Specifications
Model Names
ES200G2 (2U) · ES100G2 (1U)
Form Factor
ES200G2 2U · ES100G2 1U
Processor
5th Gen Intel Xeon Scalable, up to 225 W TDP
Processor Sockets
1 (single socket per node)
Chipset
Intel Emmitsburg PCH C741 with TPM 2.0
GPUs
ES200G2: 2x NVIDIA L40S · ES100G2: 2x NVIDIA L4
Memory
8x DDR5-4400/4800 DIMM slots, up to 128 GB each
Storage — ES200G2
2x M.2 2280/22110 NVMe + 4x U.2 NVMe
Storage — ES100G2
2x M.2 2280/22110 NVMe
Expansion — ES200G2
2x FHFL dual-slot PCIe Gen5 x16 + 1x FHHL PCIe Gen5 x16
Expansion — ES100G2
2x HHHL PCIe Gen5 x16 + 1x FHHL PCIe Gen5 x16
OCP NIC
1x OCP3.0, 10/25/40/50/100/200/400 Gbps
Power Supply
ES200G2 2400 W AC-DC · ES100G2 800 W AC-DC
Cooling
ES200G2 4x 8056 hot-plug dual-rotor N+1 · ES100G2 6x 4056 hot-plug N+1
Dimensions
ES200G2 87 x 438 x 420 mm · ES100G2 43.5 x 438 x 420 mm
Weight
ES200G2 15 kg · ES100G2 10 kg
Front / Rear I/O
Power button with LED, UID button with LED, reset, 3x USB, management port, VGA, COM
Remote Management
IPMI v2.0 compliant with Redfish API
Target Workloads
LLM inference, video analytics, virtual desktop, edge cloud
Overview
Wiwynn's ES200G2 and ES100G2 are single-socket inference nodes built to put NVIDIA data-centre GPUs into a standard 19-inch rack without the power and cooling budget of an 8-GPU training platform. The 2U ES200G2 carries two NVIDIA L40S Tensor Core GPUs at up to 350 W each, while the 1U ES100G2 carries two lower-power NVIDIA L4 cards for density-first inference fleets.
Both nodes accept a single 5th Gen Intel Xeon Scalable processor up to 225 W and expose eight DDR5 DIMM slots running at 4400 or 4800 MT/s with up to 128 GB per module. Networking is handled by an OCP3.0 mezzanine spanning 10G through 400G, which keeps the PCIe Gen5 x16 slots free for GPUs, NVMe retimers or SmartNICs.
Storage differs by height: the 2U ES200G2 supports two M.2 2280/22110 NVMe boot drives plus four front-accessible U.2 NVMe SSDs, while the 1U ES100G2 keeps two M.2 drives. Both are managed through IPMI 2.0 with a Redfish API, so they drop into existing BMC automation.
Key Benefits
Inference density without training-class power: dual L40S in 2U or dual L4 in 1U, both within a single-socket envelope. Networking headroom: OCP3.0 up to 400 Gbps leaves the PCIe Gen5 x16 slots free for accelerators and storage. Serviceability: N+1 hot-plug dual-rotor fans and 2400 W AC-DC supply on the ES200G2, with LED-guided UID and power buttons for blind rack service.
Applications
LLM and diffusion inference at the edge of the data centre, video analytics and transcoding, virtual desktop and cloud gaming infrastructure, CDN and content caching, telco edge cloud, and secondary inference tiers behind an 8-GPU training cluster.
Request a Quote — Wiwynn ES200G2 & ES100G2 AI Inference Servers — 2U / 1U Dual-GPU Nodes
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →