Late last year, Qualcomm announced the Snapdragon 835 would be its next-generation smartphone SoC. The new chip is built on a 10nm process in partnership with Samsung, but Qualcomm hadn’t unveiled much in the way of additional details. This week at CES, the company shared new details on the upcoming chip, including clock speeds, core designs, and upgrades over and above the Snapdragon 820.
The Snapdragon 835 will feature the Kryo 280 core and will have eight cores in total in a big.Little configuration. The higher-end cores will run at up to 2.45GHz, while the “little” cores are clocked at 1.8GHz max. It will feature LPDRR4X (a type of LPDDR4 developed by Samsung that uses 0.6V for I/O voltage (Vddq) rather than the standard 1.1V. Qualcomm has shrunk the overall package by 35% while claiming a 20% performance gain, and 25% faster graphics rendering.
It looks as if much of the improved performance is delivered courtesy of clock speed gains rather than any major microarchitectural changes. It’s a well-known fact that smartphones rarely run at their top frequencies for any length of time due to aggressive power management. If the 10nm chip can hold higher clock speeds than its 14nm predecessor, it can deliver better performance as a result. One reason why these gains often fail to result in visible performance improvements is because thermal and power envelopes limit their applicability to specific applications or workloads, and because efficiency improvements have naturally diminishing returns. For example, if it takes 2W to play a video and a new phone takes 1W to play the same video, that’ll have a very noticeable impact on battery life. If it takes 0.8W to play a video and improving technology cuts that to 0.4W, you’re still getting a benefit — but the objective visibility of that benefit is reduced.