• Home
  • blog
  • Journey to a Silent MicroATX Gamer

Journey to a Silent MicroATX Gamer

blog image

Another new quiet gaming PC build guide, this time featuring a Zotac GTX 970 in the unusual Corsair Air 240 MicroATX case. It is a journey, complete with scientific side trips to explore airflow impedance and its effects on cooling and noise.

This build guide began with a love-at-first-sight reaction to the Corsair Carbide
Air 240 case. The Air 240, with its side-by-side compartments, struck me as
such a clever design that I felt compelled to obtain a sample to build a quiet
gaming PC with it. After several weeks and many long days of hands-on work with
the Air 240, my initial fascination has faded but we do have our first Micro-ATX
Gaming PC Build Guide to present.

SPCR’s Quiet Gaming PC Build Guides is a series designed to bring a new level
of low-noise computing to gaming PC enthusiasts, much like that enjoyed by non-gaming
SPCR enthusiasts for years. Just duplicate our builds and you will enjoy the
same level of ultra low noise even in a powerful gaming PC. The biggest challenge
in building a quiet gaming PC is cooling the very hot components without loud
fans. We bring 12+ years of computer silencing experience to the task. Thus
far, we’ve presented four build guides for PCs that are virtually silent at
low to modest load, and barely go past 20 dBA@1m even at loads higher than any
game can reach.

This MicroATX build doesn’t reach quite as deeply into silence as the others,
but it is still a viable, interesting, very quiet gaming PC. You’ll understand
the reasons for the title when you finish reading the article.

COMPONENT SELECTION

CASE: Many prospects were considered. Some of these will feature in
other SPCR Build Guides in the future.

Silverstone Sugo SG09
is a crafty marriage of breadbox and mini-tower style of just 23 liter volume
that manages to a Micro-ATX board, a full length graphics card, multiple SSDs
and HDDs, an ATX PSU, and a 180mm cooling fan. We’re awaiting a sample of
the Sugo SG10, which has a more attractive brushed aluminum fascia.

Corsair Obsidian 350D, a 42 liter mid-tower style model, offers room
for 240mm watercooling radiators in front and on top and a clean external
design. Its closest competitor here is the Fractal Design Arc Mini R2.

Fractal Design Arc Mini R2 is another large 41 liter mATX case, but
in a more conventional min-tower format. Our
review
showed it to be a very good performer, and though not small, its
form factor makes it suitable for floor placement. Only four expansion slots
makes a quiet dual-VGA card setup challenging, as there’s not enough space
between the cards for adequate airflow, and no room below the bottom slot
for a large VGA cooler. Hence, it is limited to a single VGA card configuration
despite the large size.

In Win / Nofan Dragon Slayer is a slightly smaller (36l) tower-style
MATX case with large expanses of meshed vents on all panels but the right
side. With low noise, well-cooled components, it might house a suitably quiet
gaming rig.

The chosen case this time, the CORSAIR CARBIDE AIR 240, is a cube-style
33-liter micro-ATX model with side-by-side separate thermal zones. There is
no practical limit to PSU size, and room for at least 3 SSDs and 3 HDDs.


The Corsair Carbide Air 240 case is only a foot tall, but also 10.2″
wide.


The rear shot shows why it is so wide: The Air 240 is divided into two
side-by-side compartments, with PSU & drives on one, and motherboard
on the other.

AIR 240 CONSTRAINTS

It’s only after we examined the case sample Corsair shipped that the serious
limitations of this case became clear: The space for the motherboard, CPU &
VGA is only around 5″ wide, 10″ high and 14″ deep (after allowing
for fans). This limits the height of both VGA card and CPU cooler
to 120mm. The CPU cooler height limitation is obvious, but there are many high
performance top-down fan coolers which could be pressed into service. With decent
case airflow, quiet CPU cooling should be assured. But is 12cm for the VGA card
a significant factor? Well it turns out to be significant, indeed.


Air 240 case flipped on its left side: Oh-oh! As the circled area shows,
there is only ~2.5cm of extra height above the add-in card slots.

Reference VGA cards in recent years have all been around 10cm tall, with the
top of the PCB extending less than 1cm past the top of the PCI slot opening
(of any standard computer case). Those cards will all fit in the Air 240 without
issue, though the PCIe 6/8-pin plugs to the cards may get pressed up against
the side cover. Reference high end VGA cards are hardly famous for their low
noise, however, especially as the load goes up.

The non-reference graphics cards with larger heatsinks and fans are needed
for quiet stock cooling. And what is their size? Well, the ones that we’ve been
excited about, the quiet GTX 970 and 980 cards, are all pretty big. In fact,
they are well over 12 cm tall. The PCB of the ASUS Strix GTX 980 protrudes nearly
3.3 cm above the slot opening, and the looping heatpipe puts the total extension
at nearly 4 cm. There is no way to fit this card, nor the Strix 970, nor the
MSI GTX 970 or 980, in the Air 240.


Three high end cards from the last couple years, fitted into a motherboard
with 3 graphics card slots for a close visual comparison: ASUS DirectCU
GTX 680 on left, AMD Radeon R9 290X reference in center, ASUS Strix GTX
980 on right.

The reference card protrudes beyond the slot cover by only 1/2 cm. The
Asus DCU 680 takes a little over 2 cm, the Strix 980 takes nearly 4 cm.

GRAPHICS CARD

The size restrictions led us immediately to considerations of the video card
and the CPU cooler. We had hoped to use the Strix GTX 980 with its proven low
noise performance. That option was out, along with the 970 version. Having examined
images and specs of the MSI GTX 980/970 Gaming series, lauded alongside the
Asus Strix models as being equally quiet, we knew they would not fit in the
Air 240, either. High end AMD cards for this build were not even considered,
given their much higher power profile. Ditto previous gen nVidia cards. The
total space in the Air 240 to house both CPU and GPU is under 15 liters, and
pushing total heat past 300W would not be amenable to quiet cooling.

