Supermicro Liquid Cooling Solutions
June 2019
Data center cooling solutions are used for the maintenance of optimal
operating conditions required for the smooth operation of data centers. As
data centers process massive amounts of data heat is a directly associated
byproduct. The majority of power usage in a data center is associated
with the cooling necessary to maintain optimum operating conditions.
Liquid cooling is an approach that tends to be more ecient than standard
air cooling. Liquid cooling solutions signicantly reduce data center
power consumption and Power Usage Eectiveness (PUE). Additionally,
because liquid cooling is extremely ecient, data center can increase their
processing per square foot beyond what is possible when using forced
air cooling, increasing data center capacity. Most notably, liquid cooling
is an absolute necessity for organizations interested in trying to capitalize
on the rise of AI where GPUs often run at higher thermal requirements
than traditional CPUs. Many of today’s latest computationally intensive
workloads require the highest CPU and GPU performance possible, while at
the same time utilizing higher rack density – a recipe for overheating.
Supermicro oers liquid cooled systems in complete rack-based solutions
including cooling equipment from partners. The warm-water cooling
solution enables high-density server racks with high performance systems
together with a reduction in PUE. Overall, liquid cooling can reduce TCO by
as much as 40-50% when compared to forced-air cooling. Without need to
provision power for fans at maximum speed, system fans run at a minimum
and result in whisper-quiet operation. The in-rack Cooling Distribution Unit
(CDU) includes heat exchangers which cool the liquid loops connected to
each server through a Direct to Chip (D2C) cooler. Each D2C is an integrated
pump and cold plate directly mounted on each CPU and GPU. Each rack is
connected to facility-provided water where heat is expelled through 1.25”
exible hoses. The CDU exchanges heat across two closed systems that
never mix, the server coolant and the facilities water. As an integrated rack
system, the Supermicro Liquid Cooled Rack oers 40kW and 80kW CDU
capacities to accommodate Cooling Distribution Modules (CDM) with 76 or
108 loops to support a combination of Supermicro BigTwin
™
, TwinPro
™
, Ultra
and GPU SuperServers.
https://www.supermicro.com/en/solutions/high-performance-computing
SMCI SKU CDU capacity CDM Hose Pair Rack Size
SRS-DCLC-KIT-02* 40kW Up to 76 nodes/loops 1.5m 42U
SRS-DCLC-KIT-03* 80kW Up to 108 nodes/loops 3.5m 48U
* Please consult with Supermicro to determine the optimum CDU/CDM capacity and conguration
The information contained in this document is subject to change without notice.
Supermicro, the Supermicro logo, Building Block Solutions, We Keep IT Green, SuperServer, Twin, BigTwin, TwinPro, TwinPro², SuperDoctor are trademarks and/or registered trademarks of Super Micro Computer, Inc.
Ultrabook, Celeron, Celeron Inside, Core Inside, Intel, Intel Logo, Intel Atom, Intel Atom Inside, Intel Core, Intel Inside, Intel Inside Logo, Intel vPro, Itanium, Itanium Inside, Pentium, Pentium Inside, vPro Inside, Xeon,
Xeon Phi, Optane, 3D XPoint and Xeon Inside are trademarks of Intel Corporation in the U.S. and/or other countries.
All other brands names and trademarks are the property of their respective owners.
© Copyright 2019 Super Micro Computer, Inc. All rights reserved.
Printed in USA
Please Recycle 01_Liquid-Cooling_Rev05
Supermicro Liquid Cooling Solutions
Ultra
Supermicro Ultra SuperServers are designed to deliver the highest
performance, exibility, scalability and serviceability to demanding IT
environments. 1U and 2U Ultra systems support dual 2nd generation Intel
®
Xeon
®
Scalable processors, up to 24 DIMMs of DDR4 memory and with a
variety of Ultra Riser options oer built-in 1G, 10G and 25G Ethernet. Multiple
high performance NVMe storage congurations are supported as well as
options for SAS/SATA and hybrid NVMe drive bays. A D2C cooler is mounted
on each of the processors within the Ultra system and routed through a CDM
loop to the Liquid Cooling CDU.
GPU
Supermicro GPU systems are at the heart of today’s AI and HPC excitement by
combining the fastest processors, memory and GPUs together in a family of
systems for AI/ML, Inferencing and HPC. The 1U GPU system supports up to 4
NVIDIA
®
Tesla
®
V100 GPUs together with NVLink
®
, the highest 2nd generation
Intel
®
Xeon
®
Scalable processors, and up to 12 DIMMs of DDR4 memory
providing an extremely compact and powerful Learning or HPC system. D2C
coolers are mounted on each of the processors and GPUs within the GPU
system and routed through CDM loops to the Liquid Cooling CDU.
BiGtwin
™
The Supermicro BigTwin
™
represents agship performance for the most
demanding applications and HCI environments. The innovative design
supports up to four nodes in a 2U enclosure with no-compromise support
for processors, memory and I/O. Each node can support dual 2nd generation
Intel
®
Xeon
®
Scalable processors, up to 24 DIMMs of DDR4 memory and up to
six high speed NVMe drives. SIOM networking options include 10GbE, 25GbE,
100GbE and IB. Shared power and cooling maximize the resource-savings of
the multi-node design. D2C coolers are mounted on the processors within
each BigTwin node and routed through a CDM loop to the Liquid Cooling CDU.
twinPro
™
The Supermicro TwinPro
™
architecture is based on the Supermicro proven Twin
technology in 1U and 2U form factors, supporting two or four nodes. Each
node can support dual 2nd generation Intel
®
Xeon
®
Scalable processors, up
to 16 DIMMs of DDR4 memory and hot-swappable 3.5” or 2.5” SAS3/SATA3
storage options. SIOM networking options include 10GbE, 25GbE, 100GbE and
IB. D2C coolers are mounted on the processors within each TwinPro node and
routed through a CDM loop to the Liquid Cooling CDU.