Gigabyte G893-ZD1-AAX3

0.00$

Subtotal:

Add-ons total:

Total:

Вес: 0 kg
скачать как CSV

Technical Specifications

Form Factor and Dimensions
8U rackmount, 447 × 351 × 923 mm

Processors
Dual AMD EPYC 9005 / 9004 (Socket SP5, LGA 6096)
cTDP up to 500 W per CPU
System-on-Chip design, no separate chipset

Memory
24 DDR5 RDIMM slots
Up to 12 channels per CPU
Supported speeds: up to 6400 MT/s (EPYC 9005), up to 4800 MT/s (EPYC 9004)

GPU Support
NVIDIA HGX H200 platform with 8 GPUs (SXM form factor)
NVLink / NVSwitch interconnect with up to 900 GB/s bandwidth

Expansion and PCIe
4 FHHL x16 Gen5 slots (PEX89104)
8 FHHL x16 Gen5 slots (PEX89104)
4 FHHL x16 Gen5 slots (CPBG045)

Storage
8 front 2.5″ hot-swap NVMe bays (Gen5)
2 internal M.2 slots: 1 × PCIe Gen3 x4 (CPU_1) and 1 × PCIe Gen3 x1 (CPU_0)

Networking and Management
CFPG540 I/O module with 2 × 10 GbE LAN (Intel X710-AT2) and 1 × management port (RJ-45)
Additional rear MLAN port
BMC: ASPEED AST2600
Remote management through GIGABYTE Management Console with KVM over IP, monitoring, and power control

Power Supply
4 + 4 redundant 3000 W power supplies, 80 PLUS Titanium certified
Supports 5+3 redundancy at 30 °C
Input voltages: 115–127 V AC, 200–220 V AC, 220–240 V AC, and 240 V DC

Cooling
Onboard fans: 2 × 60 × 60 × 56 mm and 4 × 60 × 60 × 76 mm
PCIe zone fans: 4 × 80 × 80 × 56 mm
GPU tray fans: 16 × 80 × 80 × 80 mm
Operating temperature: 10 °C – 30 °C
Operating humidity: 8% – 80% (non-condensing)

Operating and Storage Conditions
Non-operating temperature: –40 °C to 60 °C
Non-operating humidity: 20% – 95%

Included Items
G893-ZD1-AAX3 chassis
2 CPU heatsinks
Rack mounting kit
Fan modules and blanking plates

Advantages and Applications

Maximum GPU density with up to eight high-end accelerators in a single chassis
PCIe Gen5, NVLink/NVSwitch, and wide memory bandwidth for optimal throughput
Redundant power and advanced cooling ensure reliability
Remote management with full KVM over IP support
Ideal for AI training and inference, HPC clusters, scientific simulations, and data center GPU workloads