Intel Xeon 6+ Clearwater Forest Puts 288 E-Cores and 18A Into Servers

As of Friday, June 12, 2026, Intel Xeon 6+ is no longer just a roadmap name. The processor family, previously known as Clearwater Forest, arrived around Computex 2026 as Intel’s newest high-density server platform for cloud-native services, telecom infrastructure, network-heavy deployments, and emerging agentic AI workloads. The headline part is the Intel Xeon 6990E+, a server CPU with 288 Efficient-cores, no Performance-cores, and a focus on parallel throughput rather than peak single-thread desktop-style speed. Intel’s own product database lists the 6990E+ as launched in Q2 2026, using Intel 18A lithography and supporting two-socket systems, which means a platform can scale to 576 physical E-cores when configured with two top-end processors. ^(intel.com)

The Xeon 6990E+ is built around density, but the spec sheet is still worth reading closely. Intel lists 288 cores and 288 threads, a 2.2 GHz base frequency, 2.8 GHz all-core turbo, and up to 3.2 GHz max turbo. Cache is a major part of the design, with 576 MB on the flagship SKU. The standard TDP is 450 W, while Intel Speed Select Technology - Performance Profile also lists a 330 W configuration with the same 288 active cores and a lower 1.7 GHz base frequency. Platform support includes 12 channels of DDR5-RDIMM-8000, up to 1.5 TB of memory depending on memory type, ECC support, 96 PCIe Gen 5 lanes, 24 GT/s Intel UPI, and 2S scalability. Intel also lists built-in support for technologies such as QuickAssist Technology, Dynamic Load Balancer, Data Streaming Accelerator, In-memory Analytics Accelerator, Intel TDX, SGX, and AVX2. ^(intel.com)

See WD_BLACK 2TB SN8100 NVMe SSD price

Clearwater Forest is not trying to be a one-size-fits-all replacement for every Xeon. The all-E-core design makes the most sense where many smaller jobs need to run at once: microservices, containers, virtual machines, web-scale services, packet processing, 5G core workloads, and CPU-side orchestration for AI systems. Intel positions Xeon 6+ for environments where watts per rack, throughput per core, and predictable latency are important. That is a different pitch from a P-core Xeon aimed at heavy vector math or maximum per-core capability. The trade-off is that Darkmont E-cores do not offer the same feature mix as Intel’s larger server P-cores; for example, The Register notes the design lacks Hyper-Threading, AVX-512, and AMX acceleration. For many scale-out workloads, though, the ability to pack hundreds of independent CPU cores into a socket can be more useful than adding more wide-core features that the software may not fully use. ^{(newsroom.intel.com)}

View NVIDIA DGX Spark Personal AI Desktop Supercomputer on partner website

The manufacturing angle is just as important as the core count. Xeon 6+ is Intel’s first data-center CPU built on Intel 18A, bringing the company’s newest process technology into servers instead of limiting it to client chips. Intel describes 18A as a 2-nanometer-class node with RibbonFET gate-all-around transistors and PowerVia backside power delivery, both meant to improve scaling and efficiency. Clearwater Forest also leans on Intel’s chiplet strategy: reports from the Computex launch describe a disaggregated design with twelve 24-core compute tiles on Intel 18A, base tiles on Intel 3 for cache and memory functions, and I/O chiplets on Intel 7, joined through Intel packaging technologies. In plain terms, this is Intel using both advanced manufacturing and packaging to fit a lot of server compute into a power-constrained platform. ^{(newsroom.intel.com)}

Buy Intel Core Ultra 7 265K here

Intel is clearly aiming Xeon 6+ at the same conversations that include AMD EPYC dense-core processors and Arm-based cloud CPUs. The most quoted comparison is Intel’s claim that the Xeon 6990E+ delivers around 30% better performance per thread than AMD’s 192-core EPYC 9965 in Intel’s tested comparison. That should be read as a vendor claim, not an independent review result, and it does not automatically mean a 30% full-chip performance lead. AMD’s EPYC 9965 has fewer cores but uses simultaneous multithreading, while the Xeon 6990E+ has 288 physical E-cores and 288 threads. The more practical takeaway is that Clearwater Forest gives Intel a very dense x86 option for buyers who want to stay in the Xeon ecosystem while chasing better performance per watt and more consolidation. The Xeon 6+ stack also includes 144-core, 192-core, 264-core, and 288-core models, giving server builders some room to balance core count, cache, memory bandwidth per core, and TDP instead of choosing only the largest SKU. ^(intel.com)

See AMD Ryzen 7 9800X3D price