It was around this time that we took delivery of a Zotac GTX 970. It came in
a big box, but the card turned out to be surprisingly small. The ZT-90101-10P
is very slightly overclocked compared to the reference GTX 970 and uses a much
shorter PCB, but in other specs is almost identical to the reference. There
would be no trouble fitting this Zotac into the Air 240. Of course, we knew
nothing about its acoustic qualities. Given its small heatsink and fans, it
would likely not meet our super-low noise standard at load, so we were prepared
to swap out the cooler for a quieter one.


Zotac 970: Huge box, small card.

A dual-slot cooling design just barely wider than the PCI slot opening.

MOTHERBOARD

There are over 100 socket LGA1150 micro-ATX boards on the market. Select only
Z97 and H97 chip variants and the number pares down to perhaps 30. Chances are,
any of them would do fine, but some have refinements that make silencing and
high power stability a bit easier. Good fan control systems and extensive BIOS
options are foremost. Here’s a small sampling of possible options:

  • ASUS Maximus VII – very high end
  • ASRock Z97M – good value model
  • ASUS Z97M-PLUS – good value model
  • MSI Z97M-G43 – good value model

We went this time with the GIGABYTE Z97MX-Gaming 5, which reputedly
has improved fan control compared to early Gigabyte outings.


Gigabyte Z97MX-Gaming 5 is a full-featured MATX board.

 

CPU

A CORE i5-4690K (3.9 GHz, 4-cores, 88W TDP), used in our first Mini-ITX
Gaming Build, was used for this build as well. For all practical purposes, any
of the following processors provide similar performance. IMO, any number of
earlier i5-3xxx processors would also work fine. In gaming, it’s usually the
GPU which sets performance level.

  • Intel Core i5-4690S (3.9 GHz, 65W TDP)
  • Intel Core i5-4670 (3.8 GHz, 84W TDP)
  • Intel Core i5-4670K (3.8 GHz, 84W TDP, unlocked)
  • Intel Core i5-4670S (3.8 GHz, 65W TDP)
  • Intel Core i5-4570 (3.6 GHz, 84W TDP)
  • Intel Core i5-4570S (3.6 GHz, 84W TDP)
  • Intel Core i5-4590 (3.7 GHz, 84W TDP)
  • Intel Core i5-4590S (3.7 GHz, 65W TDP)

Core i5-4690k mounted in Gigabyte Z97MX-Gaming 5

CPU COOLER

This is somewhat more critical than in many other systems. With the height
restriction, all 120mm fan tower coolers are out of the running. The choice
is then between smaller 92mm fan tower coolers and 92m~120mm fan top-down coolers.
Our instinct favors the latter, but actual experimentation is needed before
we can make the best choice. So a list of CPU coolers to try:





CPU cooler options, clockwise from top left: Noctua NH-D9L, Cryorig
C1, Noctua NH-C14, NZXT X41, Noctua DH-L12, Phanteks TC-12LS
  • Noctua NH-D9L – New dual-tower 92mm heatsink from an industry leader,
    a sample of which just arrived in the lab. Compact, good fit.
  • Cryorig C1 – New 14cm top-down cooler, a bit smaller than the Noctua
    NH-C14, and just 74mm tall with its thin fan. Might be worth trying if the
    C14 does not fit.
  • Noctua NH-C14 – Massive dual-14cm fan top-down cooler, but just 105mm
    tall if only the bottom fan is used.
  • NZXT X41 – It’s a great AIO water cooler though slightly noiser
    than ideal. Keep it in the wings in case it proves necessary.
  • Noctua DH-L12 – Top-down model with both 120mm and 92mm fan and 93mm
    height. It might have too big a footprint and get too close to the back of
    the graphics card.
  • Phanteks TC-12LS – New small, low profile, top-down cooler with 120mm
    fan. Might be too small but would fit easily.

(See our Recommended Heatsinks
article for a complete listing of all the reviewed CPU coolers.)

GPU COOLERS

Continuing with coolers, it is time to consider GPU coolers. It’s almost certain
the little cooler on the Zotac GTX 970 will not be able to cool the GPU quietly
enough in our tight space. A search through our own sample shelves turned up
a dozen or more VGA coolers that we’ve reviewed over the years. Alas, the vast
majority are incompatible with the latest VGA cards, and an even larger percentage
have simply disappeared from the market. There’s no point for us to transform
some old product into a cooling marvel when you can’t get one.

Arctic still offers VGA coolers in many sizes, with mutiple or single or no
fans, in a wide price range. They are very clearly focused on achieving better
cooling and lower noise. Virtually no one else is offering any VGA coolers at
all. The few that do only offer one or two models, generally “dreadnoughts”
that introduce their own size and shape constraints (yes, they make the VGA
card too big to fit in our Corsair Air 240).

So let’s examine some Arctic options, then:


Arctic Accelero coolers, clockwise from top left: Hybrid 2-120, Twin
Turbo 2, Extreme IV and Extreme 3

The above photo was taken long after the system was finalized and tested. The
Accelero Extreme 3 and Accelero Twin Turbo 2 only arrived the day before this
article was posted. Only the Hybrid 2-120 and Extreme IV were on hand when this
system was first being developed. The Twin Turbo II would have been excluded
anyway because its unnecessarily tall dimension does not allow it to fit in
the Air 240.

  • Arctic Accelero Hybrid II-120 – AIO watercooler with 120mm fan and
    big passive backplate heatsink
  • Arctic Accelero Twin Turbo II – Dual-fan cooler
  • Arctic Accelero Extreme IV – Long 3-fan cooler with big passive backplate
    heatsink
  • Arctic Accelero Extreme III – Long 3-fan cooler

Our first choice among these would have been the Extreme III, as it has the
benefit of a huge heatsink and triple fans without the big passive back heatsink
on the Extreme 4 that intrudes into the space for the CPU cooler. It’s also
only a bit wider than the Zotac 970 PCB.

