Computex 2026: Intel launches Xeon 6+ with up to 288 E-cores

At Computex 2026 today Intel unveiled its plans for Data Center Group (DGC) and how it is looking to expand its operations to satisfy the modern workload demands. The company has explained how things are changing now, due to the rise of AI, compared to earlier data center requirements. As such it has expressed the need to maximize performance-per-watt, per-core performance, as well as the number of cores per rack, and memory bandwidth.

Team Blue has also acknowledged the need for accelerator (GPU)-coupled training and inferencing nowadays, alongside the predominantly x86 CPU-only approach. Essentially, foundational data centers are now turning into AI training frontiers. In just five years, the company is expecting a 50% split in AI and traditional data center workloads with the majority of it related to AI inferencing.

As such Intel today is expanding its server CPU lineup called the Clearwater Forest that comprises new Xeon 6+ with E-cores. The tech giant had previously highlighted the benefits of this 18A-based processor lineup and how it can help in the expansion of 5G and 6G.

With Xeon 6+, customers will get up to 288 cores E-cores (based on Darkmont architecture), which will bring the necessary high core density that Intel wishes to deliver. A large amount of cache will also be on offer thanks to 576 MB of "enhanced low latency" LLC (last level cache). These server CPUs will also support high-speed 12-channel 8000 MT/s DDR5 memory, which is an impressive feat for a CPU of this calibre considering servers are supposed to be much more stable compared to say a standalone desktop environment. This suggests Intel is highly confident on its IMC (integrated memory controller).

Each compute tile of Clearwater Forest has six modules, and there are four Darkmont E cores inside each one of those modules. Hence the total number of E-cores per compute tile is 24, and 12 of these compute tiles combine to make up the biggest Clearwater Forest SKU packing 288 cores.

With Xeon 6+, Intel is promising 30% greater average performance per thread vs AMD's EPYC. In particular, the company is comparing its new flagship Xeon 6990E+ against EPYC 9965. In terms of efficiency, Intel claims the 6990E+ can deliver 30% better performance per watt than the same EPYC CPU during a typical workload with a 40% CPU usage. And compared to its own last gen Xeon 6780E, Team Blue expects a 55% improvement with a 126% lead in total performance on average.

As mentioned above, Intel wants the highest performance per watt and as such the company also highlighted its in-house Application Energy Telemetry (AET) technology on the new Xeon 6+. Essentially, AET can provide low-overhead, granular CPU-level energy consumption data for different sorts of software workloads and scenarios like app usage, VMs, and such. If you are wondering how it is managing that, AET is getting its telemetry data directly from the silicon thus cutting out the software middleman like the OS, runtime, and such, thus reducing overhead. Also granularity comes to it thanks to its ability to track workloads even as threads move across cores. AET is supported on Linux kernel 7.0 and newer and is aimed at offering better optimization opportunities.

The company is also launching its E835 Ethernet controller and adapter for server-grade usage with speeds of up to 200Gb or 25 GB/s of data throughput.

Besides the Xeon 6+, Intel is also releasing new 6300 series Xeon chips for SMB entry servers. These will pack up to 12 cores, which is a 50% bump up over the previous best parts that were octacores.

Finally the company also teased the next gen Diamond Rapids which is expected to land next year (2027).