A few years ago, Samsung announced the 8.5Gbps LPDDR5X RAM, which was the fastest in the world at the time. Technological progress is advancing, and today the company unveiled an even faster LPDDR5X.
This new version operates at 10.7 Gbps, surpassing the 9.6 Gbps LPDDR5T variant launched by SK Hynix last year. This represents a 25% increase in performance over the previous X variants, but speed isn’t the only upgrade.
Samsung builds these chips on a 12nm class process, making the chips smaller. This allowed the company to increase capacity by 30% and now offers 32GB packages with it.
Of course, this product is intended for AI applications. If you’ve been following the topic, you know that large models require a lot of memory – even top-of-the-line GPUs like the Nvidia RTX 4090 are limited in what they can use with their 24GB of VRAM. Nvidia offers professional cards with much more VRAM, but the cost is quite high.
Moreover, with new power-saving technology, the new LPDDR5X chips can reduce power consumption by 25% and can go into low-power mode for extended periods. This is beneficial for battery-powered mobile devices and servers (electricity bills are a significant part of server running costs).
Samsung plans to begin mass production of the 10.7Gbps LPDDR5X RAM by the second half of this year, pending testing with mobile chipsets. The company is also working on the next generation of LPDDR6 RAM.
Source: GSMArena