Both Hybrid II-120 and Accelero Extreme IV sport a 1″ tall, large heatsink
that fits on the trace (back) side of the PCB. It is meant to priovide good
cooling for VRM and RAM components on the graphics card, and to add additional
stiffness so the card doesn’t bend under the weight of the heatsink. But the
backside heatsink intrudes into valuable CPU cooler space. You’ll see that we
jumped some hoops to ensure this backside heatsink did not interfere with CPU
cooling.

STORAGE

This is a pure gaming rig; we’re not even going to consider high capacity storage,
you know a NAS works great. Instead, we’ll stay with a half terabyte 2.5 inch
SSD, which is plenty of room for the OS and a handful of even the most space-hogging
games. They start at just $200, so…

  • Samsung 850 EVO 512GB
  • Corsair Neutron XT 480GB

There are probably a few others which could go on the short list. We opted
for the high value, excellent performing CRUCIAL MX100 512GB.


Our review
showed the Crucial MX100 512GB to be a great value.

 

POWER SUPPLY

The Corsair Carbide Air 240 has a separate compartment for the power supply
which usually means a fanless PSU is a viable option. But because the PSU is
positioned sideways, which is not as convection-cooling friendly, a fan-cooled
PSU would be a better option. Knowing that our Silent Mini-ITX Gaming system
with a GTX 980 drew just 270W AC at full load, we knew a quiet 500W model would
suffice, but PSUs under 500W are scarce these days. As usual, the highest efficiency
80+ Gold or Platinum are preferred for their lower thermal contribution as well
as energy savings. Among the models considered:

  • Enermax Revolution X’t ERX430AWT 430W – Looks like a nice 80+ Gold
    unit, despite the bleh name/number. Partially modular output cables.
  • Rosewill Fortress 450W – Possibly a fan-cooled version of the Platinum
    80+ 500W passive I reviewed very positively. All cables attached.
  • Seasonic SSR-450RM Otherwise known as G450, a higher power
    version of the nice G360 I reviewed a while back. Partially modular cables.

We went with a sample already on hand: A SEASONIC SSR-550RM (G550),
somewhat higher power than really needed but close enough.


Seasonic G Series 550W semi-modular 80+ Gold PSU was used.

MEMORY

Precisely what RAM is used as system memory is not critical, although other
web sites have identified DDR1600 to DDR1866 as the sweet spot, somewhat dependent
on the particular game. Within this clock speed range, small variations in timing
have minuscule effect on overall performance. 8GB is more than sufficient for
any single game and general purpose multitasking. 16GB is a waste unless you
have a specific need, and RAM is one of the easiest things to add later to a
system, if you need more for some new application. Two DIMMs are ideal as it
allows for dual-channel operation, while limiting the chances of getting a bad
stick. Memory is one of the most common components to fail over time, so the
fewer the better. We also recommend choosing a brand with a good lifetime warranty
and to avoid models with overly large heatspreaders as they can interfere with
larger CPU coolers. The Kingston HyperX Genesis 2x4GB 1866MHz DDR3
RAM
has been solid for us, and it sports lower profile heatspreaders that don’t
get in the way of big heatsinks.

Alternatives:

  • Kingston HyperX Fury 8GB Kit (2x4GB) 1600MHz DDR3
  • Patriot Viper 8GB Kit (2x4GB) 1866MHz DDR3


HyperX Genesis memory kit.

SYSTEM ASSEMBLY

The fit and finish of the Corsair Air 240 will not win any prizes. It’s good
enough out of the box, but as the side panels are fitted on and off over the
course of a build, they bend and flex easily, and the so-called captive thumbscrews
stop being captive. The bottom line is that sheet metal is a bit thin, the parts
don’t fit perfectly, and screw threads seem coarse and rough. Never mind: Assembly
was easy enough, there were no nasty gotchas, and I didn’t sustain any cuts.


Looking into the main section of the Corsair Air 240 from the bottom
left corner. Two intake and one exhaust 120mm fans are standard.


The upper 3.5″ drive cage on this side needs to be removed to gain
full access to the back of the motherboard. A single thumbscrew does the
job.

The dust filter on the PSU intake is a thin perforated sheet of flexible
plastic with a magnetic perimeter. It works like a charm, peeling off
and back on with ease.

As per our usual build procedure, the CPU was installed with its cooler on
the motherboard, VGA card and SSD plugged in, mounted on an open desktop platform
and powered up with the Seasonic G550. OS installation is almost always done
prior to mounting components in the case; it is easier to access components
if any obvious problems show up at this time.


Here is the Core i5-4690K in the Gigabyte Z97MX Gaming 5 motherboard,
with Kingston HyperX RAM in the slots. Noctua heatsink mounting hardware
was installed; all the Noctua cooler optioned for this build use the same
base mounting hardware, which makes it really easy to swap them out if
necessary.

PRELIMINARY MEASUREMENTS

It is standard SPCR protocol to establish baselines by measuring the noise
of a system before it’s even turned on. Fans, specifically, are run to establish
their range of noise. In this case, since both the VGA and CPU cooling devices/fans
could be altered during development, just the Corsair Air 240 fans were measured,
mounted in their default chassis locations. The fan in the Seasonic G 550 power
supply had zero impact on the measured or heard noise during this test.

Corsair Carbide Air 240
Noise Baseline
All Fans
RPM
SPL @1m
12V
1550
33 dBA
9V
1250
29 dBA
7V
1070
24 dBA
5V
700
18 dBA
4.1V
540
15.5 dBA
Measuring mic positioned 1m at diagonal angle left/front
of case.
Ambient noise level 10~11 dBA@1m.

