowable power draw a bit. And since that is such a simple adjustment to make for the shipping product, taking these numbers as final may be a bit premature.
 
Back
Top