# Key Enhancements Release 25.06 is a major version that includes new features, scalability, security and usability improvements, and bug fixes. ## New zCompute MSP Admin Guide The [zCompute MSP Admin Guide](https://guides.zadarastorage.com/cs-admin-guide/latest/) is now available. This guide is designed to help managed service providers streamline onboarding, monitor and manage multiple tenant resourcess efficiently. ## High-performance AI Cloud Support zCompute version 25.06 introduces expanded support for high-performance AI workloads, with infrastructure updates designed for modern GPU-based compute environments. * **Nvidia hardware-based cloud support** Provides native platform support for deploying GPU-accelerated workloads within the zCompute environment, optimized for AI/ML training, inference, and data processing using NVIDIA-based hardware. This includes support for east-west networking based on Spectrum-X architecture, that enables: * High-throughput communication between GPU-enabled instances. * Low-latency data exchange required for distributed AI workloads. * Improved efficiency for multi-node training and parallel processing jobs. * **Nvidia HGX server support** Adds support for NVIDIA HGX systems, which are purpose-built platforms designed for large-scale AI and high-performance computing workloads. This includes support for GPU models such as H100, H200, B200, and B300. These GPU models are designed for advanced AI model training and inference, supporting high memory bandwidth and parallel processing capabilities required for large-scale workloads. * **Nvidia GPUs including RTX Pro 6000 Blackwell Server Edition** Support for additional NVIDIA GPU models, including RTX Pro 6000 Blackwell Server Edition, for GPU-accelerated workloads in zCompute environments. * **Revamp of infrastructure supporting new hardware** Modernization of the underlying platform, accomodating latest compute and I/O capabilities required by modern AI infrastructure. * **New hardware support** **Intel Xeon 6 CPU models** Provides updated CPU support aligned with next-generation server platforms, enabling improved compute performance and better integration with GPU-based workloads. ## New VM Instance Types zCompute 25.06 introduces new familes of VM instance types: * [zCompute Standard high-performance instance types (ZC family)](https://guides.zadarastorage.com/cs-compute-guide/latest/instance-type.html#zcompute-standard-high-performance-instance-types-zc-family) based on Xeon 6 65xx CPU models. * [zCompute Premium high-performance instance types (ZU family)](https://guides.zadarastorage.com/cs-compute-guide/latest/instance-type.html#zcompute-premium-high-performance-instance-types-zu-family) based on Xeon 6 67xx CPU models. * [zCompute Premium high-performance instance types with NVMe (ZUDx and ZUDxA families)](https://guides.zadarastorage.com/cs-compute-guide/latest/instance-type.html#zcompute-premium-high-performance-instance-types-with-nvme-zudx-and-zudxa-families) based on Xeon 6 67xx CPU models. * [zCompute Premium high-performance GPU instance types with NVMe (ZRTX family)](https://guides.zadarastorage.com/cs-compute-guide/latest/instance-type.html#zcompute-premium-high-performance-gpu-instance-types-with-nvme-zrtx-family) based on Xeon 6 67xx CPU models and Nvidia RTX 6000 PRO Blackwell Server Edition. ## GPU Networks Modern AI workloads demand a dedicated, high-performance network fabric that connects GPUs within and across compute nodes, separate from general-purpose infrastructure networks. The **NVIDIA Spectrum-X Compute East-West (E-W) network** provides a purpose-built Ethernet fabric optimized for GPU-to-GPU communication at cloud scale. ### Multi-Tenant GPU-to-GPU Network Provisioning and Management From version 25.06, zCompute’s [GPU Networks](https://guides.zadarastorage.com/cs-compute-guide/latest/gpu-net.html) exposes this fabric as a first-class, self-service resource. Tenants can provision, manage, and isolate GPU-to-GPU networks through the zCompute control plane, without any cross-tenant visibility, interference, or manual operator involvement. ### GPU Network Management (Tenant Admin) [GPU Network Management](https://guides.zadarastorage.com/cs-compute-guide/latest/tenant-account-networking-gpu-networks.html) capabilities enable tenant administrators to deploy and operate GPU-enabled instances, with the required networking and system configuration automatically handled by the platform. Key aspects include: * **GPU-enabled instance types** Instance types are available with full NUMA topology awareness, including PCI hierarchy exposure. This allows workloads to align compute, memory, and GPU resources efficiently based on the underlying hardware layout. * **Pre-configured Ubuntu machine image** A supplied Ubuntu image includes: * NVIDIA GPU drivers * SuperNIC drivers * NVIDIA DOCA stack * NVIDIA NCCL libraries This reduces the need for manual driver installation and environment preparation. * **Automated SuperNIC configuration** During instance provisioning, the platform automatically applies the required SuperNIC configuration for the selected instance type, without the need for additional manual setup by the tenant administrator. ## API updates ### Symp APIs zCompute 25.06 includes Symp API enhancements that extend API coverage for functionality introduced in this release. ### AWS APIs zCompute 25.06 includes AWS API enhancements that extend API coverage for functionality introduced in this release.