GIGABYTE’s R282-Z90 is a 2U rackmount basic goal server for mainstream computing and HPC workloads, virtualization (VMware and Citrix licensed) and distributed storage options.
The R282-Z90 options twin processor sockets for 2nd technology AMD EPYC™ 7002 / 7003 sequence processors for as much as 128 cores (256 threads) per system, 32 DIMM slots for as much as 4TB of DDR4 ECC reminiscence (or as much as to 8TB utilizing 3DS LRDIMM modules), with an M.2 storage slot and a entrance/rear storage capability of 12 x 3.5″ and a pair of x 2.5″ SATA / SAS hot-swap HDD/SSD areas and as much as to eight x PCIe Gen 4.0 slots and a pair of x OCP 2.0/3.0 slots to accommodate a lot of choices for add-on playing cards/gadgets resembling HBA, SmartNIC, FPGA and so forth. There’s additionally an optionally available NVMe equipment to transform 4 of the 12 3.5-inch SATA / SAS drive bays to NVMe-capable bays (PCIe Gen4 x4 bandwidth per drive).
Most significantly, the specs described above are all designed with a balanced NUMA system to make sure optimum efficiency and cut back efficiency bottlenecks in demanding workloads.
For instance, when constructing distributed storage options, the R282-Z90 is a perfect alternative, given the numerous PCIe slots, a balanced NUMA structure, and huge native storage capability (NVMe assist accessible), which all assist to stability loading. consumer requests and storage IOPS all through, offers quick cache pace and achieves excessive availability within the occasion of a community or storage material connection failure. Extending with JBOF operating NVMeoF might be simply configured, for instance, to offer large-scale parallel file techniques for AI and HPC use circumstances.
Equally, when designing a virtualization structure, the R282-Z90 offers the flexibility to allocate the VM based mostly on CPU affinity, particularly when the extra gadgets on the NUMA area are totally different (resembling utilizing two totally different HBA protocols ). NUMA’s balanced structure additionally avoids P2P CPU bottlenecks that may trigger vital latency to the end-user expertise. The R282-Z90’s massive native space for storing helps reduce the price of utilizing exterior storage gadgets and enlargement controller boards, particularly when contemplating software-defined storage virtualization.
For integrating compute and HPC workloads, the R282-Z90 was examined on the SPEC CPU® 2017 benchmark and the SPECjbb® 2015 benchmark and broke a number of world information. The outcomes communicate for themselves:
The R282-Z90’s versatility of use and its system structure design, confirmed by world record-breaking benchmark outcomes, make it a dependable alternative for system efficiency, ease of system integration, and adaptability in addressing varied utilization situations.
