In today's rapidly evolving AI landscape, staying ahead of the curve is not just an ambition but a necessity. At Liqid, we have alwaysbeen committed to pushing the boundaries of innovation to meet the growing demands of modern data centers. Today, I am thrilled to introduce our latestbreakthrough: the UltraStack 30.
Setting a New Standard at Dell Technologies World 2024
We are live at Dell Technologies World 2024, showcasing our ground breaking UltraStack system, which integrates 30 GPUs into a single server. This achievement sets a new benchmark forGPU density and efficiency, demonstrating our commitment to deliveringcutting-edge solutions.
Unmatched Density and Performance
The UltraStack 30 is more than just a product; it's a testament to our relentless pursuit of excellence in high-density GPU technology. Designed to deliver unprecedented performance, scalability, and efficiency, the UltraStack 30 is set to redefine what enterprises can achieve with their GPU infrastructure.
Surpassing the limitations of traditional server configurations, the UltraStack 30 is equipped to handle a wide array of datacenter workloads, which not only include AI, but other compute demanding workloads like 3D graphics, rendering, and VDI.
Key Features and Benefits
- Unprecedented GPU Density: By integrating 30 GPUs in a single server, the UltraStack 30 provides record-breaking density, enabling organizations to maximize their computational power within a compact form factor.
- Scalability at its Best: The UltraStack 30 can be configured to meet the specific needs of diverse workloads, from AI and ML to HPC and data-intensive applications, allowing seamless scaling as required.
- Energy Efficiency: With its energy-efficient design, the UltraStack 30 delivers top-tier performance while minimizing power consumption, reducing operational costs.
AI Training vs. Inference: A Shift in Demand
Due to sheer cost, we believe that AI training will increasingly be performed by hyperscalers, while inference tasks will occur much more frequently at the customer level. This means the demand for low-latency inference solutions will surge. Inference requires delivering results quickly to users, and server designs like Liqid's UltraStack 30, which composes up to 30 GPUs into a single system, eliminate inter-server node communication and latency. This architecture is crucial for achieving the low latency required for efficient inference.
Supporting AI Initiatives
"In the race to accelerate time to value for AI initiatives, organizations struggle to address the data demand generated by AI while balancing the necessary GPU density and utilization," said Scott Sinclair, Practice Director, Infrastructure & Modernization, EnterpriseStrategy Group. "Fortunately, players like Dell Technologies and NVIDIA working with innovators like Liqid are delivering impressive new configurations and performance benchmarks like the ones achieved with the Liqid UltraStack reference architecture to help organizations keep pace with modern AI demands."
Our latest 30-way UltraStack offerings build upon the 10-and 20-way solutions released in late 2024, delivering a 2x improvement in operations per second and 50% lower TCO than comparable GPU density solutions delivered by four 8-GPU servers. This new standard redefines the limits of high-density GPU systems.
Optimal Memory Bandwidth with Composable CXL
As we move forward, we recognize that optimal memory bandwidth will be crucial for maximizing performance. This is why composableCXL (Compute Express Link) is on the horizon at Liqid, promising to further enhance our capabilities in delivering high-performance, low-latency solutions.
Join Us on This Exciting Journey
We are incredibly excited about the potential of theUltraStack 30 and can't wait to see how our customers will leverage this powerful solution to drive their initiatives forward.
Thank you for your continued support and trust in Liqid.Together, we are shaping the future of high-performance computing.
Stay tuned for more updates and insights!
