How will smartphone vendors differentiate their products to make users drop an often substantial amount of money for new devices that on the surface only offer incremental improvements?
This is going to be a bit deep geek as the answer hides inside the devices. There's nothing but to divein and check out the tech, I'm afraid.
For the last bit of time, I've had an Oppo Find X5 Pro to try out. It's a well-built new Android 12 5G phone that's powerful (and doesn't overheat), with a great variable refresh rate screen that can display a billion hues on which to show off the excellent photos and videos the device can take. From that angle, the Oppo Find X5 Pro delivers.
Which it should, costing a dollar under two grand as it does.
Apart from having a mere two-times optical zoom when the likes of Apple and Samsung have three and even ten times long-lensing, the Find X5 Pro has copped flak for having a camera system that isn't much of an upgrade compared to what's on the Find X3 Pro predecessor.
That's partly true: the X3 and X5 Pros share the same 50 megapixel sensor from Sony for the main camera.
What you can't see are new features like five-axis stabilisation for the camera, and the first Oppo-designed neural processing unit (NPU).
Oppo (or its mothership BKK rather) called the chip MariSilicon X. The colour output from that is what Hasselblad tweaked for more natural images, which is why the legendary Swedish photography firm's logo is on the back of the Find X5 Pro.
NPUs are arguably more interesting than colour calibration however.
They're not new, but have come into their own for low-power machine learning on smartphones. In simple terms, an NPU is a hardware accelerator which is a specialised chip that can process particular tasks really fast. Think custom, and not general purpose computing here.
It's a bit tricky to provide performance figures that humans can relate to for NPU comparisons, but the Oppo's first NPU is said to be capable of 18 trillion operations per second.
If correct, it tops Apple's A15 Bionic which is rated at 15.8 TOPS.
The MariSilicon NPU works with a red-green-blue-white (RGBW) sensor. Normally, digital sensors have RGB pixels. Adding W to the mix means better light sensitivity and less noise in stills and videos and gives Pro snappers 20-bit ultra high dynamic range with very high contrast, and 20-bit RAW images for greater colour fidelity.
Yes, this is a tech term salad, but the end result is great looking digital approximations of reality, sorry images I mean, that appeal to the very sensitive human eye.
The neural brains of the Oppo Find X5 Pro. Photo / Supplied
Although the MariSilicon X is for imaging only according to Oppo, NPUs can be used for many more tasks.
Apple for example has stepped up the NPU game since the launch of its iOS 15 mobile operating system, and uses its Neural Engine for more than just images.
That includes smarter notifications, FaceID authentication in three dimensions (which can now recognise your mask-clad visage too!), text recognition for pictures, creating Memories in Photos and identifying image and video content, and augmented reality (AR) for directions.
Even Siri the personal digital assistant gets more smarts with neural processing, keeping the data on the device for better privacy and faster handling as the information doesn't have to be sent to cloud servers far away.
There is another NPU, Hexagon, in the Qualcomm Snapdragon 8 Gen 1 chipset that Oppo put into the Find X5 Pro, which provides similar AI/ML acceleration to Apple's Neural engine, and interestingly enough, a Leitz Look mode from another camera legend, Leica, which has launched a phone under its own brand.
Adding an NPU to a smartphone is no small feat. Apart from designing the chips, vendors have to come up with machine learning algorithms that provide tangible improvements for customers, and figure out how to explain the tech to them.
This is emerging tech that enables manufacturers to not just cheat the laws of physics for images enhanced with computational photography, but also to make smart devices smarter.
