Previous Page  30 / 52 Next Page
Information
Show Menu
Previous Page 30 / 52 Next Page
Page Background

28

Chip Scale Review November • December • 2018

[ChipScaleReview.com]

versatility of 3D-printing, the design

can be mod if ied at any moment to

accommodate the chip. This ensures

t h a t t h e l o c a t i o n o f t h e n o z z l e s

coincides with the hottest sites on the

chip. Whenever a chip with a different

layout needs cooling, the design can be

easily adapted and the cooler reprinted.

At the same time, 3D-printing also

po s e s s pe c i f ic de sig n cha l le nge s .

There are, first, restrictions on nozzle

diameters and other small feat u res

such as the cavit y height. Pr i nti ng

techniques like digital light processing

o r s t e r eol it hog r aphy le ave exc e s s

liquid resin after curing of the polymer

that needs to be removed f rom the

i nt e r na l cav it ie s. The refore, re si n

removal places constraints on nozzle

diameters and plenum thickness. A

s e cond , concomit a nt ch a l le nge i s

t he d i f f icu lt y t o che ck i f i nt e r na l

structures are free from resin, because

the cooler is printed as one part. For

this application, imec demonstrated

that the scanning acoustic microscopy

t e c h n i q u e (SAM) c a n b e u s e d t o

check for potentially blocked resin

inside the nozzles. Finally, nozzle side

walls should be suff iciently st rong

to prevent the wall from collapsing,

which could result in a “liquid short-

circuit” between the inlet and outlet

f low. All structures should therefore

be self-supporting.

The cooler put to the test

To test the design experimentally,

imec fabricated an impingement cooler

for an 8x8mm

2

test ch ip. Th is test

chip contains a number of heaters and

sensors, making it possible to create a

power dissipation pattern and measure

the temperat u re in more than 1000

points over the entire chip surface.

It was used to locally measu re the

temperat ure effect at the inlets and

outlets and evaluate the quality of the

cooler. Furthermore, the chip allows for

an assessment of the ideal number of

nozzles. For this, 3x3 up to 8x8 nozzle

arrays were printed, resulting in nozzle

diameters ranging from 300 to 800μm.

The best thermal performance was

noted for the cooler with the highest

number of nozzles (8x8), while the

3x3 nozzle ar ray per formed worse.

Moreover, compared to a single-jet

cooler with 1 inlet for the same chip,

a multi-jet impingement cooler has a

lower thermal resistance and a better

temperature uniformity for the same

f low r at e. Al l pr i nt ed noz z le s a re

within 5 to 10% of the nominal design

d i ame t e r s . Sy s t ema t i c d ev i a t ion s

– either smaller or larger – can be

compensated for in the design. Finally,

the comparison between the measured

temperature map and simulation results

with fabrication parameters show good

agreement (

Figure 6

).

How cool is the chip cooler?

To put the cooler into perspective, it

was compared to published data of other

impingement coolers in terms of thermal

resistance and pump power. For this,

coolers fabricated in different materials

were considered: Si, ceramic, metal and

plastic. The imec cooler displays a very

good thermal performance of 0.13cm

2

K/

W for the 8x8 nozzle array at a 1000ml/

minute flow rate and a 0.3bar pressure

drop between the inlet and the outlet

of the cooler. In other words, with this

cooler, the chip will only heat up 13°C

for each 100W/cm

2

, the typical power

density for a standard processor, where

the limit is around 100°C. With this

performance, cooling of 500W/cm

2

would

be feasible, making it one of the best cooler

performances in literature.

Only Si-based coolers with µm-pitch

nozzles outperform the imec cooler.

However, t hose coolers need up to

five times more pump power and are

considerably more expensive. At the end

of the day, the price tag is an important

benchmarking criterium. Making the

choice for a certain fabrication technique

or material is based on the goal to reach as

much cooling power per pump power. But

it also determines the price of the cooler,

which makes energy costs for data centers

enormous.

Summary

This benchmark ing st udy clearly

s hows t h a t i t i s no t ne c e s s a r y t o

a g g r e s s i v e l y s c a l e d ow n n o z z l e

diameters to a few tens of microns, but

that very good thermal performance

can also be ach ieved wit h nozzles

i n t he r ange of a few hund red s of

micrometers. At these sizes, cheaper

polymer fabrication technologies, such

as 3D-printing, can be used. Moreover,

the simulations indicate that the thermal

conductivity of the cooler material has

no impact on thermal performance of

the impingement cooler. Therefore,

despite being made out of insulating

materials, the polymer coolers achieve

s i m i l a r c o o l i n g p e r f o r ma n c e a s

coolers in materials with good thermal

conductivity. This makes the 3D-printed

liquid jet impingement cooler a very

cos t- ef fe c t ive a lt e r nat ive t o mor e

expensive coolers for data centers to use.

Biographies

Herman Oprins received his MSc and

PhD degrees in Mechanical Engineering

f r om KU Leuve n , Belg ium. He i s

a Senior Researcher at imec; email

Herman.Oprins@imec.be

Tiwei Wei received his MSc degree

in Optoelectronic Engineering from

Chongqi ng U., Chi na. He is a PhD

candidate at KU Leuven and imec.

V l a d imi r Che r ma n r e c e ived h i s

MSc and PhD degrees in Electronic

Engineer ing f rom Saint-Petersbu rg

Elect rotechnical U., St. Petersburg,

Russia . He is a Sen ior Re sea rche r

at imec.

Eric Beyne received his MSc degree in

Electrical Engineering and PhD degree

in Applied Sciences from KU Leuven,

Belgium. He is a Fellow and Program

Director at imec.

Figure 6:

a) (top) Model and measurement of the

8x8mm test chip thermal resistance for a b) (bottom)

4x4 array impingement cooler with 600ml/min flow rate.