The sequential contract price increases for some high-density server DRAM modules are even at nearly 30%. The average worldwide contract price of DDR4 R-DIMM 32GB modules has broken through the threshold of US$200, while the 16GB counterparts have also climbed to about US$100.

DRAMeXchange finds that the market for server DRAM modules benefits from the ongoing PC DRAM price rally. The rapid rise in prices of server DRAM modules during the first quarter of 2017 is also propelling demand from server manufacturers that want to keep sufficient inventory memory before prices go up any further.

DRAMeXchange’s market outlook for this year moreover indicates that the DRAM supply will generally remain tight as leading DRAM makers have decided to firmly limit the increase in their production capacity. DRAM makers will furthermore adjust their product mixes so that prices and margins of their server DRAM products will also remain high.

DRAMeXchange points out that Intel has already sent its next-generation server processor chips to major ODMs for sampling since the fourth quarter of 2016. Intel’s new chips are based on its latest 14nm Purley platform that supports six-channel type memory and allows a single processor to work with as many as 12 DDR4 DRAM modules, whether they are R-DIMM, NVM-DIMM and NV-DIMM types. Due to having the ability to support various types of memory, Purley-based chips will significantly boost the computing power of high-end servers.

AMD is scheduled to release its server chips based on its 14nm Zen architecture this year with the hope of capturing more customers from specialty servers and data center markets. AMD will probably acquire more 14nm wafer capacity in 2018, and by that time Zen-based chips will begin to replace the company’s older server solutions. DRAMeXchange notes that since the rival Intel has a more complete ecosystem for its server products, AMD’s Zen-based chips would face a more uphill challenge in terms of adoption in the x86 market.

Qualcomm will expand the volume production of its first 10nm ARM-based server chip at the start of 2018. While ARM has the advantage of lower power consumption versus x86, the former still lags behind the latter in terms of computing power. ARM-based solutions are mainly deployed in the backend or storage application of the data center operation, whereas servers with x86 solutions championed by Intel continue to handle high-performance computing.