back to top
More
    HomeTechSupermicro Launches 6U SuperBlade with Intel Xeon 6900 and Liquid Cooling

    Supermicro Launches 6U SuperBlade with Intel Xeon 6900 and Liquid Cooling

    Published on

    Cursor’s AI Agents Now Write Code, Run It, and Prove It Works

    Cursor just crossed a threshold that most AI coding tools have only promised. Its cloud agents no longer just generate code; they spin up their own virtual machines, run the software they build, capture video evidence, and submit pull requests that are ready to merge.

    Supermicro unveiled its densest blade server yet the SBI-622BA-1NE12-LCC 6U SuperBlade powered by dual Intel Xeon 6900 series processors with up to 256 P-cores. The system supports both air cooling (5 nodes per 6U) and direct liquid cooling (10 nodes per 6U), delivering up to 100 servers and 25,600 cores in a single rack. Announced December 31, 2025, the SuperBlade targets HPC and AI workloads across manufacturing, finance, scientific research, and climate modeling.

    What’s New

    The 6U SuperBlade represents Supermicro’s “most core-dense” blade platform to date, according to CEO Charles Liang. Each blade supports dual Intel Xeon 6900 processors up to 128 P-cores per CPU at 500W TDP plus up to 3TB of DDR5-6400 RDIMM or 1.5TB of DDR5-8800 MRDIMM. The 32-inch-deep enclosure fits standard 19-inch racks, eliminating the need for deep-rack infrastructure.

    Key hardware specs include four PCIe 5.0 NVMe drives, two hot-swap E1.S SSDs, two M.2 drives, and three PCIe 3.0 x16 slots for 400G InfiniBand or GPU acceleration. Integrated 25G Ethernet switches with 100G uplinks reduce external cabling by up to 93% compared to traditional 1U servers.

    Why It Matters

    Data centers running memory-intensive AI inference or HPC simulations can now deploy 10× the compute density of rack-mount servers in the same footprint. Shared power supplies, fans, and chassis management lower operational costs and power consumption by up to 50% versus discrete servers. Direct liquid cooling (CPU-only or CPU/DIMM/VRM cold plates) enables sustained 500W-per-socket operation without thermal throttling.

    The hot-swappable blade design and remote chassis management module (CMM) let admins control power capping, BIOS configs, and KVM access per blade even when CPUs are offline.

    Intel Xeon 6900: P-Core Performance

    Intel’s Xeon 6900 “Granite Rapids” processors use a chiplet design with up to three compute tiles on the Intel 3 process node. Each processor supports 12 memory channels, 96 PCIe 5.0/CXL 2.0 lanes, and up to 504MB L3 cache. Key advantages for the SuperBlade include:

    • 8800 MT/s MRDIMM support for bandwidth-hungry AI training
    • Six UPI 2.0 links at 24 GT/s for dual-socket coherency
    • Intel AMX with FP16 for accelerated matrix operations

    Air vs. Liquid Cooling

    Config Nodes per 6U Cores per Rack Use Case
    Air-cooled 5 blades 12,800 cores Standard HPC, moderate TDP
    Liquid-cooled 10 blades 25,600 cores AI training, max TDP, dense racks

    Liquid cooling delivers 2× the node density and supports full 500W CPU operation, while air cooling suits lower-power or budget-conscious deployments.

    What’s Next

    The SBI-622BA-1NE12-LCC is now listed on Supermicro’s product page with detailed specifications. Pricing and volume availability have not been disclosed. Supermicro previously showcased liquid-cooled roadmaps at SC25 in November 2024, signaling broader adoption of direct-to-chip cooling across its portfolio.

    Organizations planning 2026 data center upgrades should evaluate rack power and cooling infrastructure liquid-cooled configurations require CDU or facility water loops, while air-cooled models work with existing CRAC systems.

    Featured Snippet Boxes

    What is Supermicro’s 6U SuperBlade?

    It’s a high-density blade server enclosure holding up to 10 server blades powered by dual Intel Xeon 6900 processors (256 P-cores per blade). A single 6U chassis replaces up to 10 traditional rack servers.

    How many cores fit in one rack?

    Up to 25,600 P-cores across 100 server blades (10 enclosures × 10 blades) when using liquid cooling. Air-cooled configurations support 12,800 cores.

    Which workloads benefit most?

    HPC simulations, AI training and inference, financial modeling, climate/weather forecasting, and scientific research requiring high memory bandwidth and core count.

    Does it require special rack infrastructure?

    No. The 32-inch-deep enclosure fits standard 19-inch racks. Liquid-cooled models need facility water loops; air-cooled versions work with existing cooling.

    Mohammad Kashif
    Mohammad Kashif
    Senior Technology Analyst and Writer at AdwaitX, specializing in the convergence of Mobile Silicon, Generative AI, and Consumer Hardware. Moving beyond spec sheets, his reviews rigorously test "real-world" metrics analyzing sustained battery efficiency, camera sensor behavior, and long-term software support lifecycles. Kashif’s data-driven approach helps enthusiasts and professionals distinguish between genuine innovation and marketing hype, ensuring they invest in devices that offer lasting value.

    Latest articles

    Cursor’s AI Agents Now Write Code, Run It, and Prove It Works

    Cursor just crossed a threshold that most AI coding tools have only promised. Its cloud agents no longer just generate code; they spin up their own virtual machines, run the software they build, capture video evidence, and submit pull requests that are ready to merge.

    Claude Cowork Enterprise Plugins: How Anthropic Is Rebuilding the AI Workplace in 2026

    This is what separates it from generic AI assistants. The update gives IT admins, department heads, and knowledge workers a unified system to build, manage, and deploy AI agents that follow how their organization

    Anthropic Acquires Vercept: Claude Now Operates Software Like a Human

    Anthropic’s acquisition of Vercept is not a talent grab or a defensive move. It is a direct investment in making Claude the most capable computer-using AI agent available. The bottleneck has always

    Samsung Galaxy Buds4 Pro Officially Lauched: Everything You Need to Know Before March 11

    Samsung launched the Galaxy Buds4 series at Galaxy Unpacked 2026 in San Francisco, and the lineup arrives with more hardware changes than any previous Buds generation. The Buds4 Pro moves to a dual-

    More like this

    Cursor’s AI Agents Now Write Code, Run It, and Prove It Works

    Cursor just crossed a threshold that most AI coding tools have only promised. Its cloud agents no longer just generate code; they spin up their own virtual machines, run the software they build, capture video evidence, and submit pull requests that are ready to merge.

    Claude Cowork Enterprise Plugins: How Anthropic Is Rebuilding the AI Workplace in 2026

    This is what separates it from generic AI assistants. The update gives IT admins, department heads, and knowledge workers a unified system to build, manage, and deploy AI agents that follow how their organization

    Anthropic Acquires Vercept: Claude Now Operates Software Like a Human

    Anthropic’s acquisition of Vercept is not a talent grab or a defensive move. It is a direct investment in making Claude the most capable computer-using AI agent available. The bottleneck has always
    Skip to main content