DRAM
5 mentions across all digests
DRAM (Dynamic Random-Access Memory) is a type of volatile computer memory that requires periodic refreshing; global shortages of DRAM and HBM at inference sites are currently bottlenecking the growth of LLM context windows beyond 1M tokens.
DRAM Crunch: Lessons for System Design
Persistent DRAM supply constraints are forcing system architects to redesign for memory bottlenecks as a fundamental constraint rather than assuming abundant supply.
The RAM shortage could get even worse if Samsung labor protests cut production
Samsung faces a planned 18-day strike beginning May 21st over wage and bonus disputes, threatening to worsen the already critical AI-driven RAM shortage. As the world's largest DRAM and NAND producer, production cuts...
The Bromine Chokepoint: How Strife Could Halt Production of World’s Memory Chips
South Korea's 97.5% dependence on Israeli bromine for semiconductor etch gas creates a potential global DRAM/NAND production chokepoint if Iranian strikes disrupt the Negev's irreplaceable extraction complex.
Tailslayer: A technique for reducing tail latency in DRAM operations
Tailslayer addresses a critical memory system bottleneck by reducing tail latency in DRAM operations, which impacts application predictability across databases and real-time systems.
[AINews] Context Drought