The first of our quiet gaming build guides is an ATX tower featuring the highly efficient NVIDIA GeForce GTX 980. Join us as we plan, build, and stress test this beast, while trying to keep it quiet enough to satisfy our own high standards. A preview: We managed to keep it under 20 dBA@1m under all test conditions!
November 19, 2014 by Lawrence Lee with Mike Chin
November marks the onset of winter and the holidays for much of the northern hemisphere, but it’s a special time for PC enthusiasts as well. With Black Friday fast approaching it’s the perfect time to shop for components, whether for an upgrade or a complete system. PC gamers in particular have it the best of all because Autumn is also traditionally the season for major game releases. So far we’ve seen the launch of the heavily anticipated Call of Duty: Advanced Warfare, Far Cry 4, and Dragon Age: Inquisition, among many others. To play these titles smoothly and with all the eye candy enabled, you’ll need some serious hardware, especially if you plan on using a >1080p display, like a QHD or 4K model.
To help you do just that, we walk you go through the process of planning and
building a mid/high-end gaming PC. You can find build guides on countless review
sites, but being SPCR, we approach it with a heavy emphasis on noise-reduction.
We explain in detail why we selected components (and suggest alternatives),
then actually assemble the system (which most sites rarely do,
creating just a shopping list instead) and conduct a full battery of thermal
and acoustic tests. And if there are any tweaks or tricks to make the experience
quieter, we share those, too. We have several different builds in the works,
but we start off with the most common, popular one, the mid-ATX tower.
COMPONENT SELECTION
GPU: ASUS
STRIX GTX980
$580
Alternatives:
- ASUS
STRIX GTX 970 – US$380 - MSI
GTX 970 – US$350
The graphics card is the heart and soul of any gaming PC but it’s also the
biggest challenge from an acoustic perspective. The dies of modern high-end
GPUs are larger, denser, and more complex than their CPU counterparts, which
translates into substantially higher power consumption. Making things worse,
video cards are usually oriented horizontally with the cooler underneath blowing
upwards so the card’s PCB itself gets in the way of natural convection. Some
models use a blower style fan to push the hot air out, which is great for keeping
system temperatures lower but it’s often a noisier method that is not as efficient
for cooling the GPU itself. It would be great if you could mount a giant tower
heatsink like one for a CPU, but the VGA card isn’t designed to take that much
weight and such a solution would block off all the expansion slots underneath.
Having an inherently energy efficient card would be a huge advantage for a quiet
PC.
Cue NVIDIA with their new Maxwell 2 GPU architecture, which happens to be miraculously
energy efficient compared to the previous generations. The GeForce GTX 970 has
a TDP of just 145W, half that of its AMD equivalent, the Radeon R9 290X. The
higher-tier GTX 980, arguably the fastest single GPU video card on the market
today, uses only 20W more. The GTX 980 is a rather expensive card, starting
at US$550 while only offering a 10~20% performance advantage over the much cheaper
GTX 970. Clearly, the GTX 970 offers better value, but if you’re driving a UHD/4K
display it might be worthwhile to go with the 980. To duplicate or improve on
the 980’s performance you’d need a pair of cards in SLI/CrossFire which would
create substantially more heat and noise. We also prefer the challenge of the
hotter card.
|
|
For this system we chose the well-regarded ASUS Strix GTX 980. At a glance,
the stock cooler doesn’t seem that formidable (some GTX 980 models have three
fans) but it’s a DirectCU heatsink with five nickel-plated copper heatpipes
of varied thickness (the largest being 10 mm) that contact the GPU directly
rather than utilizing a baseplate to act as a go-between. According to ASUS,
the fans don’t even spin unless the GPU is heavily taxed and even some lighter
games can be played with zero noise emitted. Furthermore, the card employs a
10 phase power delivery system, making it overclocking-friendly.
|
|
The ASUS Strix GTX 980 is long enough to extend further than the width of a
typical ATX motherboard. The Strix version’s PCB measures 26.7 cm across, with
the cooler hanging over a bit for a total length 28.8 cm or 11.3 inches. Most
enthusiast cases are designed to accommodate cards this long but tighter fitting
towers with non-removable hard drive bays may be incompatible. A ventilated
metal backplate covers the entire back side of the card, lending it some much
needed rigidity. Heavy coolers tend to cause the graphics card PCB to bow.
|
|
The back panel is stocked with three DisplayPort, one HDMI, and one DVI-I connectors, and can drive up to four displays simultaneously. The GTX 980 supports HDMI 2.0 as well, which supports 4K resolution at 60 Hz. There’s only a small ventilation port at the back and as the cooler employs top-down blowing fans most of the heat produced by the card will remain in the case. You can also see from this angle that the card is wider than normal, extending past the edge of the expansion slot by 3.4 cm; This will interfere with some narrower cases.
