The InnoPro-8000 is purpose-built for demanding AI inference and accelerated computing workloads. Its 5U architecture combines dual-socket CPU performance with dense GPU capacity, supporting two 4th/5th Gen Intel Xeon Scalable processors with a processor TDP of up to 350 W and up to eight double-width GPUs with a combined GPU power budget of up to 6000 W.

Sixteen DDR5 memory slots support DDR5-4800/5600 and a total memory capacity of up to 4 TB. A high-bandwidth expansion design provides ten MCIO x8 interfaces from each CPU with PCIe 5.0 x8 signaling, along with additional PCIe 4.0 and PCIe 3.0 connections for storage and system expansion.

Storage options include one onboard M.2 PCIe 3.0 x4 slot, support for up to four U.2 devices, and optional expansion for up to eight SATA 3.0 drives. RAID 0/1/5/10 and VROC options help customers tailor storage performance and data protection to their workloads. Integrated 1GbE and 2.5GbE networking supports straightforward deployment in enterprise and research environments.

An optimized airflow path, dedicated CPU and GPU cooling fans, and up to five hot-swappable CRPS power supplies with 4+1 redundancy help maintain stable operation under sustained load. Multi-layer BIOS, driver, and operating-system tuning further improves direct-connect multi-GPU communication for LLM inference and other bandwidth-intensive applications.

Key capabilities:

– 5U rackmount AI server supporting up to eight double-width GPUs
– Dual 4th/5th Gen Intel Xeon Scalable processors, up to 350 W TDP per CPU
– 16 DDR5 DIMM slots, DDR5-4800/5600, up to 4 TB
– Twenty CPU-side MCIO x8 interfaces with PCIe 5.0 x8 signaling
– One onboard M.2 PCIe 3.0 x4 slot and support for up to four U.2 devices
– Optional expansion for up to eight SATA 3.0 drives
– RAID 0/1/5/10 and VROC options
– 1GbE network interface with PXE support plus one 2.5GbE interface
– Up to five 1300 W/1600 W/2000 W/2400 W CRPS hot-swappable power supplies
– 4+1 power redundancy for high availability