Next-gen AI-first smartphones may feature LLW memory, and not LPDDR

Smartphones are running into a memory problem, and it’s not something a software update can fix. As you may be aware, phone makers are trying to run larger AI models directly on devices, and fortunately, mobile processors are capable of running these LLMs locally. 

The current LPDDR memory technology, however, has limits on how quickly it can feed data to these processors. That’s one reason why AI datacenters rely on HBM, or High Bandwidth Memory. 

According to a new report, a new memory format called LLW, short for Low Latency Wide DRAM, is being developed to address exactly that issue.

LLW could be the HBM solution for smartphones

LLW reportedly draws inspiration from HBM, which achieves high speed by placing multiple memory layers close to the processor and connecting them via an extremely wide interface. The result is significantly higher bandwidth, 10 to 15 times faster than the current best LPPDR memory.

However, HBM also demands complex packaging and cooling solutions that aren’t practical for smartphones. LLW is expected to solve this by providing more bandwidth and lower latency than conventional LPDDR memory without introducing the thermal and space constraints that come with stacked memory designs.

That said, it’s worth taking these claims with some caution. The information available so far comes from industry bloggers and tipsters rather than official company announcements. 

Speaking of which, a leak by tipster Fixed Focus Digital suggests LLW could reduce power consumption by around 50% while improving performance by roughly 1.5 times. 

As for availability, the same leak suggests that LLW is still some years away from widespread adoption. More specifically, it won’t be until the second half of 2027 that we can expect its larger-scale deployment. Xiaomi and Huawei are said to be potential early adopters, although neither company has publicly discussed the technology.

For more daily updates, please visit our News Section.

Stay ahead in tech! Join our Telegram community and sign up for our daily newsletter of top stories!

(Via 1, 2)