The three fans sound quite similar to the Corsair Air AF120 Performance Edition
we reviewed in 2013. The differences are small: A little less rubber in the
damping sleeve at the mounting holes, and a rating of 0.3A compared to 0.33A
in the Air AF120. Like its retail cousin, the Air 240 case fan retains a slightly
buzzy quality at higher speed and some rattling/clicking at low speed. It’s
not a great sound fan, but run slow enough, it should be OK.

The minimum 15.5 dBA@1m SPL obtained with all three fans going as slowly as
possible without stalling is a bit daunting. It is higher than we’d like. But
fans can be swapped out if necessary.


The fan used in the Corsair Carbide Air 240 is a slight variant of
the Corsair Air AF120 Performance Edition.

TEST & ANALYSIS TOOLS

The OS used is Microsoft Windows 7 Ultimate, 64-bit. We have yet to be sold
on any version of Windows 8 on a desktop system.

Conditions throughout testing: 21~23°C temperature, 11~12 dBA background
noise.

For the record, CPU and GPU manufacturers have cautioned that Furmark and Prime95
are extreme stress applications that no real software, including games, come
close to emulating. The key issue here is extremely high AND
continuous load. Real games and apps have ups and downs in power
demand. Often not large, but they are there. In contrast, the amount of up and
down we observe with either of these stress programs is typically 1% or less.
We use Furmark because it’s relatively fast and two threads of Prime95 run simultaneously
is a pretty good simulation of how a real game loads a CPU. We have tried playing
games with the systems we build in these guides; they rarely hit more than 85%
of the power load observed under Furmark + Prime95.

CONFIG 1

The first system configuration consisted of the following:

  • Intel Core i5-4690K
  • Gigabyte Z97MX Gaming 5 motherboard
  • Zotac GTX 970
  • Noctua NH-D9L dual-tower 92mm fan
  • Kingston HyperX Genesis 4GB x2 RAM
  • Crucial MX100 512GB SSD
  • Seasonic G 550W

The PCIe power cables to the Zotac GTX 970 card had to be bent over and pushed
back for the side cover to go on. It was a tight fit, with the cables pressing
up against the clear acrylic window.

The stock fans were kept in their original positions: Two intakes in front,
one exhaust on top rear. The fan in the Noctua NH-D9L CPU cooler was oriented
to exhaust its heat up, straight toward the top mounted case exhaust fan.

Total AC power draw with Furmark + Prime95x2 peaked around 270W.

Zotac’s utility for the GTX 970 had very little to offer, so I immediately
tried the now-familiar ASUS GPU Tweak, which thankfully worked to control the
Zotac VGA fans without any problems.

This configuration was a failure. The base SPL was around 20 dBA@1m, and under
load, there was no way to keep the Zotac GTX 970 fans from hitting the nVidia
hard-coded >90°C full speed fail-safe. Not without speeding up the case
fans to the point where the whole system was always too loud; ie, over 30 dBA@1m.
On top of it all, the CPU temperature easily jumped 15°C whever the GPU
was loaded up. The case fans simply can’t seem to move the heat quickly enough
out of the case to keep this from happening.

One AIDA64 sensor screen capture of many similarly poor temperature
results with this configuration. Note the CPU fan RPM; it’s not quiet
at this speed. Ramping the case fans up to full speed was really the only
way to keep both CPU and GPU down.

I tried rotating the case so that it lay on its right side. This
is one of the Corsair-approved positions for the case. There was little if any

difference in either thermal or aural performance. This is confirmation that
the rubber feet affixed on the bottom are tall enough not to cause additional
impedance to airflow.

There seemed to be three interrelated issues:

  • The Zotac GTX 970 cooler can’t cool the GPU at load without running its
    fans at pretty high speed.
  • The case fans can’t evacuate hot air fast enough to keep the CPU from being
    affected by the GPU heat.
  • The CPU cooler can’t cope with both CPU heat and the additional heat
    from the nearby GPU.

Other than the nCase M1, the 15 liter space for the main components in the
Air 240 case is the smallest we’ve tried in our Gaming PC Guide series thus
far. The Rosewill Legacy W1, for example, has a main compartment volume of 23~24
liters, despite its smaller m-ITX only format. In the much smaller space of
the Air 240, I’m trying to quietly cool some 230~240W of heat with smaller heatsinks
and fans. No surprises. Time to bring in bigger guns.

CONFIG 2

Components: Same as Config 1, except Zotac GTX 970 now equipped with
Accelero Extreme III heatsink

The first recipient of a bigger gun was naturally the Zotac GTX 970 card. Removing
the stock heatsink and fans was a tedious job, but not difficult if you’re reasonably
handy and have a good collection of small screwdrivers.


Arctic’s image showing Accelero Extreme IV mounted on a VGA card.


Same backside heatsink is used on the Accelero Hybrid II.

I mentioned earlier that both of the Arctic Accelero coolers on hand —
Hybrid 2-120 and Extreme IV — are equipped with big intrusive heatsinks
that mount on the backside of the VGA card. It’s their way of cooling non-GPU
components on the VGA card and seems effective enough. The backside heatsinks
were not welcome in this build, but they are integral to the mounting system
for both, so I had to find an alternative solution. That solution came in the
form of a backplate from a Scythe Musashi
VGA cooler
, which is long discontinued and not compatible with the newer
graphics cards. It allowed the back heatsink to be left off, but the long thumbscrews
still had to be used, sticking out nearly an inch from the backside of the VGA
card.

I wanted to avoid the potential added noise of the water pump in the Hybrid,
so the Extreme IV was tried first. Mounting the Accelero Extreme IV on the Zotac
GTX 970 was not exactly difficult, but requires patience and a good collection
of small tools. Long needle nose pliers, several small screwdrivers (including
a long one), and bright working lights were all used for the job. Because the
Accelero coolers are designed to fit on a wide variety of VGA cards, the hardware
doesn’t always fit perfectly. You do have to use your eyes and your sense of
feel to figure out just how far to tighten each of the four bolts that clamp
the cooler to the GPU. This is the most critical part of the process, and putting
too much tension in one corner could have nasty consequences, so I took my time
to ensure even pressure all around. The long L-bracket was not used, of course,
as it needs the back heatsink to be installed. Not using the back heatsink basically
turned the Extreme IV into an Extreme III.