CPU: Intel
Core i5-4670K Quad-Core 3.4 GHZ 6 MB Cache
– US$220
Alternatives:
- Intel
I5-4440 – US$180 - Intel
Core i5-4460
– US$190
Currently AMD’s strength is based on the number of cores they pack into lower
priced processors. The best example of this is the six-core AMD
FX-6300 Black Edition
which can be had for a mere US$100. Many applications, especially games,
can’t take advantage of that many threads, however; Single-threaded performance
is more important, an area where Intel holds a notable advantage. That being
said, CPU bottlenecking shouldn’t be an issue with a single GTX 980 for any
mid-range or higher processor. The real back-breaker in the Intel-AMD debate
however is the superior energy efficiency of Intel’s lineup. This makes their
chips much easier to cool, a key advantage when attempting to silence a PC.
|
|
An ideal choice is the Core i5-4670K, a quad core chip running at 3.4 GHz (up
to 3.8 GHz with Turbo Boost) with an unlocked multiplier for easy overclocking
to improve overall performance further. Honestly, any of the 4xxx series would
perform similarly in terms of gaming, so which particular model isn’t overly
important.
Unfortunately we didn’t have an i5-4670K on hand so an i7-4770K
became our stand-in. With a TDP of 84W, it should have similar thermal properties
which is all that’s needed from a CPU for the type of testing we’ll be conducting.
The i7-4770K is a standout performer but it’s actually not a great buy currently
as it has been superseded by the higher-clocked i7-4790K which occupies a similar
price-point. Rather than lowering the price on older models, for whatever reason,
Intel tends to just let them wither on the vine.
CPU Cooler: Scythe
Kotetsu
– US$40
Alternatives:
- Coolermaster
Hyper 212 Plus Universal – US$33 - Scythe
Mugen 4–
US$50 - Be
Quiet! Shadow Rock Slim
– US$50 - Silverstone
Argon 3
– US$50
|
|
Haswell processors don’t run particularly hot but the Intel stock cooler is
woefully insufficient for a silent PC. We’re going with one of our current favorite
value cooling solutions, the Scythe
Kotetsu, which delivers excellent performance at a low price. This modest
tower cooler is equipped with a single 120 mm fan with an overall pleasant/smooth
sound.
Many gamers regard an AIO water cooling unit as a defacto choice for a high-end
build but we have to disagree as they are neither cost or noise efficient. They
also usually ship with acoustically poor fans, and the squealing and gurgling
sounds from the pump raises the noise floor, making it difficult to get the
noise level down when the system is idle.
COMPONENT SELECTION (Con’t)
ATX Motherboard: Asus
Z97-PRO – US$200
Alternatives:
- ASUS
Z97-P
– US$115 - ASUS
Z97-A
– US$140
The motherboard isn’t usually a factor in gaming performance as long as it
has a full-sized PCI-E slot, but it’s not a component you want to skimp on.
You should select a model with all the features you need but there are some
intangibles that can have an impact on noise. Controlling fans can be a critical
part of optimizing a PC for low noise operation. Some cases ship with fan controllers
and there are aftermarket models, but they are usually inferior to the best
fan controllers already built into the motherboard.
For the chipset, Z97 is ideal as it accepts the newer Haswell Refresh LGA1150
chips without a BIOS update, and it is somewhat future-proof as it will be compatible
with Intel’s next line of chips, codename Broadwell. It also supports CPU overclocking,
but if that’s deemed unnecessary, an H97 board would do just fine.
|
|
The ASUS Z97-PRO has the
best fan controls of any motherboard to come into our labs, making it an easy
choice. As a bonus, it also has excellent energy efficiency during heavy load.
Enthusiasts will enjoy its UEFI BIOS, which sports tons of options and little
conveniences that make it a joy to use, as well as its high-end power regulation
system. It’s stocked with more features than most users need so if you want
stay with basics, there are lower-tier models in the ASUS Z97 line that provide
similar core features.
|
|
Of paramount importance is the Z97-PRO’s six fan headers, five of which are
controllable by both PWM and DC (voltage) control. Fan speed behavior can be
configured in the UEFI BIOS or via ASUS Fan Xpert utility and each individual
header can be set to react to any of the onboard temperature sensors, making
it a truly dynamic system. It also has a pair of full-sized 16x PCI-E 3.0 (8x/8x
in dual configuration) so there’s always the option down the road to add an
second video card for SLI.
