Intel is expanding the capabilities of its Xeon 6 processor platform to tackle massive AI bottlenecks. The company announced select SKUs of its Xeon 6700P chip will support super fast 8000 MT/s RDIMM memory. Production availability is targeted for later this year with rollouts expected during August and September.
Data center leaders are realizing that computing power is only half the battle. If a system cannot move data fast enough, expensive AI accelerators spend valuable cycles sitting idle. Srini Krishna, a Fellow in the Intel Data Center Group, recently wrote in an official corporate opinion piece that memory speed is now a major competitive advantage. This memory upgrade addresses that problem by feeding data directly to hungry processors.
The upgrade to 8000 MT/s RDIMMs delivers a 25% speed increase over older 6400 MT/s memory. It also reduces latency by 6% to keep operations moving. Tests show up to 20% more total memory bandwidth, which is a massive boost for applications sensitive to data limits. Intel is the 1st to bring this level of speed to mainstream server platforms. It will require only a BIOS update for existing systems.
Organizations preparing for even heavier computational workloads have more to look forward to. Intel already has a road map extending into next year. The company plans to support Gen 2 MRDIMM technology reaching 8800 MT/s for its high core count Xeon 6900P platforms. These systems scale from 72 up to 128 cores. This path ensures that memory performance scales alongside raw computing density.
