Home/Products/Datacenter servers
InInventec

Datacenter-Class Rack Servers

Inventec rack-scale platforms for AI training, HPC and cloud — NVIDIA Grace, HGX and Intel Xeon 6 systems from one of the world's largest server ODMs.

K709G6
Inventec · 2U1N · NVIDIA MGX

K709G6

NVIDIA Grace CPU Superchip server

An NVIDIA MGX-compliant 2U node pairing the 144-core Grace CPU Superchip with up to four double-width GPUs — high-throughput HPC and cloud at lower power.

144 coresNVIDIA Grace Superchip
4× GPUdual-slot Gen5
960 GBLPDDR5X
Form factor2U1N, NVIDIA MGX architecture
ProcessorNVIDIA Grace CPU Superchip, up to 144 cores @ 3.2GHz
Memory480GB or 960GB LPDDR5X, up to 4266 MT/s
GPUUp to 4 dual-width Gen5 FHFL GPUs, NVLink support
StorageUp to 12 U.2 NVMe SSDs
NetworkingOCP 3.0 SFF slot · BMC management
PowerCRPS 1+1 redundant, 100–240V AC / 240V HVDC
Ideal forHPCCloudAI inference at scale
K900G7
Inventec · 2U4N

K900G7

High-density Intel Xeon 6 HPC server

Four Xeon 6 nodes in 2U for dense compute — DDR5-6400, Gen5 NVMe and up to 2,400W CRPS power.

4 nodesin 2U
350WXeon 6 per node
Gen5E1.S / U.2 NVMe
Form factor2U4N
ProcessorIntel Xeon 6, up to 350W TDP
Memory8 DDR5 DIMMs per node @ 6400 MT/s (1DPC) or 16 @ 5200 MT/s (2DPC)
Storage4× E1.S, or 4× U.2 + 2× E1.S NVMe — PCIe Gen5 x4 per drive
NetworkingOCP 3.0 · AST2600 BMC
Power / coolingCRPS up to 2,400W · designed for 35°C ambient
Ideal forHPC clustersSimulationDense compute
P9000G7

Representative image — final chassis may differ.

Inventec · 10U

P9000G7

NVIDIA HGX B300 AI training system

Next-generation AI infrastructure: NVIDIA HGX B300 with dual Xeon 6 for training and fine-tuning foundation and robotics models.

HGX B300NVIDIA baseboard
2× Xeon 6host CPUs
8× U.2NVMe
Form factor10U, air-cooled
GPUNVIDIA HGX B300
ProcessorDual Intel Xeon 6 (P-core / E-core)
MemoryDDR5, 6400 MT/s @ 1DPC · 5200 MT/s @ 2DPC
StorageUp to 8 U.2 NVMe SSDs
Ideal forFoundation-model trainingRobotics simulationFine-tuning

Need a different configuration?

Memory, storage, wireless, OS images and mounting can all be configured to your spec — we'll confirm lead time and pricing.