The result again were better, especially GPU temperature, but the CPU continued
to run quite hot. The best balance of cooling and noise never got much better
than 27 dBA@1m.

I tried changing up the fans:

  • Original fan config: Two front fans blowing in, one top fan blowing
    out.
  • Fan config b: One front fan blowing in, two top fans blowing out.
  • Fan config c: One front fan blowing out, two top fans blowing in.
  • Fan config d: Two front fans blowing out, one top fan blowing in.

As you might imagine, this was a fair bit of work. Unfortunately, while there
were changes in temperature balance, but none of these changed the overall picture
much.

On a hunch, I tried using a simple baffle of cardboard to keep the GPU heat
from getting to the CPU. This didn’t work either.


The cardboard baffle was a reasonably close fit to separate VGA and
CPU thermal zones. Bottom front fan blew out, and the top two fans above
CPU blew in. This didn’t work well.

 


AIDA64 sensor screen capture of one result with Config 2 + thermal
separation baffle. Increasing the case fan speeds to 8~900 RPM cools the
CPU better with slightly slower fan speed, but the overall SPL does not
improve.

CONFIG 3

Components: Same as Config 1, but Zotac GTX 970 now equipped with Accelero
Hybrid II cooler.

The hope was that watercooling the Zotac GTX 970 would keep its heat from affecting
the CPU and allow the whole system to be cooled better, more quietly. Mounting
the Accelero Hybrid II on the GTX 970 was, again, a job requiring patience,
good small tools and some manual dexterity. The large back heatsink would have
to be used, as the pump/heatblock prevented the use of small aluminum heatsinks
affixed earlier to the RAM chips on the VGA card. This meant the primary PCIe
16X slot could not be used; it would have to go into the 8X slot two spots down.
Not a big deal as PCIe 3.0 8x is not a bottleneck for any games or applications
in this single VGA card system. For more info, see Puget Systems’ investigation
of the
Impact of PCI-E Speed on Gaming Performance
.


The VGA card bends on the left. Clips are normally used to pull the
board closer to the back heatsink. It is is tricky to get the 4-pin fan
plug into the VGA card header next to the power sockets.


Clips affixed on underside. They’re made of somewhat flexible plastic.
Hard to find spots that don’t interfere with motherboard mounting.


This L-bracket is supposed to strengthen and stabilize the card, but it
cannot fit when the VGA card is put in the second PCIe 8X slot of a MicroATX
case.

This is just a DIY building tip: Bamboo skewer used to stick-handle a
fan plug into a hard-to-reach header.


Accelero Hybrid II on Zotac GTX 970, installed in lower PCIe 8X slot.
The radiator is mounted on the lower front panel position, with its fan
blowing out, same as the other front fan. The two top fans were later
installed above the CPU, blowing down. The CPU fan is blowing in the same
direction as the front fans.

After our experience reviewing
the Accelero Hybrid II
, hopes were high, and they were not dashed. Configuration
3 kept the GPU temperature well in hand, not allowing it to go beyond 70~75°C
when its radiator fan was set to run as slowly as 700 RPM. The Arctic fan is
quiet enough than even at 1000 RPM, its contribution is basically inaudible
in the noise of the three Corsair case fans. Strangely, CPU cooling was still
less than ideal, with temperature reaching close to 80°C under either full
Prime95 load or Furmark + P95x2. The best high load SPL achieved was just under
27 dBA@1m; at idle it was 19 dBA@1m. This is not quite up to the standard of
the other Gaming PC Guides SPCR has run this season, but it’s a quiet PC.

Still, annoyed at the CPU temperatures (and the noise level), I decided to
try a CPU cooler change.

CONFIG 4

  • Intel Core i5-4690K
  • Gigabyte Z97MX Gaming 5 motherboard
  • Zotac GTX 970 w/ Arctic Accelero Hybrid II watercooling
  • Noctua NH-C14 14cm fan top-down cooler
  • Kingston HyperX Genesis 4GB x2 RAM
  • Crucial MX100 512GB SSD
  • Seasonic G 550W

This was a relatively painless change, as the Noctua heatsinks share the same
base mounting hardware. One cooler was easily swapped for the other.


The Noctua NH-C14 was used only with one fan mounted under the fin stack,
blowing down. It only fit with the Arctic Accelero Hybrid II because the
VGA card was in a lower PCIe slot. Top fans blowing in, front fans blowing
out.

This config finally produced results I’d been expecting long before, with both
GPU and CPU temperatures at just 71°C after half an hour of Furmark + Prime95x2
pulling 270W AC. The final SPL: 25 dBA@1m in full load; 18 dBA@1m in idle.


A bit of Furmark captured with AIDA64 report.

A WEEK OF PERCOLATION

It’s nice to call it R & R, but in reality I was simply too busy to start
writing this article till almost a week after all the system tweaking and testing.

After I’d written much of the above, something about the whole build was bugging
me. Why didn’t the system do better? Surely, with three 120mm case fans, top
notch coolers on both the CPU and the GPU, and all those vents everywhere, the
system in this Corsair Air 240 could do better!? Why were those fans so ineffective?

It was while discussing these questions that Larry Lee, longtime SPCR lab worker
and reviewer, noticed the plastic mesh dust filter lying on my desk. I’d pulled
it off the side panel of the Rosewill W1-S case, a move that increased the airflow
to the ASUS Strix GTX 980 card in our first Mini-ITX
Gaming Build
and dropped the noise level to just 13/20 dBA idle/load. The
sight of this filter naturally led to the question, if removing that filter
was productive, how about the dust filters in the Corsair Air 240?

