Apple
announced the iPhone yesterday, along with its A10 microprocessor. We’ll have
to wait a bit to see how the new CPU compares with the A9 in terms of overall
performance, but Apple shared some significant details of what to expect with
the new chip.
Past
rumors have suggested a second-generation 14nm SoC built at TSMC, but Apple
said nothing about either its process node or foundry manufacturer of choice.
What we do know is that the new A10 is a quad-core, 3.3-billion transistor CPU
(the A9 CPU contained more than two billion transistors) with an estimated 40%
performance advantage over A9 (according to Apple).
Big.Little-ish
Ever
since it debuted the iPhone 4s, Apple has stuck to a dual-core policy for its
iPhone, even as its Android competitors steadily boosted their core counts.
It’s become common for flagship Android devices to field 4-8 cores — either a
single cluster of high-end CPUs like the Snapdragon 820, or a combination of
high-end and low-power CPUs in a unified cluster via ARM’s big.Little.
When
ARM debuted big.Little it was by no means certain that the idea would take off.
As we covered at the time, making the large and small CPU clusters talk to each
other effectively required some heavy lifting from both SoC manufacturers and
Android itself. Intel, meanwhile, believed that it could use Dynamic Frequency
and Voltage Scaling to effectively address the market with a single CPU or CPU
cluster, rather than relying on clusters of high-performance and
high-efficiency CPU cores that share data and workloads amongst themselves.
Apple
isn’t calling its own cluster implementation big.Little, but its implementation
of the same concept appears similar. The A10 will combine two high performance
cores with 40% better performance than the A9 with two high-efficiency cores
that draw 1/5 the power of the A10’s high-end chips. The controller that
manages these workloads is custom Apple silicon, so we don’t know how well Apple’s
solution will compare with ARM’s. Apple could be facing a learning curve here —
the first chips to implement big.Little didn’t actually do so properly and it
took several technology and Android iterations before the standard was fully
supported in both hardware and software. Alternately, Apple may have been
developing the technology through several revisions, and only rolled it out
when it knew it had everything nailed down.
Apple
is also claiming that the new GPU inside the iPhone 7 will deliver 50% more
graphics performance than A9 and 240x more performance than the original
iPhone. These gains appear to have been plotted against the iPhone 6s rather
than the 6s Plus. The iPhone 7’s battery life is said to have improved as a
result of these changes, with more than two hours of run-time compared with the
iPhone 6s and one hour more than the iPhone 6s Plus.
Apple’s
decision to use a similar technology to big.Little should pay dividends in
terms of battery performance, but we’ll have to wait for more details before we
can compare the two technologies. The latest revisions of big.Little allow for
workloads to be shared across all the CPU cores in a device, provided its
thermal budget allows for this. Apple may have duplicated this functionality
with the A10 Fusion. Apple is also claiming that its screens are now 25%
brighter than before, and these types of changes can have a significant impact
on battery life.
Heading
into the event there were rumors that the iPhone would use an Intel-branded
modem rather than a Qualcomm chip, but if this is true Tim Cook didn’t mention
it. The new Fusion technology Apple is debuting here is a significant shift for
its device strategy and it’ll be interesting to see how the new hardware
impacts overall SoC performance. The GPU is presumably by Imagination
Technologies — they’ve built the hardware for every iPhone and there’s no sign
Tim Cook changed suppliers for the iPhone 7.
for more development and traing cum Webiste Development call +919568180808 or go to SPOKES TECHNOLOGIES
No comments:
Post a Comment