KAYTUS Unveils Cutting-Edge V3 Server Family for Next Open, Green and Scale-Up AI Infrastructure

KAYTUS, a leading provider of IT infrastructure, has announced the launch of its V3 server family. This latest lineup supports the powerful Intel® Xeon® 6 processors and AMD EPYC™ 9005 series processors, offering 17 versatile server models designed for general computing, multi-node configurations, critical computing, storage optimization, and rack-scale systems. The V3 server family is engineered with enhanced intelligence and groundbreaking cooling solutions, ensuring seamless support for diverse AI applications. Packed with advanced computing and storage technologies, it maximizes CPU bus bandwidth, memory bandwidth, and overall capacity. The dual-socket models boast up to 576 cores and clock speeds reaching 5.0 GHz, along with an impressive 136% boost in memory bandwidth. Performance tests reveal that in large-model AI inference scenarios, such as LLaMA2, the V3 series achieves up to 3x higher overall performance compared to its predecessor.

This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20241106878096/en/

KAYTUS V3 Server Family (Graphic: Business Wire)

KAYTUS V3 server empowering all AI scenario applications

The rapid expansion of diverse AI application scenarios has introduced new computing challenges. Each AI application comes with unique characteristics and system requirements, demanding more than just advanced algorithms. A robust, diverse computing power ecosystem and optimized data processing capabilities are equally crucial for success.

The newly released V3 servers support a diverse range of computing platforms, including CPU, GPU, and FPGA, and promote full-stack openness across components, firmware, nodes, and rack-scale systems. The V3 server family tackles computing efficiency with a unified foundation, enabling both node scale-up and network scale-out to align with the scaling demands of AI models. This comprehensive approach ensures support for a wide array of AI scenarios, driving unmatched performance and adaptability.

At the component level, the V3 server supports memory expansion through CXL adapter cards and BOX, fully adhering to CXL specifications. This advancement effectively addresses the imbalance between computing power and data transport capacity, ensuring seamless data flow and boosting overall system performance.

At the firmware level, the V3 firmware incorporates an open firmware management platform built on OpenBMC, showcasing a highly modular design with a layered, decoupled architecture. This setup facilitates asynchronous iteration and functionally decoupled upgrades via standard interface protocols, enabling unified management across a wide range of computing platforms.

At the node level, the V3 server family, built upon the server compute module design specification (OCM), features a modular, decoupled architecture that enables “computing unit decoupling.” This architecture forms the smallest computing unit centered around the CPU and memory, with support for standardized high-speed and low-speed interconnect interfaces. By componentizing processor computing units, the V3 series achieves comprehensive compatibility with both Intel and AMD processors, creating a unified computing architecture, thus reducing user adoption barriers. Users can easily find tailored solutions for various application needs. For example, the KR2190V3 model is fully decoupled, supports both Intel BHS and AMD Turin processors, and facilitates computing standardization for rapid cross-platform and multi-scenario deployment.

The V3 server family also introduces a diversified open acceleration computing platform built on the UBB2.0 specification, compatible with multiple OAM v2.0 modules adhering to OCP open acceleration standards. Featuring a fully modular design and outstanding system energy efficiency, this platform supports the iterative development of large models and AI applications, further driving the innovation and integration of computing power.

At the rack scale level, KAYTUS has introduced a rack scale server compatible with select OCP open standards, designed to embody the architecture of the future. This rack scale server accommodates both 19-inch and 21-inch nodes, offering versatile computing capabilities, open decoupling, and seamless business deployment right out of the box.

Data constitutes the backbone of AI models, making accelerated data processing a critical requirement. Large language models (LLMs), especially multimodal models, now face the challenge of handling significantly larger datasets, necessitating systems with increased memory capacity for efficient data preprocessing and storage. The V3 server is fully optimized for data storage, supporting single-drive capacities of up to 128TB, with a single server capable of accommodating up to 6PB and a single rack achieving a 2.7x boost in storage capacity. This ensures exceptional data processing and storage performance, effectively meeting the high demands of modern AI applications.

Native intelligence enhances computing power efficiency

As IT infrastructure continues to expand in computing power and load, ensuring the long-term stable operation of server hardware is essential. This stability is often challenged by three main issues: unreliable server alerts, inaccurate diagnostics, and failed failure predictions. To combat these challenges, KAYTUS has developed a comprehensive fault diagnosis system, incorporating over 20 new features that enable real-time monitoring and alerts for server health status throughout their entire lifecycle.

The V3 server family has received intelligent upgrades in fault prediction and alerting. For memory, AI training on extensive fault data from thousands of servers enables real-time monitoring of memory errors, achieving an impressive 85% accuracy rate for UCE (Uncorrectable Error) detection. In hard drives, advanced fault warning technology integrated with array cards sets proactive flags. Periodic BMC queries automatically upload alerts for failing drives, ensuring timely replacements. Additionally, diagnostic capabilities have been strengthened with an independent diagnostic module capable of detecting nanosecond-level noise in server motherboard digital signals. This real-time monitoring of key components enhances root cause localization accuracy to 95%, significantly boosting system reliability.

