Slide 1: Exploring Supermicro's H14 Solutions Powered by New AMD Instinct MI350 Series GPUs
Exploring Supermicro’s H14 Solutions Powered by New AMD Instinct MI350 Series GPUs
Slide 2: Speakers
Speakers • Cyril Barranta, Digital Marketing, Supermicro (Host) • Ted Marena, Business Development Director, AMD • Tharun Kuppireddy, Sr. Solution Manager, Supermicro 2 Accelerate Everything © 2025 Super micro
Slide 3: Housekeeping
Housekeeping Attachments Q&A On-demand 3 Accelerate Everything © 2025 Super micro
Slide 4: Agenda
• Introduction and Housekeeping • AMD Instinct MI350 Series Overview • Supermicro H14 GPU Servers • Q&A 4 Accelerate Everything © 2025 Super micro
Slide 5: Advancing the AI datacenter with Supermicro
Advancing the AI datacenter with Supermicro 5 |
Slide 6: Advancing the AI Data Center with Supermicro
Advancing the AI Data Center with Supermicro Software Solution CPUs GPUs Networking Systems Design Open Software Stack Annual Roadmap, DPUs, UALink + Ultra Rack Level products World’s Best Server CPU Upstreamed Pytorch, Leadership Memory and Ethernet DLC & Air-cooled Servers for Cloud, Enterprise, A
Slide 7: Supermicro H14 Block Diagram
Supermicro H14 Block Diagram AMD Instinct AMD Instinct MI350X/355X MI350X/355X AMD Instinct AMD Instinct MI350X/355X MI350X/355X AMD Instinct AMD Instinct MI350X/355X MI350X/355X AMD Instinct AMD Instinct MI350X/355X MI350X/355X PCIe® [and OAM] switch EPYC 9575F EPYC 9575F Salina DPU Pensando Pollar
Slide 8: Leadership roadmap commitment continues
Leadership roadmap commitment continues 2026 2025 2024 2023 MI400 SE RIES MI350 SE RIES MI325X Next-Gen MI300X Architecture Compute and HBM3E Memory Memory Leadership Increased Gen AI Compute Leadership 8 | Roadmap subject to change
Slide 9: MI350X / MI355X Product Overview
MI350X / MI355X Product Overview • CDNA 4 architecture using N3P process technology • Support for new 4- and 6-bit data formats • Improved HBM subsystem • Up to 288GB of HBM3E, 8.0 TB/s HBM BW • Upgraded performance for inference and fine-tuning • Orderable now from Supermicro! • DLC MI355X: AS-4126
Slide 10: Open, Modular Software Stack
Open, Modular Software Stack AI Models and Algorithms Support all major frameworks and models Libraries Expanded Gen AI optimizations New algorithms Compilers and Tools New libraries Runtime Expanding platform support AMD GPUs Extended developer support TRANSITIONING WORKLOADS TO AMD INSTINCTTM GPUS
Slide 11: The Most Demanding Data Center Workloads
The Most Demanding Data Center Workloads Requires End-to-End Leadership in CPU, GPU, and Networking SMC was the first OEM to qualify Instinct GPUs with EPYC 9575F and Pollara NICs 11 |
Slide 12: Summary Slide
Summary Slide • Supermicro is the lead time to market OEM for AMD Instinct GPUs • Both air-cooled MI350X and DLC MI355X are orderable now • Supermicro’s AMD Instinct based servers feature AMD EPYC CPUs and Pensando AI NICs • Get started now! MI325X available today for remote POC evaluations • Access
Slide 13: Supermicro H14 GPU Solutions
Supermicro H14 GPU Solutions w/ AMD Instinct MI350 Series GPUs
Slide 14: A+ Accelerated MI350 Series Ready Systems
A+ Accelerated MI350 Series Ready Systems Air - Cooled Liquid-Cooled AS -8126GS-TNMR AS -4126GS-NMR-LCC w/ AMD Instinct MI350X W/ AMD Instinct MI355X AVAILABLE NOW! 14 Accelerate Everything © 2025 Super micro
Slide 15: Supermicro | AMD Data Center Solutions
Supermicro | AMD Data Center Solutions Time To Online Rack Scale Plug and Play Solutions Air and Liquid Cooled Options Close Partnership with AMD DCBBS for End-to-End Solutions 15 Accelerate Everything © 2025 Super micro
Slide 16: Datacenter Building Block Solutions
Datacenter Building Block Solutions System Level Rack Level Rack-Scale PnP Level Datacenter Infrastructure Level 16 Accelerate Everything © 2025 Super micro
Slide 17: H14 8U OAM GPU System AS-8126GS-TNMR
H14 8U OAM GPU System AS -8126GS-TNMR Integrated Performance, OAM MI350X 8-GPU AIR-COOLED Key Features • Dual AMD 5th Gen EPYC processors with up to 192 cores each and high-frequency SKUs up to 5GHz gives better performance when combined with MI350 series GPUs • 8x PCIe Gen 5.0 X16 LP, and up to 4 P
Slide 18: H14 4U GPU System AS-4126GS-NMR-LCC
H14 4U GPU System AS -4126GS-NMR-LCC Performance Density, AMD MI355X 8-GPU LIQUID-COOLED Key Features *Offered as Rack-level deployment • Dual AMD EPYC processors with up to 192 cores each and high- frequency SKUs up to 5GHz gives better performance when combined with MI350 series GPUs • 8x PCIe Gen
Slide 19: Broadest Portfolio of AMD Server Families
Broadest Portfolio of AMD Server Families CloudDC Hyper Hyper-U GrandTwin /FlexTwin All-in-One Servers with Flexible I/O Options Industry Leading IOPS Server Enterprise-Focused Servers Delivering Memory Leading Multi-Node Architecture with High Power for Cloud Scale Data Centers with Energy Efficien
Slide 20: Supermicro + AMD Scalable AI Cluster
Supermicro + AMD Scalable AI Cluster Large Model Inference AI Workloads Mid-Range Inference Large Training Models Mid-Range Training Platform Scale Small GPU Pod Medium GPU Pod Large GPU Pod Typical Pod Size 8 GPUs 32 - 96 GPUs >256 GPUs GPU-GPU Bandwidth 560GB/s 560GB/s 560GB/s Scale Out Bandwidth
Slide 21: Key Takeaways
Key Takeaways Supermicro Supermicro is first- servers with Supermicro has to-market with AMD MI350 series are the largest Instinct powered available. Remote portfolio of AMD’s AI data-center POC with H14 + A+A+A systems solutions MI350 series GPU across all regions systems coming soon! 21 Accelerate
Slide 22: Q&A
Q&A 22 Better Faster Gr eener © 2024 Supermicr o
Slide 23: Disclaimer
DISCLAIMER Super Micro Computer, Inc. may make changes to specifications and product descriptions at any time, without notice. The information presented in this document is for informational purposes only and may contain technical inaccuracies, omissions and typographical errors. Any performance tes
Slide 24: www.supermicro.com/aplus
www.supermicro.com/aplus 24 Better Faster Gr eener © 2024 Supermicr o