Snapdragon X70 Modem Trims Power
Qualcomm’s next 5G modem uses AI to reduce power and improve performance while maintaining the data rates of its predecessor.
Now that 5G has topped out in peak data rate, Qualcomm has changed its focus to power efficiency. Its next-generation Snapdragon X70 modem increases the use of neural networks to manage power, with reductions of up to 60% in some situations. AI-based feedback can boost data rates by up to 24% as well. The company plans to ship the new modem late this year. We expect it’ll first appear in the Snapdragon 8 Gen 2 processor. It will also ship as a standalone modem for automotive, PC, and industrial systems. The changes are mostly in firmware and thus could quickly trickle down to other Snapdragon modems.
The Snapdragon X65 modem also implements some AI capabilities, but the X70 expands on them. The most common application is channel management: a neural network monitors the channel-state information (CSI) and recommends actions to optimize reception. But the results are impressive; Qualcomm reports data-rate gains of 12–26% relative to an unoptimized channel, depending on the user’s distance from the base station and the type of data traffic. It didn’t provide a comparison against the previous-generation AI, but that gain is certainly smaller.
Like the X65, the X70 achieves a peak data rate of 10Gbps but tops out at 7.5Gbps on sub-6GHz networks. To reach the higher rate, the modem can aggregate up to four carriers (4xCA) for a maximum channel bandwidth of 300MHz; the older X65 also supports 300MHz, though only with 3xCA. Few operators have licensed even 200MHz of sub-6GHz spectrum, but with new spectrum auctions coming, this capability may become more useful. For millimeter-wave (mmWave), the X70 can aggregate 8x100MHz channels, generating 7.5Gbps; the peak 10Gbps rate requires combining sub-6GHz and mmWave bands.