Intel’s server roadmap and product families have been pretty calm in recent years. The company reorganized its server chips into the Xeon Scalable family and away from the old E7/E5/E3 designations in 2017, and it overhauled the Core X series, but it hasn’t made any dramatic changes to Xeon since AMD launched the Epyc family. Those changes are apparently coming in the first half of 2019, with the launch of Intel’s latest high-end Xeon parts, Cascade Lake Advanced Performance. Unlike standard Cascade Lake chips, these chips will offer substantially more cores and support for up to 2-socket systems.
Intel’s current line of Xeon Scalable Processors like the Xeon Platinum 8180, top out at 28 CPU
It’s not surprising to see Intel making this move. When AMD had 32-core CPUs
Intel predicts performance that’s 1.21x better in Linpack than Xeon Scalable 8180 (3.4x better than AMD Epyc 7601), 1.83x better in Stream Triad (1.3x better than Epyc 7601), and a 17x improvement in AI/Deep Learning inference compared with Xeon Platinum at launch. We don’t have a launch date for Cascade Lake other than 1H, so it’s a good bet that whatever Intel fields will go head-to-head with AMD’s Epyc 2 silicon, codenamed Rome, and built on 7nm at TSMC. As large as 12 memory channels is, it’s not crazy — as CPU core counts scale upwards, motherboards need more memory channels just to keep things relatively balanced.
Like standard Cascade Lake, Cascade Lake AP will feature support for Intel Optane Persistent Memory DIMMs and the hardware-level fixes for problems like Spectre and Meltdown. In the absence of a formal release date, we’re betting these chips don’t tip up much before summer — by which point they should have company from AMD.
Now Read: Intel Smashes Earning Targets, Denies 10nm Problems, Intel’s 9th Gen CPUs Are Still Hard to Find at Retail, and Intel Core i9-9900K Review: Welcome to an Intel-AMD 8-Core Slugfest
