AceleMax ICG-160
1U 6x GPU Density Optimized Server for Immersion Cooling
Designed for liquid immersion and offers the highest GPU density configuration available today. The server design reduces server power usage by up to 20% in submerged environments, and has six enterprise grade GPU accelerator cards on a single 1U chassis.
Ideal for data center applications including:
- Artificial Intelligence (AI)
- High-Performance Computing (HPC)
- Geographic Analysis
- High-Frequency Trading (HFT)
- Edge Computing
Reference #: Q710040
Processor
Single AMD EPYC™ 7002 series processor, 7nm, Socket SP3, up to 64 cores, 128 threads
Memory
- 16 x DDR4 DIMM slots
- Up to 2,048GB RDIMM/ 1,024GB LRDIMM/ 2,048GB LRDIMM 3DS
- Memory speed up to 3200 MHz
Graphics Processing Unit (GPU):
Up to 6 x single-wide or 4 x double-wide GPU cards in a 1U immersion cooling chassis
Expansion Slots
- 6 x PCIe Gen 3 x16 slots for GPUs
- 1 x FH/HL PCIe Gen 3 x16
Network Controller
- 2 x 10GBASE-T ports
- 1 x 10/100/1000 management LAN
Storage
2x M.2 PCIe SSD
Chassis
- 1U rack-optimized chassis
- Modified power supply: 1600 watt redundant, 80 PLUS Platinum
Supported Immersion Cooling Rack: GRC Chassis
- Power dissipation per rack: 25kW – 100kW
- Number of U per rack: 42U – 52U
Immersion Cooling Rack includes:
- Rack(s) filled with liquid immersion coolant
- Coolant distribution unit (CDU)
- Monitoring and alerts application
- Mirrored System, with 2N redundant pumps
- Integrated cable management
- Rack-mounted service rails for easy server maintenance and hot swaps
Immersion Cooling Solution Value Proposition:
- Cut cooling energy by up to 95%
- Attain an mPUE <1.05
- Lower upfront costs 50%
- Reduce server power draw 10-20%
- Cool up to 200 kW in a single rack
- Enables high-density data center layouts
- CPU/GPU-agnostic
- Fast deployment: 8-12 weeks
- Enables Scaling in 1 to 4 rack increments
Pushing Operational Limits and Efficiencies with Liquid Immersion Cooling
As hardware density and costs continue to rise, liquid cooling is transforming the datacenter industry, pushing the operational limits of both CPU- and GPU-based servers while roughly halving the construction costs of traditional data centers.
To read more, download solution brief: