The Effect of CPU Location in Total Immersion of Microelectronics
Click here to load reader
Transcript of The Effect of CPU Location in Total Immersion of Microelectronics
![Page 1: The Effect of CPU Location in Total Immersion of Microelectronics](https://reader038.fdocuments.in/reader038/viewer/2022100514/577c79731a28abe05492b16a/html5/thumbnails/1.jpg)
7/26/2019 The Effect of CPU Location in Total Immersion of Microelectronics
http://slidepdf.com/reader/full/the-effect-of-cpu-location-in-total-immersion-of-microelectronics 1/4
Abstract — Meeting the growth in demand for digital servicessuch as social media, telecommunications, and business and cloudservices requires large scale data centres, which has led to an increasein their end use energy demand. Generally, over 30% of data centre
power is consumed by the necessary cooling overhead. Thus energycan be reduced by improving the cooling efficiency. Air and liquidcan both be used as cooling media for the data centre. Traditionaldata centre cooling systems use air, however liquid is recognised as a
promising method that can handle the more densely packed data
centres. Liquid cooling can be classified into three methods; rack heatexchanger, on-chip heat exchanger and full immersion of themicroelectronics. This study quantifies the improvements of heattransfer specifically for the case of immersed microelectronics byvarying the CPU and heat sink location. Immersion of the server isachieved by filling the gap between the microelectronics and a water
jacket with a dielectric liquid which convects the heat from the CPUto the water jacket on the opposite side. Heat transfer is governed bytwo physical mechanisms, which is natural convection for the fixedenclosure filled with dielectric liquid and forced convection for thewater that is pumped through the water jacket. The model in thisstudy is validated with published numerical and experimental workand shows good agreement with previous work. The results show thatthe heat transfer performance and Nusselt number (Nu) is improved
by 89% by placing the CPU and heat sink on the bottom of the
microelectronics enclosure.
Keywords — CPU location, data centre cooling, heat sink inenclosures, Immersed microelectronics, turbulent natural convectionin enclosures.
I. I NTRODUCTION
HE wide spread growth of data centers in recent years isresulting in an increase in end use energy demand of
which more than 30% is required for their cooling [1].Therefore more efficient cooling methods are essential, butmost data centres use air cooling techniques. However,approaches that use liquids offer greater efficiency of heatremoval due to their density and are readily adopted insituations of high density Datacom equipment. The revivalfrom the 1960s [2] of liquid-cooled datacom equipment bringswater from the edge of the data centre to the racks or bringswater into the rack to enable liquids to get closer to the mainsources of heat dissipation. There are different types of liquidcooled microelectronics which bring water into the rack or viawater jackets in contact with a dielectrically immersedmicroelectronics enclosure.
Heat transfer in total immersion of microelectronics is
A. Almaneea (corresponding author), N. Kapur, J. L. Summers and H. M.Thompson are with the Institute of Engineering Thermofluids (iTF), School ofMechanical Engineering, University of Leeds, United Kingdom (e-mail:[email protected]).
achieved by natural convection of the dielectric liquid, wherethe heat is transferred from all of the microelectroniccomponents, including the CPUs, to the water jacket on theopposing surface. This study develops a theoretical method forquantifying the improvements of heat transfer when the heatsink and CPU vertical location is varied as shown in Fig. 1.
The analysis presented in this paper is based on naturalconvection of liquid in an enclosed cavity, where there is heatflux into and out of the cavity on each vertical wall; heat is
transferred by natural convection, without the use of pumps.Previously, [3] established the first theoretical benchmark forheat transfer via laminar flow in a square enclosure where theleft wall is cold and the right wall is hot. Tian [4] performed
benchmark experiments for the turbulent flow in a squareenclosure and he plot thermal counter and vector for first timein turbulent enclosure. In the case of rectangular enclosures,[5] has investigated the fluid behaviour in both cases oflaminar and turbulent flows, where it was found that for aspectratios between 1 and 40 and Prandtl numbers between 1 and20 the fluid flow becomes turbulent when the Rayleighnumber, Ra>10 7. For the localised heat source in an enclosure[6] numerically investigated a vertical rectangular cavity withthree heat sources on one wall and the opposing wall ismaintained at a cold temperature. This study determinedtheoretically the optimum spacing between the heat sourcesand reported spacing arrangements that can drop thetemperature by 10%. Heat sources in enclosures have beeninvestigated by many further studies [7]-[9]. However, the useof fins to increase surface area has been the focus of muchfewer studies. Nada [10] carried out an experimental study onfinned vertical rectangular enclosures. In this case horizontalfins were adopted on the hot wall side and the opposing wallwas cold. This study also investigated the effect of differentRayleigh numbers (Ra) and different fin lengths and spacing.