RAM: Kingston
HyperX Genesis Kit 2x4GB 1866MHz DDR3
US$100
Alternatives:
|
|
Precisely what RAM is used as system memory is not critical, although other
web sites have identified DDR1600 to DDR1833 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 of money unless
you have a specific need for it, and RAM is one of the easiest things to add
later to a system, if you really need it 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 RAM has been
solid for us, and it sports lower profile heatspreaders that don’t get in the
way of big heatsinks.
SSD: Kingston
HyperX 3K 240 GB SATA III – US$120
Alternatives:
- Crucial
MX100 256GB – US$110 - Samsung
850 Pro 256GB
– US$170 - Crucial
MX100 512GB – US$210 - Samsung
840 EVO 1TB
– US$400
Solid-state storage may be the most significant advance in the last decade
for silent computing. With no moving parts, they generate zero noise, but also
have ridiculously low latency, resulting in fast loading times. As games continue
to grow in complexity, having an SSD becomes increasingly advantageous. A 240~256GB
model should be considered a starting point; with Windows and a few triple-A
gaming titles installed, a smaller drive could be filled close to capacity.
Also, with ever increasing memory density, fewer dies are needed, which can
mean fewer read/write channels being used and slower performance for SSDs below
240-256GB capacity.
|
|
The Kingston HyperX
3K is one of many SSDs using the SandForce SF-2281 controller which
employs built-in file compression to improve speeds, and is a well-priced, good
performer. For those looking for more value, the Crucial
MX100 is only a tad slower and offers the best price per byte; the 512
GB model we reviewed is probably the value SSD of the season. The Samsung
850 Pro is the creme de la creme, but you won’t gain any real benefit
in gaming. If you’re an inveterate gamer and want fast loading and interactive
menus for all your umpteen games, then the $400 starting price for 1 TB SSDs
might not seem too onerous, and it obviates the need for any mechanical spinning
drives in the system to slow you down or make noise.
| Postscript 31 Nov: SSD Substitution: Samsung |
Hard Drive: Seagate
Barracuda 2 TB
– US$80
Alternatives:
- Seagate
Barracuda 3 TB
– US$100 (for performance) - Western
Digital Caviar Green 3 TB
– US$105 (for low noise) - WD
Red 3 TB NAS Drive
– US$125 (for low noise, warranty)
Most PC users store more than just games, but also high definition video or
lossless audio, so it’s entirely possible that even the largest capacity SSD
you can afford might not be enough. If you have very deep pockets, you could
shell out for several 1TB models, but the rest of us lowly peasants must rely
on mechanical hard drives to handle the overflow.
|
|
For a silent PC, a “green” hard drive is perfect as they are quiet
and produce limited vibration. However, if you need it to store games or applications,
a 7200 RPM model like the Seagate Barracuda 2TB will make for a more responsive,
albeit noisier experience. We’re actually going to use the hybrid version of
the Barracuda, the Seagate
Desktop 2 TB Solid State Hybrid Drive, but the acoustics are the same
as it’s the same physical drive with the addition of 8GB of NAND Flash cache.
Actually this $100 hybrid is a reasonable OS drive option if your budget is
tight and a slower in start to your PC or games isn’t going to bother you.
COMPONENT SELECTION (Con’t)
Case: Fractal
Design Define R4
– US$110
Alternatives:
- Antec
P280
– $90 - Silverstone
Raven RV05
– US$115 - Corsair
Obsidian 550D
– US$130 - Silverstone
Fortress FT05S
– US$180
There is so much diversity in the market that it’s easy to get overwhelmed
when choosing a case. Providing a good supply of fresh air from the outside
to the components inside is vital as it allows for cooler temperatures and thus
lower fan speeds. However, too much ventilation allows more noise to escape.
It’s a delicate balancing act that stymies many manufacturers or simply don’t
understand.
