82 MACWORLD MARCH 2020FEATURE APPLE’S A14 PROCESSORTransistor count by itself is a poor means
of estimating processor performance, but it
does tell us how much “room” Apple has to
work with. It means more cores, or bigger
cores, or more cache, or any combination of
these improvements.CPU PERFORMANCE
For the A13, Apple didn’t do a lot to
change its fundamental CPU architecture.
It’s still the same number of cores, and
the cores aren’t dramatically different.
Instead, the company relied on the faster
clock speeds afforded by the enhanced
7nm process to improve performance.
In single-threaded performance, the
A13 delivers a 20 percent improvement on
the A12 (in Geekbench 5). It’s a bit more
than I predicted it would be last year, and
the fastest mobile CPU around.Simply following the trend line (which is
quite consistent for recent A-series
processors) we could expect the A14 to
score close to 1,600. The 5nm process
alone should be about 8 percent faster
than the enhanced 7nm process Apple
uses in the A13, which would give us a
score in the mid-1,400s. My guess is that
Apple will likely wind up in the 1,500-1,600
range, due to both higher peak clock
speeds and some architectural
improvements made possible by the much
higher transistor budget.
Multi-core performance is harder to
predict. The A12 and A13 had four small
high-efficiency cores and two big high-
performance cores. Apple already very
efficiently schedules workloads to take
advantage of all cores as much as
possible, so it’s unlikely that we’ll see a bigMulti-threaded performance would go up
if Apple didn’t add more cores, but that’s a
conservative prediction. I expect more.
Single-threaded performance is expected to
improve, but not as much as multi-threaded
performance.A1192111141334A12 A13 A14*1600
1400
1200
1000
800
600
400
200
0GEEKBENCH 5
SINGLE-CORE SCOREHIGHER SCORES ARE BETTERA11 A12 A13 A14*5000
4000
3000
2000
1000
0GEEKBENCH 5
MULTI-CORE SCOREHIGHER SCORES ARE BETTER3463
2398 2774