Aside from the magnetic perforated film dust filter on the PSU side, I had
not considered the dust filters on the fan locations in this case, which are
all on the panels covering the other compartment where the motherboard, CPU
and VGA are housed. I’d noted that they were there but not thought much beyond
that. They are not meant to be removed. How much airflow do they impede? Suddenly
this seemed a very worthwhile question.

I stopped writing and went back to investigative mode. A couple of hours later,
I had some interesting answers.

DUST FILTERS

Higher priced computer cases started to sport dust filters over intake vents
about a decade ago. The idea was simple enough: Keep dust from accumulating
on the fans and cooling fins to maintain good cooling. This was particularly
important in systems where the fans run at high speed constantly; a description
which applies to a lot of systems back in those days. The fast spinning fans
naturally suck up a lot of dust and deposit them all over the case, especially
on heatsinks fins and fan blades. Of course, the dust filters themselves get
choked up and if they’re not regularly cleaned, the end result is perhaps even
worse than no filter: No airflow.

Today, almost every aftermarket case is fitted with dust filters, often regardless
of whether the vent is meant for intake or exhaust. This is pertinent: You don’t
need a dust filter on an exhaust fan; that just impedes airflow for no good
reason. Dust filters on the best cases now are easily accessible so they can
be readily and frequently cleaned. It looks mostly like progress.

But dust filters, like protective fan grills, exact a price on airflow. Just
how much depends on the exact design of the filter, and whether this matters
depends on how important that lost airflow might be. In our Rosewill W1-S Gaming
PC, removing that side vent dust filter stopped the GPU from reaching the 90°C
point where its fans go berserk. (True of all GTX 970s and 980s, AFAIK.) So
the airflow lost to the that filter was pretty important.

I took a close look at the Corsair Air 240 filters. First, the removeable
magnetic sheet on the PSU side.


You might recall this photo from many pages back: I thought it was the
cat’s meow. Looking more closely…


…I’m not so sure. The perforations look pretty small compared to the
amount of plastic around them.

The dust filters for the left compartment of the Corsair Carbide Air 240 are
integrated into the plastic frames that are the front, top and bottom panels.
A row of thick plastic columns are on the outside, followed by a perforated
metal sheet with 3 mm hexagonal holes surrounded by 1mm of material. The final
layer is the dust filter, and it is exactly the same material as the magnetic
filter on the PSU vent.


The Air 240 vent grills, viewed from the inside. There are multiple perimeter
tabs that keep the plastic dust filter material in place. (Click pic to
see large version.)


From left: Air 240 top panel with filter removed; the removed Air 240
filter; the filter from Rosewill W1-S case; stock Air 240 bottom panel.
(Click pic to see large version.)

The second photo above is instructive. The top and bottom panels of the Air
240 are identical. You can see the white (OK, off-white; it’s old & beat
up) of the vinyl tile floor through the top panel without the dust filter. You
can’t see the floor at all through the stock bottom panel on the far right.
The difference between the Corsair and the Rosewill filters in the center is
self-evident as well.

I would guess that the Corsair Air 240 dust filters block at least 50% of the
light.

What about airflow? 🙂 It so happens that SPCR Labs is equipped to test this
scientifically.

EFFECT OF DUST FILTERS ON AIRFLOW

Our Kanomax 6803
Anemometer
does not get much mention in SPCR, and only brief use in
extended fan tests. This is an expensive ($845 today) instrument of high precision
able to read as little as 40 FPM air velocity to as high as 7800 FPM with 1%
accuracy. I spent an hour with the Kanomax 6803, the mounted fans in the Air
240 case, and the removed dust filters. The results are discussed in the captions
for the revealing photos below.


The vane of the anenometer was placed over the top fan, set up as an intake.
Note that the dust filter was removed from this panel. The linear velocity
of the air reads 254 Feet Per Minute (from which CFM is calculated).


The removed filter was placed atop the fan. FPM dropped to just 159. That’s
a 38% drop. Or if you start with the filter on, a 60% increase when the
filter is removed.


The anenometer was placed at the front vent, whose filter was also removed.
The fan is blowing out, and air speed measures 300 FPM.


With the filter in place, 139 FPM. That’s a loss of 53% with the filter,
or if you start with the filter in place, a gain of 116%.


The filter from the Rosewill W1-S case in place of the Air 240 filter:
The air speed drops only by a third instead of more than half.

One interesting aspect of these experiments is that I could easily hear
the effect of the filters on fan speed: It actually went to a lower pitch and
got slightly rougher sounding, with more turbulence. It was clear that the fan
speed was reduced by the filter.

All this tells us that when used on the intake side of the fan, Corsair’s dust
filter blocks about 40% of the airflow. Used on the exhaust side, it blocks
over 50%. With this solid new evidence of just how restrictive the dust filters
in the Corsair Air 240 are, there’s no getting around my need to revisit this
build with all filters removed. Yes, PC silencing is an iterative process.

CONFIG 5 – NO DUST FILTERS!

  • Intel Core i5-4690K
  • Gigabyte Z97MX Gaming 5 motherboard
  • Zotac GTX 970 w/ Arctic Accelero Extreme III
  • Noctua NH-D9L twin-tower + second Noctua NF-A9 PWM fan
  • Kingston HyperX Genesis 4GB x2 RAM
  • Crucial MX100 512GB SSD
  • Seasonic G 550W

I did make changes to the components, namely, the Accelero Extreme III on the
Zotac GTX 970 and the Noctual DH-D9L with two in-line fans. By this time, the
original system had been long pulled apart, so changing the components didn’t
take any more effort.

The first change was to see if the reduced impedance of the bottom intake vent
would help VGA cooling. The Accelero Extreme III has three fans which draw directly
from the bottom vents. The absence of the radiator (which is part of the Hybrid
II) would help keep in/out airflow impedance to a minimum. And besides, the
Extreme III seemed a little easier to mount than the Hybrid II.