The best performing ATX cases we’ve tested are the SilverStone
Raven/Fortress
series which use a layout with a rotated motherboard and large 18 cm fans to
drive bottom to top airflow. However, these are considered niche cases by some,
and the Fortress line is somewhat pricey, while the Raven’s aesthetics are off-putting
to some. We decided this time to use a more traditional tower, the Fractal Design
Define R4, which shares DNA with the classic Antec
P183 and P280, but
in a more modern, versatile design. The Corsair
Obsidian 550D is another viable option in the same genre as the FD Define
R4 and Antec P280, and it will likely be featured in another build guide in
the near future.
|
|
The case features acoustic damping on the side panels and behind the front
door, and thick padded blocks to cover unused side/top fan placements. Airflow
is often an issue in this genre of cases but Fractal has dealt with this nicely
by employing slits all along the sides of the front bezel. If you’re familiar
with the R2/R3,
the R4 is essentially a wider version with larger 140 mm fans and removable
hard drive cages. The area of a 140mm circle is some 56% larger than a 120mm
circle, so venting is improved substantially, though that number doesn’t translate
linearly to cooling.
|
|
The only flaws in its design are somewhat poor physical support for the top
panel which can result in vibration-induced noise if top fans are employed,
and a completely unnecessarily grill at the front. When the door is closed,
the latter seals off the entire area so there’s no way dust is getting in. The
side vents are all behind this portion, making it completely superfluous. Going
back to the vented top panel, the relatively flimsy feel seems unavoidable;
all cases with largish top vent have a similar issue, even the old P180, which
was built of heavy steel metal by today’s standards. Also, we’re not using the
top vents at all.
|
|
The case ships with two 140 mm fans, one exhaust and one intake. The front
fan clips onto a plastic mount and can be oriented in three different positions.
The fan is a Silent Series R2 1000 RPM variant with fairly decent acoustics
though the one in front is louder because a dust filter and the unnecessary
grill are pressed up against the intake, creating some extra turbulence. Thankfully
the door masks much of this noise.
|
|
The internal layout is fairly common for a tower case, with a bottom-mounted
power supply position, and drive cages at the front. Padded holes are scattered
around the edges of the motherboard tray to push cables out of sight and out
of the airflow path. As optical media is starting to fade away there are only
two 5.25 inch bays while below it is space for eight 3.5/2.5 inch drive mounts.
Removable drive cages can amplify hard drive vibration if they are too loosely
secured but these fit quite snugly compared to most cases. For a gaming system,
it’s best to take the middle cage out of it’s not being used, as it impedes
the front intake fan.
|
|
The R4 has its own fan controller that is powered by a 4-pin molex connector and supports three 3-pin fans. It’s wired to a three speed switch at the front that can run the fans at 5V, 7V, or 12V. It’s a nice addition but we would’ve preferred an adjustable knob for more granular control.
Power Supply: be
quiet! Straight Power 10 600W – US$125
Alternatives:
- Corsair
CS 650
– US$95 - Corsair
RM 650
– US$110 - Seasonic
X-650 – US$120
Today’s power supplies are highly efficient and most models don’t produce much noise. This is especially true when they’re bottom-mounted, like they are in the Define R4, as the floor position gives it direct access to cool air from underneath the case. From a thermal standpoint, it’s essentially isolated from the rest of the components and only has to cool itself.
You may think a passively cooled unit would be idea for a quiet system, but
this is not the case generally for gaming system that have hot components. The
~10% of power even the most efficient PSUs lose in converting AC to DC turns
into heat, which rises and increases the thermal stress on the GPU and other
components. In a system like this, the maximum amount of PSU heat would be perhaps
30W, but that can be the difference between all the fans running at minimum
speed verus running them 300 RPM faster, which means higher noise. We’d rather
have the 120~140 mm fan inside the power supply running faster rather than putting
extra demand on the smaller, less efficient video card fans.
|
|
be quiet! makes some of the quietest power supplies on the market and
the Straight Power 10 is a fine example, remaining practically silent for much
of its lower range, just like the Dark
Power Pro 550W we reviewed a while back. be quiet! products are
sold by NCIX
in the US and Canada. If you need an alternative, we generally recommend models
manufactured by Seasonic (which includes some models sold by Corsair, Antec,
XFX, and others). We can attest to the quietness of the Seasonic
X-series, and according to techPowerUp!,
the Corsair CS650M and RM650 are fairly quiet at loads between 200~400W, the
range where most single GPU builds fall in.
|
|
Our be quiet! Straight Power 10 600W has about twice as much capacity
as we really need, but this allows room for future upgrades. It’s 80 Plus Gold
certified for high efficiency. Compared to a Bronze unit, it will use a bit
less power and run slightly cooler, but this may not be enough to actually lower
the noise it produces in any meaningful way. The claimed baseline noise specification
is 10.7 dBA@1m, which means it won’t register on our sound level meter in our
10~11 dBA ambient chamber.
ASSEMBLY
Putting together the system was fairly straightforward as the Define R4 is
reasonably spacious and all the components are installed with traditional methods.
