Micron Technology Unveils MRDIMMs to Scale Up Memory Densities on Servers

[btarunr]

Micron Technology, Inc., today announced it is now sampling its multiplexed rank dual inline memory module (MRDIMMs). The MRDIMMs will enable Micron customers to run increasingly demanding workloads and obtain maximum value out of their compute infrastructure. For applications requiring more than 128 GB of memory per DIMM slot, Micron MRDIMMs outperform current TSV RDIMMs by enabling the highest bandwidth, largest capacity with the lowest latency and improved performance per watt to accelerate memory-intensive virtualized multi-tenant, HPC and AI data center workloads.1 The new memory offering is the first generation in the Micron MRDIMM family and will be compatible with Intel Xeon 6 processors.

"Micron's latest innovative main memory solution, MRDIMM, delivers the much-needed bandwidth and capacity at lower latency to scale AI inference and HPC applications on next-generation server platforms," said Praveen Vaidyanathan, vice president and general manager of Micron's Compute Products Group. "MRDIMMs significantly lower the amount of energy used per task while offering the same reliability, availability and serviceability capabilities and interface as RDIMMs, thus providing customers a flexible solution that scales performance. Micron's close industry collaborations ensure seamless integration into existing server infrastructures and smooth transitions to future compute platforms."



By implementing DDR5 physical and electrical standards, MRDIMM technology delivers a memory advancement that allows scaling of both bandwidth and capacity per core to future-proof compute systems and meets the expanding demands of data center workloads. MRDIMMs provide the following advantages over RDIMMs:

• Up to 39% increase in effective memory bandwidth
• Greater than 15% better bus efficiency
• Up to 40% latency improvements compared to RDIMMs

MRDIMMs support a wide capacity range from 32 GB to 256 GB in standard and tall form factors (TFF), which are suitable for high-performance 1U and 2U servers. The improved thermal design of TFF modules reduces DRAM temperatures by up to 20 degrees Celsius at the same power and airflow, enabling more efficient cooling capabilities in data centers and optimizing total system task energy for memory-intensive workloads. Micron's industry-leading memory design and process technology using 32 Gb DRAM die enables 256 GB TFF MRDIMMs to have the same power envelope as 128 GB TFF MRDIMMs using 16 Gb die. A 256 GB TFF MRDIMM provides a 35% improvement in performance over similar-capacity TSV RDIMMs at the maximum data rate. With 256 GB TFF MRDIMMs, data centers can drive unprecedented TCO benefits over TSV RDIMMs.

"By leveraging DDR5 interfaces and technology, MRDIMMs provide seamless compatibility with existing Xeon 6 CPU platforms, giving customers flexibility and choice," said Matt Langman, vice president and general manager of Datacenter Product Management, Intel Xeon 6 at Intel. "MRDIMMs provide customers a full choice of higher bandwidth, lower latencies and capacity points for HPC, AI and a plethora of workloads, all on the same Xeon 6 CPU platforms that also support standard DIMMs. Our customers will benefit from Micron's broad portfolio of MRDIMMs ranging from 32 GB to 256 GB densities and in standard and tall form factors that will be validated with Intel Xeon 6 platforms."

"As processor and GPU vendors have given us exponentially more cores, the memory bandwidth required to deliver balanced system performance has lagged. Micron MRDIMMs will help close the bandwidth gap for memory-intensive workloads like AI inference, AI retraining and countless high-performance computing workloads," said Scott Tease, vice president and general manager of AI and High-Performance Computing at Lenovo. "Our collaboration with Micron is stronger than ever, and laser-focused on delivering balanced, high-performance, sustainable technology solutions to our mutual customers."

Micron MRDIMMs are available now and will be shipping in volume in the second half of calendar year 2024. Subsequent generations of MRDIMMs will continue to deliver up to 45% better memory bandwidth per channel over similar-generation RDIMMs.6 For more information on Micron MRDIMM innovations, visit: Micron MRDIMM memory.

View at TechPowerUp Main Site

 

[Wirko]

AMD will probably bring MRDIMM compatibility with Zen 5 Epyc processors. At least it should. 192 cores with only 12 memory channels will need all the bandwidth they can get.

 

[eidairaman1]

Basically Datacenter level Hardware here

 

[ypsylon]

With all advancement with RAM in stick form I think, server segment has only one way forward - HBM (like Xeon Max). There is finished real estate on server motherboard (especially 1U blade half node) and future generations surely will come with 16, 20 or 24 channels.

 

[Wirko]

With all advancement with RAM in stick form I think, server segment has only one way forward - HBM (like Xeon Max). There is finished real estate on server motherboard (especially 1U blade half node) and future generations surely will come with 16, 20 or 24 channels.

Think about how much a terabyte of HBM would cost. Too much for just about every application.

Now if they manage to push the data rate of DIMMs further up, which still seems possible, there will be no need for more channels in the near future.