Inventec Artemis II Rack vs Lenovo Lenovo NVIDIA GB300 NVL72

Both are built on the NVIDIA GB300 NVL72 baseboard, so their accelerator performance is identical rather than merely similar. The choice between them is chassis, cooling, expansion and supplier, not FLOPS.

What actually differs

SpecificationInventec Artemis II RackLenovo Lenovo NVIDIA GB300 NVL72
VendorInventecLenovo
Part number—7DJVCTO2WW
Form factorrackrack
Released——
Accelerator72x GB30072x GB300
Aggregate memory——
Aggregate bandwidth——
System power136.0 kW135.0 kW
InterconnectNVLink across all 72 GPUs over passive copper cable cartridges, with nine 1RU NVLink switch trays; scale-out is ConnectX-8 at 800 Gb/s InfiniBand east-west, carried on two I/O boards holding two SuperNICs each, plus a BlueField-3 DPU at 400 Gb/s north-southFifth-generation NVLink across all 72 GPUs through nine switch trays, joined by a rear cable cartridge backplane; the NVLink domain is contained within one rack, with no inter-rack scaling, and each compute tray adds up to three PCIe Gen 5 x16 slots for networking

Rows shown in grey are identical between the two.

Inventec Artemis II Rack

The Blackwell Ultra generation of Inventec's Artemis rack, built on the NVIDIA GB300 NVL72 and the MGX platform, again pairing 36 Grace CPUs with 72 Blackwell GPUs across eighteen compute trays and nine NVLink switch trays in a 48RU rack. The change from the first Artemis is in the network: east-west moves from ConnectX-7 at 400 Gb/s to ConnectX-8 at 800, carried on two dedicated I/O boards with two SuperNICs apiece, so back-end bandwidth keeps pace with the GPUs. Cooling narrows to row-based and sidecar, with leakage detection, and Inventec rates the rack at a maximum of 136 kW. No throughput or memory figures are published. This entry links to no reference design because the only GB300 NVL72 record in this catalogue is NVIDIA's own DGX appliance, which is a finished competing product rather than a reference platform.

Vendor product page

Lenovo Lenovo NVIDIA GB300 NVL72

Lenovo's build of NVIDIA's GB300 NVL72, a 48U MGX rack holding 72 Blackwell Ultra GPUs and 36 Grace CPUs across eighteen 1U compute trays, with nine switch trays and either six or eight power shelves joined by a rear cable cartridge backplane. Two rack base options exist, C5RK with non-redundant power and six shelves and C5RJ with redundant power and eight; they differ only in power, not in compute. Cooling is deliberately hybrid rather than fully liquid: CPUs, GPUs and the NVSwitches are liquid cooled while the OSFP modules, drives and power distribution board are air cooled, which Lenovo puts at roughly 90 percent of rack heat captured to liquid and 10 percent to air, and which lets the whole thing drop into a standard 19 inch cabinet. Each rack consumes 135 kW of thermal design power, rising to about 155 kW at peak depending on workload. Each compute tray carries two Grace processors of 72 Arm Neoverse V2 cores with up to 960 GB of LPDDR5x between them, up to eight E1.S NVMe drives and an M.2 boot device. Lenovo publishes no accelerator memory total and no throughput figure; the memory it does state is the Grace side rather than HBM.

Vendor product page

© 2026 Flopper.io - Compare the hardware powering AI