Apple (AAPL 0.32%) takes its mobile chip development very seriously. From what I have observed, the company goes to great lengths to get some of the world's best chip architects to join its ranks. For example, earlier this year Apple hired away Intel chief processor architect and fellow Per Hammarlund to work on next generation processors.
This is a company that can attract and retain world-class talent.
Recently, there have been reports discussing the fact that Apple is also working on in-house graphics processor designs (the company currently licenses graphics processors from Imagination Technologies (NASDAQOTH: IGNMF). In this article I'd like to show why it is imperative that Apple rolls its own graphics processors over the long term.
Apple is routinely bested by Qualcomm in terms of smartphone graphics
In order to enable the best graphical experiences on the iOS platform, Apple needs to offer developers leadership graphics performance in its mobile processors. Generally speaking, Apple's chips offer best-in-class graphics performance at launch, but they are generally eclipsed in this area once Qualcomm (QCOM +0.61%) releases its flagship chips.
For example, even the much-maligned Snapdragon 810 offered better graphics performance than Apple's A8 chip, and it would seem that Qualcomm's upcoming Snapdragon 820 -- according to a number of recent performance previews -- will once again have a lead over the A9 in graphics once it hits the market in the spring.
Although consumers probably don't pay too much attention to performance benchmarks when choosing phones, having better graphics performance enables developers to develop richer, more immersive gaming experiences. In short, graphics performance is a fundamental enabler of better user experiences.
In order to further improve the competitiveness of Apple's graphics, I believe the company will need to develop its own graphics architectures.
What will a custom GPU bring to the table?
Apple licenses graphics designs from Imagination Technologies and very often has a "head start" over other licensees. It also tends to choose bigger, more powerful graphics configurations than most other licensees in order to deliver excellent graphics performance.
Interestingly, one could reasonably make the argument that Apple could easily increase graphics performance without resorting to a custom GPU by choosing bigger graphics configurations. However, there are some significant trade-offs associated with doing so.
The first is cost: adding in a larger configuration of a given architecture leads to an increase in silicon area used relative to using a smaller configuration, which increases manufacturing cost. The second is power: assuming a similar architecture, a higher core count configuration will use more power than a lower core count configuration if clock speeds are held constant (thus leading to performance increase roughly proportional to the number of additional cores added).
By developing a customized architecture, assuming Apple has a capable GPU team and is willing to fund that team generously, the company can potentially achieve major performance enhancements that are more area/power friendly than simply tossing in more of the Imagination-designed graphics cores.
Although Imagination has significant experience developing GPUs while Apple is relatively new to this (it still hasn't shipped a custom mobile GPU), I think Apple can afford to throw more money at its graphics designs than Imagination can.
More money thrown at the problem won't necessarily lead to success, Apple's strong execution in developing custom CPUs, as well as in developing the rest of the ingredients required to field successful mobile chips, gives me confidence it will succeed with its graphics efforts.
The big unknown is when we should expect to see a custom Apple GPU design, but I'd be surprised if it didn't show up within the next two or three years.
