AMD EPYC™ 9005 Processors
Powering Next-Gen AI and Cloud Performance
Dual AMD EPYC™ 9005 processors with “Zen 5” architecture deliver unmatched compute density and efficiency for AI-driven data centers.
Delivers up to 22% more FLOPs per Watt
diamond’s unmatched thermal conductivity enables exceptional heat dissipation, cooler operation, and greater performance headroom—keeping GPUs fast, efficient, and stable even in extreme environments.
Lower GPU temperature by
up to 10°C
Boost token throughput by
up to 15%
Increase FLOPs per Watt by
up to 22%
Throttle-free performance at
up to 48°C (120°F)
Reallocate up to 100% of cooling power to compute
Densify compute and immediately unlock additional capacity within new and existing data centers
CapEx & OpEx cost savings
Empowered by Akash's Diamond Cooling®, G8825Z5 with AMD Instinct™ MI350X GPUs delivers extreme AI and HPC performance
8 AMD Instinct™ MI350X GPUs
Dual AMD EPYC™ 9005 processors with “Zen 5” architecture deliver unmatched compute density and efficiency for AI-driven data centers.
Up to 192 “Zen 5” cores and 5 GHz boost for demanding AI, cloud, and enterprise workloads.
Purpose-built for AI inference and GPU acceleration, delivering superior performance per rack.
5 nm process technology enhances power efficiency and enables flexible deployment across modern data centers.
4th Gen CDNA™ + 288GB HBM3e for massive model inference & HPC
FP16/FP8/FP6/FP4 boost speed with lower power
Secure, multi-tenant ready, seamless GPU-to-GPU
Features Path-Aware Congestion Control for reliability
Delivers 400Gbps via PCIe Gen 5.0 interface
Maximizes efficiency with Intelligent Packet Spray
The AMD ROCm™ software stack delivers an open, frictionless platform for AI and HPC — with drop-in framework support, full open-source access, and seamless integration across leading ecosystems.
With support for 24 DIMM slots and up to 6TB of DDR5-6400 RDIMM memory, the system delivers extreme bandwidth and capacity to handle memory-intensive AI training, HPC simulations, and large-scale data analytics with speed and efficiency.
The system provides a mix of half-height and full-height PCIe 5.0 x16 slots, delivering high bandwidth and versatile expansion for both front-end and back-end network fabrics to enable GPU cluster scaling and ensure future-ready system expansion.
Featuring 15 rear-mounted hot-swap fans, the system ensures efficient airflow and stable cooling for GPU-intensive workloads. Power is supported by 5+1 hot-swap 16,500W redundant PSUs (80+ Titanium), delivering high efficiency and enterprise-grade reliability for continuous 24/7 operation. Diamond Cooling® delivers 22% more FLOPs per Watt and reduces GPU temperatures by up to 10°C.
Unleash extreme AI and HPC performance with the MiTAC G8825Z5 – an ultra-dense 8U server designed for the most demanding workloads in large language model (LLM) training, generative AI, and scientific computing.
The MiTAC G8825Z5 is an advanced 8U server designed to support large-scale AI infrastructure and high-performance computing (HPC) environments.
Overall, the MiTAC G8825Z5 was executed quite well. For those who want simplified maintenance, having all front I/O with only power supplies and fan modules at the rear will be an attractive option as well.
To find a solution and configurations, get quotes, or any technical support from us, please fill in the inquiry form for your needs. We will contact you shortly.
SUBSCRIBE NOW to get the latest news.