Back to all
    Back to all
    Back to all
      Back to all

        Supercloud availability and flexible access

        In highly available cloud regions or as powerful on-premise deployments at large scales.

        availability-v2

        Supercloud is growing its network of cloud regions in Australia, hosted from secure, reliable and backed-up facilities that utilize Firmus’ sustainable hosting technology.

        Alternatively, the same Firmus sustainable hosting technology can be deployed on-premise to provide customers with petaflop-scale data processing capacity to support critical location-specific applications.

        Growing a network of
        cloud regions in Australia

        • On-premise unlocks access on a scale never seen before

        • The power of Firmus’ sustainable technology unlocks access to on-premise cloud services and bare metal at unprecedented scale.

          Available in an array of configurations, Supercloud on-premise is a way to access cutting-edge compute in almost any edge location, even without traditional data centre infrastructure present.

        cube-with-shadow
        [Modular Sustainable supercomputer]
        [SUPERCLOUD ON-PREMISE FEATURES]
        • [bare metal hpc]

          Use for bare metal HPC, augment existing on-prem supercomputing capacity, or make use of the full Supercloud AI stack in an on-prem format.

        • [touchless operation]

          Supercloud can provide full service and operational uptime guarantees, with no knowledge transfer required to operate.

        • [sustainable supercomputing]

          The building block of a sustainable supercomputer. Each Cube 300 hosts as much compute as an NVIDIA DGX Superpod, in a compact, sustainable format.

        • [Deploy anywhere]

          Fully self-contained and reliable, Supercloud on-premise requires minimal on-site infrastructure to deliver at 99.95% SLAs.

        • [flexible use, lower cost]

          Supercloud on-premise lets you access powerful on-site compute without capital cost or complexity. Long-term rental options materially lower TCO and risk of access to cutting-edge technology.

        • [Upgrade cycle]

          Deploy A100s today, plan for what’s to come tomorrow. Easily upgrade on-prem to latest-generation GPUs and CPUs, such as NVIDIA’s H100, through structured upgrade agreements.

        Cube 300
        configuration options

        A range of current-generation A, B, or C SCE metal is hosted over 135 RU, allowing for flexible builds based on customer use cases. In full A1 configuration (A100 GPU), Cube 300 delivers up to 5.5pf of double precision AI performance equivalent to a DGX Superpod.

        Supercloud on-prem can be delivered in a range of configurations and over varied lease periods. The configuration options listed are examples of what can be delivered, and solutions can be tailored to customer requirements. This extends to networking type, technology, and speed.

        • All A1
        • All B1
        • ALL C1
        • ALL SCS
        8,000
        AMD 7452
        250
        A100, 80GB, NVLINK
        63
        200GB
        BLUEFIELD 2
        958,000 GB
        NVME GEN 4
        32,000 GB
        157PF (FP16); 5.1PF (FP32); 5.1PF (FP64)
        12,000
        Intel 6342
        370
        A40 48GB, NVLINK
        126
        200GB
        BLUEFIELD 2
        1,900,000
        NVMe Gen 4
        64,000
        19PF (FP32); 0.67PF (FP64)
        64,000
        AMD 7713
        0
        -
        63
        200GB
        BLUEFIELD 2
        3,830,000
        NVMe Gen 4
        258,000
        1.3PF (FP32)
        0
        AMD 7452
        0
        -
        126
        200GB
        Bluefield 2
        52,000,000
        NVMe Gen 4
        0
        -