The Qualcomm Snapdragon 820: everything we know and what we expect

The existence of a new, more powerful SoC that's destined to succeed the troubled Snapdragon 810 isn't a secret. Chipmaker Qualcomm put the Snapdragon 820 in the public mind as far back as March, at MWC 2015. But the company's forward-looking announcements didn't get people interested in its next-gen premium chipset quite like its predecessor's thermal woes did. The "Snapdragon 810 overheats!!!" that the Internet collectively shouts at Qualcomm is, unfortunately, not a meme, but practical reality.

Thus, at this point in mobile computing history, the usual question that arises when approaching an up and coming Qualcomm high-end processor — something to the tune of "How much of a beast is this thing going to be?" — is now precluded by a more pressing question. Something to the tune of "is the SD820 going to be another disappointment like the Snapdragon 810?"

While it's too early to be specific about that, the Snapdragon 820 is definitely due for having an early look at. This is Qualcomm's next generation premium tier SoC (system on a chip) platform. It's designed on a current FinFET process node and is built upon a custom 64-bit CPU architecture named Kryo.

The importance of custom cores

Utilising 14nm FinFET fabrication technology by Samsung and GlobalFoundries means the Snapdragon 820 will be built on the most efficient process that's commercially available now. Meanwhile, custom cores mean a return to Qualcomm's usual way of doing high-end CPUs. The company doesn't make custom cores (with a license to use the ARM instruction set) just for show, but rather with a specific purpose. Historically, this purpose has been to optimise the living daylights out of them. The 32-bit ARMv7 Krait cores that powered the S4 and 800-series of CPUs are either faster, or more efficient than stock ARM Cortex cores running at the same clock speed.

For example, Qualcomm managed to achieve Cortex-A15 class performance with its Krait 400 cores, but avoided the A15's thermal dissipation and power consumption problems which, if you may remember, led the Nvidia Tegra 4 to its demise. And farther back in time (think 2009), Qcom's custom Scorpion cores successfully one-upped Nvidia and Texas Instruments' competing "stock ARM" chipsets by introducing asymmetric use and pushing the Cortex-A8 architecture to its power and performance limits.

Thus, the announcement of Qualcomm's new Kryo microarchitecture, which will debut inside the Snapdragon 820, has gotten semiconductor enthusiasts frenzied. Expectations are at an all time high, because the chipmaker isn't just faced with continuing its successful legacy, it also needs to say sorry for the Snapdragon 810, which failed to deliver on the usual high expectations. And frankly, the technological industry knows no better excuse than coming up with an awesome product. It means the manufacturer listened and put in genuine efforts to improve its output.

Of course, Qualcomm wheeling out the Kryo architecture in the coming months has much less to do with emotional pursuits than it has to do with the company following its established roadmap. Kryo simply wasn't ready for the Snapdragon 810, but device vendors wanted high-end 64-bit silicone ASAP. It was an "oblige, or lose business" situation, and Qualcomm had to oblige.

But now, the time for Kryo has come. Unfortunately, the successor to Krait has not been properly profiled as of now, but we do know that the architecture allows for a maximum clock speed of 3GHz per core, which is Qcom's biggest CPU speed achievement thus far. Of course, maximum clock speed can only be achieved in short performance bursts, but a 3GHz burst is still a 3GHz burst.