Eco-friendly liquid cooling designed for high efficiency AI workloads

The surge in AI applications has driven a substantial rise in global electricity consumption within data centers. As cooling systems account for a significant share of this energy use, optimizing their efficiency is now more crucial than ever. Embracing innovative cooling solutions is key to building energy-efficient, sustainable data centers that meet the growing demands of the AI era.

The KAYTUS V3 servers set a new standard for energy efficiency with comprehensive liquid cooling optimization. Every KAYTUS server supports liquid cooling and incorporates cutting-edge, safe, and efficient technologies such as two-phase cold plates, negative pressure systems, specialized coatings for motherboard PCBs, and proactive power shutdown mechanisms to address potential leaks. These measures collectively reduce energy consumption and elevate the performance and sustainability of data centers.

KAYTUS’ liquid cooling servers achieve an industry-leading 85% cooling efficiency. With innovations such as memory “sleeper” liquid cooling and waterless heat exchange solutions for hard drives, the system supports various memory thicknesses and hard drive configurations. The liquid cooling rack-scale server enhances heat dissipation, provides flexible node support, strengthens leak protection, and incorporates standardized liquid cooling components, setting a new benchmark for performance and reliability in data center cooling.

The V3 servers also set new standards in overall energy efficiency, incorporating component-level temperature monitoring, an advanced single-fan control strategy, and high-performance fans, enhancing cooling efficiency by 20%. Additionally, all V3 servers come equipped with titanium power supplies as standard, offering an impressive conversion efficiency of over 98%. These power supplies maintain high conversion efficiency across a broad workload range, from 20% to 50%, ensuring optimized performance and energy savings.

KAYTUS Showcasing cutting-edge products tailored for AI-centric applications

The KAYTUS V3 server family stands out with revolutionary performance and efficiency, emerging as a powerful engine in every AI scenario. This groundbreaking series has been meticulously optimized for computing density, storage density, I/O expansion, and full-cabinet deployment, fully meeting the demands of AI computing, data storage, and cloud environments. The V3 servers are already making their mark, being gradually deployed among KAYTUS's leading and premier customers.

The KAYTUS V3 family offers solid support for database scaling, AI inference, and cloud computing with its formidable computing performance and expansion capabilities. The standout KR2280V3 server features an impressive 21 PCIe 5.0 slots and supports up to 45 hard drives, positioning it as a leader in multi-scenario applications. With AI computing performance boosted by 70% and array computing performance doubled, customers can tackle more complex data analysis, accelerate AI model training and inference, and make rapid, informed decisions in a competitive market.

For applications seeking ultimate storage and bandwidth, the KR1280V3 server stands out with exceptional capabilities, providing 512TB of flash storage and 300GB/s of storage bandwidth in a compact 1U space, with 32 E1.S drives directly connected to the CPU. With increasing GPU counts and loads, low-latency data transfer optimization has been enhanced by 13%, ensuring smoother, more predictable performance. This advancement delivers unparalleled speed and reliability for managing large data sets, significantly enhancing the performance and user experience of data-intensive applications.

In terms of energy efficiency and heat dissipation, the K22V3 server features an innovative 2U2N architecture with two independently hot-swappable nodes, which significantly reduces migration costs after server failures. This design boosts heat dissipation efficiency by 40% and cuts energy consumption by 8.5% compared to traditional 2U single-path servers, allowing customers to maintain high performance while lowering operational costs and promoting green, eco-friendly data center operations. The K24V3 server takes it further with support for 8 processors and 1536 cores in a 2U space, delivering a memory bandwidth of up to 6.76TB/s and liquid cooling capabilities exceeding 85%. This robust design ensures superior protection for high-density computing, offering unmatched computing power and stability for big data processing, guaranteeing the continuity and reliability of mission-critical business operations.

For storage-optimized applications, the KR4266V3 server stands out by supporting 48 3.5-inch HDDs in a 4U space, delivering up to 1.4PB of cost-effective storage capacity. It also supports QLC U.2 SSD configurations with single-drive capacities of 128TB, allowing for an impressive 6PB total capacity per server and providing 260GB/s of sequential read storage bandwidth. This setup offers customers a cost-effective solution for storing vast amounts of data while ensuring rapid access and processing. The result is improved storage efficiency and enhanced flexibility in data management, perfectly tailored for today's data-driven demands.

The KRS6000S V3 full-cabinet server is set to make waves with its cold plate heat dissipation density of up to 200W/cm² and advanced negative pressure and two-phase cold plate liquid cooling technology, ensuring zero leakage risk. Scheduled for launch in the middle of next year, this innovative server offers customers an efficient, reliable, and eco-friendly solution for building large-scale, high-power-density data centers.

About KAYTUS

KAYTUS is a leading provider of IT infrastructure solutions, delivering a diverse range of innovative, open, and eco-friendly products for cloud, AI, edge computing, and other emerging applications. With a customer-centric approach, KAYTUS is agile and responsive to user needs through its adaptable business model. Discover more at KAYTUS.com

View source version on businesswire.com: https://www.businesswire.com/news/home/20241106878096/en/

Media
media@kaytus.com

Permalink: http://www.businesswire.com/news/home/20241106878096/en