It was found that increasing the fin length increased the Nusselt number and improves the heat transfer.In the present work a model of turbulent natural convection
flow and conjugate heat transfer is used to investigatevertically placed microelectronics in a sealed enclosure filledwith dielectric liquid. The model geometry and water jacketinlet temperature remain constant. The CPU heat flux andlocation are varied with the aim of quantifying heat transferimprovements based on heat sink and CPU location thatreduce the CPU case temperature.
II. THE MODEL DESCRIPTION
The geometry and boundary conditions of the liquidimmersed microelectronics model is shown in Fig. 1. Heat is
The Effect of CPU Location in Total Immersion of
MicroelectronicsA. Almaneea, N. Kapur, J. L. Summers, H. M. Thompson
T
World Academy of Science, Engineering and TechnologyInternational Journal of Electrical, Computer, Energetic, Electronic and Communication Engineering Vol:9, No:4, 2015
431International Scholarly and Scientific Research & Innovation 9(4) 2015 scholar.waset.org/1999.5/10000942
I n t e r n a t i o n a l S c i e n c e I n d e x , M e c h a n i c a l a n d
M e c h a t r o n i c s E n g i n e e r i n g
V o l : 9 ,
N o : 4 ,
2 0 1 5 w a s e t . o r g / P u b l i c a t i o n / 1 0 0 0 0 9 4 2
![Page 2: The Effect of CPU Location in Total Immersion of Microelectronics](https://reader038.fdocuments.in/reader038/viewer/2022100514/577c79731a28abe05492b16a/html5/thumbnails/2.jpg)
7/26/2019 The Effect of CPU Location in Total Immersion of Microelectronics
http://slidepdf.com/reader/full/the-effect-of-cpu-location-in-total-immersion-of-microelectronics 2/4
g bthte
pl
Ce3htoTh
ndthBflFca
wa(7reca
nerated fromthe water th
e water jac perature 30
anes for the
MSOL 4.3. perimental w0mm x 200t wall. The b the temperate average ere and the ex
Fig. 1 The gCOMSOL (
In this study,mely dielecte to naturalrough channefore solvingws are checr the waterlculating the
ere D is theerage flow v.96x10 -4 Pa.spectively. Tn therefore b
the CPU unat passes thr et at the fl K. The conjimmersed li
The simulatiork for naturam x 40mm eoyancy drivere differenceror betweenerimental wo
eometry of thea) Isometric vi
III. MATHE
the model coic liquid thaconvection
ls to carry ththe full conjed to determiflowing in thReynolds nu
water channelocity (1.2x1) are the de Re is 53 for considered a
erneath the hugh the soli
ow rate 3x1ugate heat mquid approac
n model is vl heat convecclosure with
n flow inside between thethe simulatirk in [10] is 4
ymmetry simuw (b) Side vie
MATICAL MO
nsists of twot circulates iand the wate heat awayugate heatne whether te channel thi ber, Re
μ
l diameter (3.0-2 m/s), ρ w (ensity anda flow rate o
s laminar [11]
eat sink and block repres0-7 m3/s andodel with syh is simulat
alidated agaiion in [10],fins attachedthe enclosurehot and coldn model pre.8%.
lation model us (c) Front Vie
EL
types of fluiside the enc
er that is pfrom the enclodel, the twe flow is tur s is determi
.5x10 -3 m), V96.6 kg/m 3)iscosity of
f 1.15x10 -7 m.
ooledenting
inletmetryd via
st thehere ato theis duewalls.sented
ed in
flow,losurempedosure.
fluidulent.ed by
is thend μ w water,/s and
dielthe
whgrais tvisexpChiwhiis t
S
I
byenedo
ρ
Forwhi
wh
Iturbcanasintr k, aequ
or natural cectric fluid flRayleigh nu
re H is the eity, c p is spe
he dielectricosity, β isansion. All tet al [13] anch is greaterrbulent [5].
W
Properties
pecific heat capaThermal expansiDynamic viscoshermal conducti
side the enclthe conserv
rgy for turbulain.ontinuity equ
omentum eq
. .p
he body forcan increase i
ch creates the
re g = (0, 0,he dielectric
turbulentulent modeloffer solutioinvestigated
oduces two and specific diations are bas
ρ .