The CPU cooler change was made because I like the idea of CPU fans blowing
the heat out of the case, rather than a top-down style cooler which blows the
heat all around for case fans to then push out. With a second 92mm fan in push-pull
mode, I figured the two fin stacks would work most effectively.

Again, installing the Accelero Extreme III was mainly fussy, but not “hard”.
Getting even pressure on the GPU was the essential task. Despite the inclusion
of many different height spacers, you need to pay close attention to ensure
even, tight pressure on the GPU.


Accelero Extreme II installation, viewed from top edge of VGA card.


Config 5 without dust filters, all components installed, with arrows showing
airflow. Top fan blows down on the Noctua CPU cooler, whose two fans get
also air from the open back panel vents & blow towards the front.
The retail Noctua NF-A9 PWM fan package includes a handy 4-pin Y-splitter
cable, used to keep both fans on the CPU fan header of the Gigabyte motherboard.
Accelero Extreme III fans pull air from bottom vent and blow up. Front
fans blow out: When gaming, the area in front of the PC does get warm.
Note small piece of black colored close-cell foam positioned between the
second and third VGA fan. This was stuck in place with a couple drops
of super glue. It heps to keep the VGA card from drooping on that side
with the weight, reducing the pressure on the PCIe slot. If you do something
similar, make sure no fan blades get stopped.


The front portion of the top vents were blocked off with a piece of heavy
paper. This is to keep airflow in control. Ideally, the front fans should
not pull air in from those top vents, which would be an airflow short
circuit that bypasses the CPU heatsink.

Control of the Arctic VGA cooler was an issue. I’d been using the ASUS GPU
Tweak
utility for GPU fan control before, but the lowest speed that would
allow is 35%. With the fans on the Arctic cooler, 35% is very smooth and broadband,
but far too loud, in the mid-20s dBA@1m. The small 4-pin plug on the Arctic
cooler has four conductors for its fans, like the original Zotac cooler, but
for some reason, none of the montoring software picked up the RPM.

Since GPU Tweak could not get me to a lower speed, I uninstalled it and tried
MSI RIVA Tuner. It got to the Arctic cooler to 28% fan speed but this
was still far too noisy.

Stymied, I resorted to the 4-pin Molex to 7V/12V adapter Arctic supplies with
its VGA coolers. This connects the cooler fans directly to the PSU for either
12V or 7V operation. It’s a pretty basic fan control in this age of uber-sophisticated
software utilities. I didn’t even try the 12V plug, going straight to 7V.

This was the right setting. I could tell the SPL was no higher than 20 dBA@1m,
and it was a smooth sound, far better than any of the earlier iterations of
this build. Those filters definitely exact an acoustic price.

After playing around with the Advanced Smart Fan section in Gigabyte’s System
Information Utility while keeping the system under high load (Furmark + Prime95x2),
I settled on a couple of relatively simple custom profiles for the CPU fan and
the three case fans.


Custom CPU fan profile.


Custom profiles were similar for all three case fans.

With these settings, after half an hour of Furmark + Prime 95×2, the CPU was
stable at 70°C, the GPU at 83°C. The latter was much higher than with
the Accelero Hybrid II-120, but the noise was far lower. All the other temperatures
reported by system sensors showed nothing amiss. The measured SPL was 19 dBA@1m
at this point. A close look and listen confirmed that the last fan in the system,
the one in the Seasonic G550 power supply, might be spinning slightly faster
than at idle, but at a low enough level that its acoustic contribution was moot.

Thinking through the system details, I wondered again if the bottom feet gave
enough clearance for good unimpeded airflow to the VGA fans. You may recall
I turned the system on its side so the bottom intakes were facing left in an
earlier iteration, and with the dust filter in place, almost nothing changed.
This time, instead of putting the system on its side, I propped up the front
with a small block of wood, adding a little more clearance between intake vent
and the desktop. The effect, this time, was almost immediate, with GPU temperature
starting to drop the moment this was done. Five minutes later, the GPU temperature
had dropped 7°C, and this also helped the CPU, which dropped 3°C. All
this is summarized in the following images and tables.


Filterless Air 240 system propped up on a block of wood for a bit more
bottom intake clearance.



There’s barely any difference in measured or perceived noise between idle
and full load.


AUDIO RECORDING

This recording was made with a high resolution, lab quality, digital recording
system inside SPCR’s own 11 dBA ambient anechoic chamber, then converted to
LAME 128kbps encoded MP3s. We’ve listened long and hard to ensure there is no
audible degradation from the original 24/88 WAV files to these MP3s. They represent
a quick snapshot of what we heard during the review.

The recording starts with 7 seconds of ambient noise, then ~10 second segments
of the system at various states. For the most realistic
results, set the volume so that the starting ambient level is just barely audible,
then don’t change the volume setting again while listening.

JOURNEY’S END

To those who read through my entire journey, thanks for persevering; I hope
it was worth your time. To those who jumped straight to this page for the snapshot
conclusions, it’s possible that some of what I write won’t jibe fully till you
head back and read the article all the way through.

This Corsair Air 240 based Quiet Micro-ATX Gaming PC is a bit different from
the other gaming systems we’ve detailed this season. It’s base SPL is several
decibels higher than the others, most of which approach the ambient levels of
the quietest homes. This one is more audible, especially as it is meant to be
placed atop the desk. The noise is mostly broadband, as you’ll have heard in
the audio recording, and it hardly gets any louder even when subject to extended
artificially high load. My final measurements indicate just 1 dBA@1m difference
between idle and load SPL. Some users will find this more constant noise preferable
to a machine that’s quieter at idle but ramps up to a more audible level under
load. It’s often the change many of us notice most.

The Zotac GTX 970 graphics card and the Intel i5-4690K make this a very capable
gaming rig. The performance of the GTX 970 is only a half step behind the much
pricier 980 and the quad-core CPU hardly imposes any limits with today’s most
demanding games.

