Aivres AMD AI Servers:
Raising the Bar on Performance and Efficiency
Aivres AI systems based on leading-edge AMD technologies set new standards for generative AI, agentic AI, exascale AI and HPC infrastructure in the data center. Based on Aivres’ modular building-block architecture, these advanced servers integrate the latest AMD processor and accelerator platforms to deliver scalable, ready-to-deploy systems for the largest AI models.
AMD Instinct™ MI350 Series
AMD Instinct™ MI350 Series GPUs leverage the 4th Gen AMD CDNA™ architecture to deliver leadership efficiency and performance for generative AI and high performance computing workloads in data centers, including training massive AI models, high-speed inference, complex scientific simulations, data processing, and computational modeling.
Enterprise AI Performance, Leadership TCO
Delivering standout performance to deploy gen-AI and agentic AI in existing infrastructures
144 GB HBM3E memory • up to 4 TB/s peak memory bandwidth • up to 4.6 PFLOPs peak performance for MXFP4 / MXFP6
The New Performance Standard for AI & HPC
8 GPU OAMs connected via 4th-Gen AMD Infinity Fabric™ in industry-standard, ready-to-deploy platform
2.3 TB HBM3E memory • 64TB/s peak aggregate memory bandwidth • 161 PFLOPs peak performance for FP4 / FP6
AMD Instinct™ MI350P
KR6268-MI350P 6U PCIe Enterprise AI Server
Boost Your Infrastructure with Scalable Enterprise-Ready AI Performance
- 8x AMD Instinct™ MI350P PCIe GPU Cards
- 2x Intel ® Xeon® 6 or AMD EPYC™ 9005 processors
- 144 GB HBM3E memory per GPU
- Supports up to 24x 2.5″ or 16x E3.S drives
AMD Instinct™ MI350 Series / MI325X
KR9298 10U 8-OAM Server
Powerful Air-Cooled Extreme AI Platform
- AMD Instinct™ MI350 Series / MI325X 8-GPU platform
- 2x AMD EPYC™ 9005 processors
- 2TB total HBM3e memory capacity, 8 TB/s peak memory bandwidth, TDP 1000W
- Supports up to 24 DIMMs slots, 6400 MT/s DDR5 memory in 1DPC
- 10x PCIe 5.0 x16 FHHL + 2 PCIe 5.0 x16 HHHL slots
- 12V and 54V partitioned power distribution
AMD Instinct™ MI300 Series
AMD Instinct™ MI300 Series accelerators are built on AMD CDNA™ 3 architecture to power the most demanding AI and HPC workloads, offering exceptional compute performance, large memory density with high bandwidth, and support for specialized data formats.
2 TB HBM3E memory • 48TB/s peak memory bandwidth • 20.9 PFLOPs peak performance for FP16 / BF16
AMD Instinct™ MI300X
KR6298-E2 6U 8-OAM Server
Scalable Platform for Data Center AI
- AMD Instinct™ MI300X 8-GPU platform
- 2x AMD EPYC™ 9004 processors
- 1.5 TB total HBM3e memory capacity, 6 TB/s total peak memory bandwidth, TDP 1000W
- Supports up to 24 DIMMs slots, 4800 MT/s DDR5 memory
- 2x PCIe 5.0 x16 FHHL + 8x PCIe 5.0 x16 HHHL slots
- 24x SATA/NVMe drives, up to 16x NVMe U.2
