Memory Technology
Ongoing coverage on Memory Technology.
Latest coverage
Chinese RAM hits 9000 MT/s and even matches the speed and latency of the memory in our western gaming PCs
Chinese memory manufacturer Colorful has achieved a new milestone, reaching 9000 MT/s with its domestically produced DDR5 RAM on the AM5 platform. While Western memory has surpassed this speed years ago, this development signifies rapid progress in China's DRAM technology. This advancement, coupled with increasing demand from the AI boom, contributes to global memory supply and potentially impacts pricing for consumers.
Samsung announces next-gen 3D memory when all we want is reasonably priced DDR5
Samsung has announced its latest advancements in 3D memory technology, primarily targeting data center applications. The announcement comes amidst consumer demand for more affordable DDR5 memory for PCs and consoles.
NAND and HBM have been merged into a dorky megazord for datacenters by Sandisk and SK hynix
SK hynix and Sandisk have unveiled High Bandwidth Flash (HBF), a new memory technology merging High Bandwidth Memory (HBM) and NAND capabilities for data centers. HBF aims to provide high-speed data transfer with expanded capacity, leveraging an open standard via the Open Compute Project and the Universal Chiplet Interconnect Express (UCIe) interface. This development comes amidst a growing demand for AI infrastructure and ongoing memory supply shortages.
Meta's solution to the global memory shortage is to use DDR4 in a DDR5 server, with a custom chip making the…
Meta has developed a custom chip named Vistara, utilizing Compute Express Link (CXL) technology, to integrate slower DDR4 memory modules into servers designed for DDR5. This solution allows Meta to expand server memory capacity beyond DDR5 limitations, using DDR4 for 'cold storage' of less frequently accessed data, effectively increasing total system memory to 1 TB per node.
Gaming GPU Memory Tech Powers Micron’s Cheaper AI Hardware Push
Micron is developing a new AI hardware strategy by stacking up to four layers of GDDR memory modules, typically used in gaming GPUs, to significantly reduce costs. This approach aims to make AI hardware more affordable, with testing planned for this year and early samples potentially available in 2027. While this could benefit AI adoption, it may also impact the availability and pricing of gaming hardware due to increased demand for GDDR memory.
