Lenovo SR670 | 2U Dual-Socket GPU Server, Up to 8x GPUs, 24x DDR4

Q: What GPU configurations does the SR670 support?
A: The SR670 offers flexible GPU configurations: up to four full-height, full-length double-wide GPUs using PCIe x16 slots, or up to eight single-wide GPUs using PCIe x8 slots. It supports NVIDIA Tesla V100, T4, and other high-performance GPUs ideal for AI training, inference, and HPC workloads.

Q: Can the SR670 be used in a cluster environment?
A: Yes, the SR670 is designed for cluster deployments using high-speed fabric and networking to scale out as workload demands increase. It supports multi-user environments and can be managed with LiCO platform for HPC and AI applications across multiple AI frameworks including TensorFlow, Caffe, and MXNet.

Q: What memory capacity does the SR670 support?
A: The SR670 supports 8, 12, 16, or 24 DDR4 DIMMs with dual processors installed. It uses Lenovo TruDDR4 memory operating at speeds up to 2933 MHz, providing substantial memory bandwidth for demanding computational workloads.

Q: What management capabilities are included?
A: The SR670 includes Lenovo XClarity Controller for comprehensive systems management including remote monitoring, configuration, and firmware updates. The OCP adapter supports shared BMC network sideband connectivity for out-of-band management.

Q: What workloads is the SR670 optimized for?
A: The SR670 is optimized for GPU-accelerated workloads including artificial intelligence (AI), machine learning (ML), deep learning (DL), inference, and high-performance computing (HPC). It excels in industries leveraging GPU computing such as retail, manufacturing, financial services, and healthcare.