Baseband Chips are the not so glamorous side of mobile phones, however designers such as ARM are pushing the boundaries of these often overlooked processors.
Simply put, the baseband chip in every mobile phone is what ‘connects’ you to the cell tower. Therefore its performance is absolutely crucial to the operation of the device. With next generation Long Term Evolution (LTE) networks putting greater demands on baseband chips, designers are being pushed to the limit.
ARM’s very public launch of the Cortex R5 and Cortex R7 chips shows just how serious chip designers take this often overlooked component. Richard York, director of product marketing at ARM said that the attention afforded to the chips was “to give visibility to the underlying technology, the technology that brings the data to the user”.
While York aims to shine the spotlight on ARM’s baseband chips, his engineers are lavishing their considerable talent on improving the real-time performance of baseband processors. York said bluntly, “real-time performance is vital”.
When York was pushed for those often quoted headline figures you hear from chipmakers claiming that their latest product is X times faster than the previous generation, he explained that it is not always about increasing visible performance figures. “It’s about meeting the deadline,” said York, adding “there is no need to be faster than you need to be”.
That may sound like a manufacturer artificially sandbagging its products but York’s statement makes perfect sense. Baseband chips need to meet a deadline set by a protocol and beating the deadline could pose just as many problems as missing it.
Next generation wireless networks such as LTE and LTE-Advanced propose lower latency, meaning baseband chips will have to work harder. Typically when it comes to talking about 3G, LTE, WiMax and various other wireless technologies, bandwidth figures are mentioned, however one of the most important aspect of LTE is the stricter latency requirements it places on devices.
LTE requires radio access network latency roundtrip times of just 10ms. Putting that into perspective, the HSDPA standard requires latency to be no greater than 100ms, while UMTS ups that to 150ms. Not only do baseband chips have to transfer greater amounts of data than ever before, but they have to meet stricter guidelines in their delivery schedule. But what exactly does a latency of 10ms mean to the user?
Bandwidth is easy to quantify with claims of being able to watch high definition video on your smartphone, however latency is somewhat harder to pinpoint. As York put it, for most users “latency is the time taken for the webpage to start loading“.
Of course York simplified his answer, but his point is perfectly valid. Assuming ample connectivity is present, then the so-called ‘start up’ time for a webpage to download is probably the greatest performance barrier the user has to overcome. For a webpage that isn’t very important, but for live video, it becomes absolutely vital.
York says that the performance of ARM’s latest baseband chip is not just in the silicon itself but in the way the chips are used. “Our job it to make software development easier for real-time applications,” he said.
York also pointed out that while chip designers such as ARM regularly mention the word efficiency in their marketing materials, for him the term takes on a secondary meaning, one that is just as important as a chip’s power usage. York claims, “ARM tries to increase the coder’s efficiency in making use of the chip,” or to put it bluntly, make it easier to extract the most from the chip.
ARM’s focus on developers seems to be paying off, with York claiming that developers are driving the demand for multi-core baseband chips. York admits that he was surprised that all of ARM’s Cortex R5 customers want the dual core configuration of the chip, saying that in the embedded market “multi-core is ubiquitous,” adding that many of ARM’s customers found running multiple cores at lower voltages gave better performance per watt than running a single, higher voltage core.
Typically with embedded chips power usage is touted as a key performance trait, however according to York it isn’t just about lowering the power usage but rather doing more within a particular thermal design power envelope. York characterised this seemingly never-ending quest for a suitable mix of power and performance by saying, “ARM’s customers are never happy with power usage.”
ARM’s latest processors highlight the advantages of next generation wireless connectivity. LTE’s greater demands on real-time performance should mean that smartphones and tablets can finally become untethered computers rather than fancy gadgets.
So perhaps it’s time for ARM’s own system-on-chip designs to stand back and let their less fashionable cousin, baseband chips, take some of the limelight, as it looks like they will play a big role in the adoption of next generation wireless networks.
Source: The Inquirer