QCT – a.k.a. Quanta Cloud Technology – is a bigger player in the data center server business than some might think. The Taiwan-based company, a long-term partner of Intel’s, reports that it produces one in every seven servers sold worldwide and, according to the Omdia industry analyst firm, QCT ranks in the top two globally for GPU servers shipped in 2023.
Tailored to the high-requirement needs of cloud service providers, telecoms and enterprises running public, hybrid and private clouds, QCT servers combine hyperscale hardware efficiency with infrastructure software from diverse industry leaders. QCT solutions address next-generation data center design and operation challenges that have become increasingly complex with the explosion in demand for generative AI and related workloads. Formed in 2012, QCT has years of expertise providing end-to-end HPC and AI solutions for companies.
QCT’s QuantaGrid, QuantaPlex and QuantaEdge product lines feature tool-less modular components, advanced air cooling and liquid cooling options, various high-performance storage options, support for single and dual-width accelerators, and tool-less/hot swappable designs.
At the recent COMPUTEX technology conference in Taipei, QCT debuted a range of new servers supporting the latest Intel® Xeon® 6 processors and Intel® Gaudi® 3 AI accelerators, leveraging the new processors and AI accelerators to optimize compute power and efficiency.
The new QCT QuantaGrid and QuantaPlex servers support a variety of data center workloads, ensuring reliability for enterprise and edge computing applications while delivering power efficiencies at impressive price-to-performance ratios. The result is exceptional performance and efficiency for high-density and scale-out workloads.
Let’s look at QCT’s new servers in more detail.
- Increased efficiency: Intel® Xeon® 6700 processors with E-cores support higher core density and exceptional performance per watt, 1.5x improvement in performance per watt compared to 5th Gen Intel® Xeon® processors, 2.4x improvement in performance per watt compared to 4th Gen Intel® Xeon® processors, and better energy efficiency and sustainability features to reduce one’s operational carbon footprint.
- Enhanced performance: Intel® Xeon® 6700 processors with E-cores include up to 144 cores per CPU and faster memory at the same power envelope as the previous generation, up to 8 memory channels of DDR5 per CPU. Additionally, QuantaGrid S55R-1U and QuantaGrid S55J-2U are powered by Intel® Xeon® 6700 processors with P-cores.
- Greater expandability and flexibility: QCT servers with these processors offer extensive expandability with up to 88 lanes of PCIe 5.0, 64 lanes of CXL 2.0, and 4 UPI 2.0 links that also support up to 24GT/s, representing a 1.8x increase in bandwidth over the previous generation.
- Quality and security: Intel® Xeon® 6 processors support more encryption keys on the latest Intel® Trust Domain Extensions (Intel® TDX) 6 firmware. These processors additionally continue to support Intel® Software Guard Extensions (Intel® SGX) with hardware-based security for confidential computing.
In addition, powered by 5th/4th Gen Intel® Xeon® Scalable processor family, the QuantaGrid D74H-7U is a powerhouse acceleration server purpose-built to tackle the most complex AI and HPC workloads. This server can accommodate up to 32 DDR5 DIMMs and various vendors’ GPU accelerators, including the Intel Gaudi 3, and is optimized for massive data sets, huge AI models such as for natural language processing, large language models (LLMs), and supercomputing applications.
The QuantaGrid D74H-7U architecture is designed around a tool-less modular philosophy. Major components can be removed from the system chassis without any unmounting from the rack for better serviceability and increased system uptime. The modularity in this innovative design allows for forward support of upcoming CPUs and GPUs, including Intel’s next-gen Gaudi AI accelerator.
Finally, the QuantaGrid D75E-4U is a high-performance server that unlocks ultimate flexibility to support LLM/AI/HPC-related workloads with next-gen PCIe GPUs. It is powered by two Intel® Xeon® 6 processors with up to 350W TDP and it supports up to 8x DW air cooling next-gen PCIe GPUs with 600W TDP. All of its PCIe Gen5 I/O expansion slots are designed to reserve up to 150W. This server’s target launch date is 2025 Q1.
Source: insideHPC