ρ . ω
onvection inw behaviour
ber [12], Ra
nclosure heiific heat capthermal conthe dielectriese fluid the listed in Tabhan 10 7 and
TAORKING FLUID T
Ab
cityontyvity
osure the fluition equatioent natural co
ation
.ρuation
μ μ
, F , dependsn temperatur driving buoy
).luid has dens
ρ(T) = 17
atural convhas been fous that are cloin [14]-[16]
dditional varissipation rate,ed on [17], w
.μ μ
.μ μ
ide the sealcan be indica
μ
ht, g is the acity, ρ D is diductivity, μ D
c’s coefficiermal propertile I. The Ra indicates ther
LE IERMAL PROPER
brev.
C p Β 1µD 1λ D
flow is turbs for mass,
nvection of t
0
23μ
mainly on thee there is a dancy force, e
ρT
ity-temperatu
16.2 -2.2T
ction in end to be a ro
se to experim. The k-ωables; turbule ω. The two aich are:
∗ k p
ω α
ed enclosure,ed by consid
cceleration dlectric densit is the diel
t of volues are taken
equal to 1.8fore that the
IES
Dielectric liquid
1140 (J/kg.K).151496x10 -3 (K -1
.124782x10 -3 (Pa.6.9x10 -2 (W/m.K
lent and gov momentume dielectric li
. 23ρk
density variaecrease in de pressed as:
e variations
losures, theust model
ental observaturbulent
nt kinetic endditional tran
∗
p
thering
e to, λ D
ctricetricfrom108,
flow
1 ).s)
rnedand
quid
tion.nsity
k-ωhichtionsodelrgy,
sport
World Academy of Science, Engineering and TechnologyInternational Journal of Electrical, Computer, Energetic, Electronic and Communication Engineering Vol:9, No:4, 2015
432International Scholarly and Scientific Research & Innovation 9(4) 2015 scholar.waset.org/1999.5/10000942
I n t e r n a t i o n a l S c i e n c e I n d e x , M e c h a n i c a l a n d
M e c h a t r o n i c s E n g i n e e r i n g
V o l : 9 ,
N o : 4 ,
2 0 1 5 w a s e t . o r g / P u b l i c a t i o n / 1 0 0 0 0 9 4 2
![Page 3: The Effect of CPU Location in Total Immersion of Microelectronics](https://reader038.fdocuments.in/reader038/viewer/2022100514/577c79731a28abe05492b16a/html5/thumbnails/3.jpg)
7/26/2019 The Effect of CPU Location in Total Immersion of Microelectronics
http://slidepdf.com/reader/full/the-effect-of-cpu-location-in-total-immersion-of-microelectronics 3/4
wn
d
wusth
li pmfoth
I
Cilothd
The turbulen
The producti
p μ The energy e
ere λ D is thember based oThe empirica
1325, ∗ 12 For the pumain, the goContinuity e
Momentum
ρ . Energy equa
In the solid d
ere, the λ s ise propertiese water jackeThe workinguid (Table Isses throughicroelectronicr which the te density whi
. EFFECT O
The followiU and heatmersed micr
cations are ve top to the bfined as (S)
viscosity ca
μD
on term is fou
: quation woul
ρc T .mal conductin the Kays-Cl turbulent m
, 12, ped water cerning equatiuation
.quation
.pion
ρcomain, only t
0 the thermalf copper forand its fins.fluids in this
). The waterthe channel
s enclosure ihermal propech is a functio
CPU LOCATI
g investigatisink locationelectronics eried as show
ottom of thehere the valu
be defined a
nd from the fl
23 .
be written a
λ c ity and Pr T iawford equatdel constants
9125, ∗ 910annel which
ons are:
0
μ
.λ T e energy equ
.λ T conductivitythe heat sink
paper are wis used as as of the wa filled withties are keptn of temperat
ON ON HEAT
on determine on the heatnclosure. The in Fig. 2. Tnclosure andes are listed i
uid velocity a
23ρk .
:
Ts a turbulention [17]. parameters a
, 0.41,is a lamina
ation is requi
or the solidsand alumini
ater and a diecooling liquiter jacket ahe dielectricconstant excre.
