Macworld - USA (2020-03)

84 MACWORLD MARCH 2020

FEATURE APPLE’S A14 PROCESSOR

When it comes to using the GPU for
computation, I think we can expect a similar
speed bump. A bigger GPU plus faster
clock speeds and more memory bandwidth
could give us 50 percent or more—making
the Geekbench 5 compute score jump to
the 9,500 to 10,000 range.

IMAGE
PROCESSING
AND NEURAL
ENGINE
The A-series
processors do a lot
more than just house
CPU and GPU cores.
They contain
specialized hardware
for specific tasks like
image processing (to

turn data from the cameras into stellar photos

and video), video encoding and decoding,

and a Neural Engine that powers machine

learning tasks that are used throughout iOS.

In the A13, Apple added specialized

hardware to handle matrix multiplications

and an updated “machine learning

controller” to balance those compute tasks

among the various parts of the

processor—CPU, GPU, and Neural Engine.

It did not, as I predicted, add cores to the

Neural Engine, but since all parts of the

CPU, GPU, and NE run at higher clock

speeds, performance improved.

This is an area Apple cares deeply

about, and is a necessary part of

improving photo and video quality,

augmented reality, and many of Siri’s

functions throughout the operating system.

With the higher transistor budget afforded

by the 5nm manufacturing process, I think

GPU Compute is increasingly important. I
wouldn’t be surprised if it’s 50 percent faster
in the A14.

Apple added new matrix math hardware to the CPU in the A13. The

A14 may not have a similar “new” feature but more Neural Engine

cores seems like a reasonable expectation.

A11 A12 A13 A14*

8000

6000

4000

2000

0

GEEKBENCH 5

COMPUTE SCORE

HIGHER SCORES ARE BETTER

6370

4521

3272