SPCR’s Quiet MicroATX Gaming PC Component List
SPCR Build Components
Street Price
Alternatives
Corsair Carbide Air 240
$85
Fractal Design Arc Mini R2
Corsair Obsidian 350D
Silverstone Sugo SG10
In Win / Nofan Dragon Slayer
Zotac GTX 970
ZT-90101-10P
$330
PNY VCGGTX9704XPB
Gigabyte GV-N970IXOC-4GD
Gigabyte GV-N970WF3OC-4GD
EVGA 04G-2974-KR
Arctic Accelero Extreme III
$65
Arctic Accelero Hybrid II-120
Gigabyte Z97MX-Gaming 5
$130
ASRock Z97M
ASUS Z97M-PLUS
MSI Z97M-G43
Intel Core i5-4690K
$190
Core i5-4690S
Core i5-4670
Core i5-4670K
Core i5-4570
Core i5-4590
Noctua NH-D9L
w/ 2nd NF-A9 PWM fan
$50
Noctua NH-U9S
NZXT X41
Noctua DH-L12
Noctua NH-C14
Crucial MX100 512GB SSD
$200
Samsung 850 EVO 512GB
Corsair Neutron XT 480GB
Seasonic G550
SSR-550RM
$85
Rosewill Fortress 450W
Enermax Revolution X’t 430W
Seasonic G450 SSR-450RM
Kingston HyperX Genesis 2x4GB 1866MHz
DDR3
$95
Kingston HyperX Fury 8GB Kit (2x4GB) 1600MHz
DDR3
Patriot Viper 8GB Kit (2x4GB) 1866MHz DDR3
TOTAL
$1230
NOTES
* The alternate GTX 970 cards listed all feature narrow
coolers and PCBs that should fit the confines of the Corsair Air 240.
* Noctua DH-U9S will work best with a second 92mm fan.
* Other case alternatives don’t have the same VGA card size restriction.
Retail prices are subject to constant fluctuations.
Please use the shopping links to check on current pricing; don’t rely
on the prices cited in non-linked text.

This quiet gaming rig would not have been possible without the power efficient
performance of the GTX 970 and the Intel Core i5-4690K. The peak power draw
of 267W AC is probably a hundred watts less than what a similar performing system
with a last gen video card would have required. The reduced heat is key in allowing
this system to be so quiet.

The other components all played their part, of course: The flexible and stable
Gigabyte Z97MX-Gaming 5 motherboard, the quiet Seasonic G550 PSU, the Kingston
RAM, the Crucial MX100 512GB SSD. Most crucial in this build is the Arctic Accelero
Extreme II: Not only is it a highly effective VGA cooler, its footprint is narrow
enough to fit in the Air 240. The Noctua NH-D9L
with its extra fan for push-pull was also critical in keeping the CPU cooled
quietly enough, although the alternative options would probably done similarly
well.

The most compromised component in this build turns out ironically to be the
one that inspired it. The Corsair Carbide Air 240 is a fascinating concept poorly
executed. There are three main issues:

  • Basic build quality is mediocre at best. The side panels — nay, all
    the panels! — are too thin and flexible and too many parts fit sloppily.
    Make it better and charge a bit more!
  • The dust filters built into the top, front and bottom panel vents are far
    too restrictive of airflow. This might be less of an issue if you’re running
    case fans at full speed, but 50% less airflow has got to be significant for
    anyone seeking good cooling, whether for an overclocked rig or a low-airflow
    stealth machine. This flaw should be easy for Corsair to correct.
  • The case probably should not be made much wider if it’s going to fit comfortably
    on many desks, but the internal width of the main compartment is unfortunately
    narrow.

The best and best cooler-equipped VGA cards are all too wide to fit in the
main compartment. You simply cannot fit any of the recent generation ASUS Strix
GTX 980/970 or MSI Gaming GTX 980/970 cards in the Air 240; these are reputedly
the quietest graphics cards on the market. If you want a GTX 980 in this case,
it must be one fitted with a narrower cooler, which almost assuredly will have
smaller fans and be too noisy for a quiet rig. All this could be avoided if
the main compartment of the Air 240 was 1.5 cm or maybe even just 1 cm wider.

Could the right side compartment have been squeezed one centimeter narrower
and the extra space given to the left side? Judging by the space on either side
of the vertically poisitioned PSU, I believe so. A half cm wider overall dimension
would have made no change desktop usability, either.

Lest this article end on a negative note, I draw your attention to the results
of the dust filter and fan airflow experiments conducted in the course of this
build. It’s positive reinforcement of a basic tenet for PC silencing SPCR has
espoused for years: Minimize impedances for best airflow and low noise. Yes,
dust filters can be convenient and useful, but if you’re seeking to eke out
the best airflow with the lowest noise, they’re not in your best interest. Just
accept the need to be a bit more vigilant about dusting and know that you have
to clean dust filters just as often anyway.

* * *

This is certainly my last article for the week, so have a Merry Christmas!
(What? I dare mention the word? How un-PC! Mmmmm…. How commonsense! It’s what
the day is called, OK?)

This concludes another SPCR Silent Gaming PC Build Guide. Please support SPCR
and help us present many more build guides by using our sponsor advertising
links for your shopping.

Many thanks to Intel,
Corsair, Arctic,
Crucial, Gigabyte,
Kingston, Seasonic
and Noctua for sponsoring
the components in this build guide.

* * *

Articles of Related Interest
Quiet Mini-ITX Gamer
Build Guide

Quiet ATX Gamer, R5 Version
SPCR’s Quiet ATX
Gaming Build Guide

Silent Mid Gaming PC Build Guide

Case
Basics & Recommendations

Recommended Heatsinks

* * *

Discuss
this article in the SPCR Forums
.

Leave a Comment

Your email address will not be published. Required fields are marked *