RANSFER R E
s the effecttransfer with
CPU and hehe variation ithe CPU loca Table II.
s
randtl
e
5.2
fluid
ed
whichm for
lectricd thatd theliquid pt for
SULTS
of thein thet sinkfrom
tion is
prefrofluwer hig
occmm
byrestowhea
whithe
prewatdielencfluicolthe
pladrifluirole
lar recithesm
Witreci
Fig. 2 CPU loc
CaseCaseCase
he effect ofented in Fig. 70 to 100
load q L 100e kept constaest Nu impli
urs when thewhich is for
89% betweenectively. Thiard the bottot transfer perf s the dielec
ch drives theenclosure itence of the ter jacket anectric liquidlosure. Thisd and hence
water jacketflow domaine. Fig. 4 illen eddy exisd below thein the globalhen the CP
e section of trculation sitsfluid beneathll surface ar
h the CPUrculation fills
ations, where S
TACPU L
1
the CPU an. 3. The heat
(where heat%) while thet at 3x10 -7 m
es the best h
CPU is locatcase 3. Fig. 3 case 1 andindicates tha of the encl
ormance.tric liquid is buoyancy inis directed
op wall wher d then itsto travel bauoyancy for transfers the. The locatio over whichstrates thising at the lePU is not h
flow.is located to
e channel do between theit hardly mova on the co
located at th the entire ca
is varying as s
LE IICATION
S (mm)
300200100
d heat sink lflux load (q
flux q is 888flow rate an3/s and 303eat transfer p
d toward theshows that t
3, where S=t the CPU anosure has gr
heated itsuced circulatowards thee it transfersdensity increk down to te creates theheat from th of the CPUmeaningfulost clearlyel of the CPated and the
wards the toes not experie
PU and the tes. Consequel wall for r
e bottom ofvity and the
own in Table I
ocation onL) is also va8.88 W/m 2 at
inlet temper respectively.erformance a
bottom at S=e Nu is incr
300 and 100heat sink lo
ater effect o
ensity decreion. At the told plate bheat to the cases causinghe bottom ocirculation o heat sink tictates the sieat transferith the buoy
U and above.efore does pl
of the channce any flowop of the waltly there is omoving the
the channel,ffective area
I
u isyingheat
atureThe
nd it
100asedmmatedthe
ases; p of
theolerthethe
f thethe
e ofakesancyTheay a
el, atheandly a
heat.
theover
World Academy of Science, Engineering and TechnologyInternational Journal of Electrical, Computer, Energetic, Electronic and Communication Engineering Vol:9, No:4, 2015
433International Scholarly and Scientific Research & Innovation 9(4) 2015 scholar.waset.org/1999.5/10000942
I n t e r n a t i o n a l S c i e n c e I n d e x , M e c h a n i c a l a n d
M e c h a t r o n i c s E n g i n e e r i n g
V o l : 9 ,
N o : 4 ,
2 0 1 5 w a s e t . o r g / P u b l i c a t i o n / 1 0 0 0 0 9 4 2
![Page 4: The Effect of CPU Location in Total Immersion of Microelectronics](https://reader038.fdocuments.in/reader038/viewer/2022100514/577c79731a28abe05492b16a/html5/thumbnails/4.jpg)
7/26/2019 The Effect of CPU Location in Total Immersion of Microelectronics
http://slidepdf.com/reader/full/the-effect-of-cpu-location-in-total-immersion-of-microelectronics 4/4
wlaco
t
icoismtuinTein
siC
th
ich meaninggest – henceoler when it
Fig. 3 Varia
Fig. 4 Effectifferent cases.ransfer perfor
3x10 -7
This papermersed micnvection andsimulated us
icroelectronicrbulent and tside the watee model set
perimental re good agreemDifferent cask location tU and heat
e Nu number
ful heat transthe temperat
irst meets the
ion of heat flux
of the CPU lochese velocityance when them3/s and water
V. CO
as investigatoelectronicsconjugate heing COMSOs enclosure,e k- turbuler channel is
up was cosults and shoent with previes were modeat can yield
sink arrange
by 89%.
er takes placre of the die
CPU.
load for differ
ation on the velrofiles are sele
heat flux loadinlet temperat
NCLUSION
d heat transin a dielectt transfer for
v4.3. Sincethe dielectr
nce model isaminar beca pared withed that the
ous work [10led to establithe best heaent was able
on the coollectric fluid
ent CPU locati
ocity for the thcted for the bes 100%, flow r re is 303 k
er of enclosric liquid.a vertical enc the Ra>10 7
c fluid regiemployed. Thse Re is ver
previous pubodel in this s.h the CPU at transfer. Thto show a b
all isill be
n
eet heatate is
d andaturallosurein theme ise fluid
low.lishedudy is
d heate bestost in
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
Shah, A., et alcooling ensembElectronic Systeon. 2008: IEEE.Anderson, D.,model 91: MacResearch and Dde Vahl Davis,mark numericalFluids, 1983. 3(Tian, Y. and T.filled squareInternational Jo866.MacGregor, R.,and high Prandt1969. 91: p. 391Phan-Thien, Y.mounted on aTransfer: Part AHeindel, T., Sconvection froTransfer, Part AKeyhani, M., Lconvection in aTransfer (TransEngineers), SeriWroblewski, D.mounted protrugeometry and116(1): p. 112-1
Nada, S., Natuclosed narrowInternational Jo679.Çengel, Y.A., R fluid sciences 2Holman, J., Hea
Chi, Y.Q., Jona Nik Kapur andEnclosed, ImmThermal Measu2014, IEEE.Zitzmann, T., eCFD. in Proc.Simulation 200Rundle, C. andin a square cavifluid flow inConference, CalAounallah, M.,convection inInternational Jo1693.Wilcox, D.C.,industries La Ca
R EFE
. Impact of racle. in Thermalms, 2008. ITHE . Sparacio, andine philosophy avelopment, 196
G., Natural convsolution. Interna): p. 249-264.
Karayiannis, Loavity: part I:
urnal of Heat a
Free convectionl number fluids.-403.., Nhan, An opvertical substr
: Applications, 2. Ramadhyani,
an array of pr Applications, 19. Chen, and D.
enclosure withctions of the A
es C);(United St and Y. Joshi, Lsion in a three-
boundary condit19.al convection h
enclosures withurnal of Heat a
.H. Turner, and J08. p. 534.
t transfer, 9th. 20
than Summers,Harvey Thompsrsed, Direct Li
rement and Man
al. Simulationof the 9th Inte. 2005: Montréal
. Lightstone. Vty for applicatioatria geometriegary. 2007.
et al., Numerian inclined sqrnal of Heat an
Turbulence monada, CA.
ENCES
-level compactiand Thermomec
M 2008. 11th I
. M. Tomasulo,nd instruction-ha. 11(1): p. 8-24.ction of air in aional Journal for
turbulence natuthe thermal ad Mass Transfer
through verticalrans. ASME, Jo
imum spacing pte in an enclo00. 37(6): p. 61
and F. Incropetruding heat so
96. 29(1): p. 1-1itts, The aspectrotruding heat sME (Americantes), 1991. 113(4quid immersiondimensional encions. Journal o
eat transfer inheated rectangud Mass Transfer
.M. Cimbala, Fu
02, McGraw-Hil
eter Hopton, Ken, Case Study ouid-Cooled Ser
agement Sympos
f steady-state narnational IBPSA: IBPSA.alidation of turbu of CFD modelli. in 2nd Cana
cal investigatioare cavity wit Mass Transfer,
eling for CFD.
n on the dataanical Phenometersociety Conf
The IBM Systedling. IBM Jour
square cavity: a Numerical Meth
ral convection ind fluid flow, 2000. 43(6): p.
plane layers-mournal of Heat Tr
roblem for threeure. Numerical-630.a, Conjugaterces. Numerical.ratio effect onurces. Journal o
Society of Mech).cooling of a sublosure: the effe heat transfer,
orizontal and vlar finned base, 2007. 50(3): p.
damentals of th
. p. 335-337.
th Deakin, Alanf a Data Centreer, in Semiconium (SEMI-TH
tural convectionConference Bu
lent natural convng to heat transf dian Solar Bui
of turbulenta hot wavy
2007. 50(9): p.
Vol. 2. 1998:
centerna inrence
/360nal of
benchods in
an airields.
. 849-
deratensfer,
chipsHeat
aturalHeat
aturalHeat
anical
trate-ts of1994.
rtical plate.. 667-
rmal-
Real,UsinguctorRM).
usingilding
ectionr and
ldings
aturalwall.
1683-
DCW
World Academy of Science, Engineering and TechnologyInternational Journal of Electrical, Computer, Energetic, Electronic and Communication Engineering Vol:9, No:4, 2015
434International Scholarly and Scientific Research & Innovation 9(4) 2015 scholar.waset.org/1999.5/10000942
I n t e r n a t i o n a l S c i e n c e I n d e x , M e c h a n i c a l a n d
M e c h a t r o n i c s E n g i n e e r i n g
V o l : 9 ,
N o : 4 ,
2 0 1 5 w a s e t . o r g / P u b l i c a t i o n / 1 0 0 0 0 9 4 2