We ran into only one issue, the motherboard mounting holes weren’t completely
threaded so we had to use a nut-driver (you can also use pliers) to get the
standoffs fully installed.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
SOFTWARE
The process of building a quiet PC doesn’t end once all the physical parts
are put together. Left at the mercy of stock settings, even the best selected
components can produce more noise than is necessary. Controlling fan speeds
is key and this is most easily performed through the motherboard software.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
TESTING
System Configuration:
- Intel Core i7-4770K processor – 3.5 GHz (3.9 GHz with Turbo Boost), 22nm, 84W, integrated HD 4600 graphics
- Scythe Kotetsu CPU
cooler - ASUS Z97-PRO
motherboard – Intel Z97 chipset, ATX - ASUS Strix GeForce GTX 980 graphics card – 2048 CUDA cores, 1178 MHz clock (1279 MHz with GPU Boost), 7010 MHz memory
- Kingston HyperX Genesis memory – 2x4GB, DDR3-1600, C10
- Kingston HyperX 3K solid-state drive – 240GB, SATA 6 Gbps, 2.5-inch
- Seagate Desktop SSHD hybrid drive – 2TB, 7200 RPM, 8GB NAND Flash, SATA 6 Gbps, 3.5-inch
- Fractal Design Define R4 case – ATX
- be quiet! Straight Power 10 power supply – ATX, modular, 600W
- Microsoft
Windows 7 Ultimate operating system, 64-bit
Measurement and Analysis Tools
- Media Player
Classic – Home Cinema to play H.264/VC-1 video. - TMPGEnc Xpress video encoder for stress testing.
- Resident Evil 6 standalone benchmark for stress testing.
- Prime95
processor stress software. - FurMark
openGL GPU stress software. - CPU-Z
to monitor CPU frequency and voltage. - GPU-Z to
monitor GPU temperature, load level, and fan speed. - AIDA64
to monitor system temperatures and fan speeds. - ASUS Fan Xpert 3
to monitor system temperatures and monitor/change fan speeds. - Extech
AC Power Analyzer 380803 AC power meter, used to measure system power consumption. - PC-based spectrum analyzer:
SpectraPlus with ACO Pacific mic and M-Audio digitalaudio interfaces. - Anechoic chamber
with ambient level of 11 dBA or lower
Baseline Noise
Before any stress testing we took a quick look at the noise produced by each part. We did this by leaving the system on but idle, and stopping/unplugging all the other fans (and the hard drive) to isolate individual components.
| GTX 980 + Define R4 System Component Noise Levels (idle, power supply on) | ||
| Noise-producing Components | Fan Speed | SPL @1m |
| Baseline (PSU on only) | N/A | 11 dBA |
| HDD (hard-mounted) | N/A | 14 dBA |
| GPU fan | 740 RPM (min) | 16 dBA |
| CPU fan | 340 RPM (min) | 11~12 dBA |
| 650 RPM (~5V) | 12 dBA | |
| 900 RPM (~7V) | 14~15 dBA | |
| 1120 RPM (~9V) | 19~20 dBA | |
| 1400 RPM (max) | 25 dBA | |
| Case fans | 500 RPM (min) | 14 dBA |
| 620 RPM (~7V) | 17~18 dBA | |
| 740 RPM (~9V) | 21 dBA | |
| 920 RPM (max) | 25~26 dBA | |
| Measuring mic positioned 1m at diagonal angle left/front of case. Ambient noise level 10~11 dBA@1m. | ||
The baseline, with only the PSU on, measured only 11 dBA@1m as expected from
be quiet!’s 10.7 dBA claim, barely above the noise floor of our ultra-quiet
hemi-anechoic chamber, so its fan was effectively a non-factor. The hard drive,
hard-mounted in the bottom drive cage, brought the noise level up to 14 dBA@1m.
The graphics card actually didn’t emit any noise as its fans don’t spin in this
state, but when we manually activated the fans at minimum speed (740 RPM or
36% according to GPU Tweak), it generated 16 dBA@1m, making it the loudest component.
Using Fan Xpert 3, we found the CPU fan could produce anywhere from 11~12
dBA@1m to 25 dBA@1m. The two case fans were slightly louder. Incidentally, we
found the two Fractal fans can spin a bit slower, by about 50 RPM by hooking
them up to the case’s fan controller and flipping the switch to 5V. While the
Fractal 140mm case fans weren’t bad at all, they did have something of a growly
quality, and it’s quite possible that swapping them out for quieter, smoother
sounding fans could buy us a decibel or two when the system is at low load.
Keep that in mind as we guide you through our measurements and analysis.
The GPU fans are easily the roughest sounding of all the fans. Smaller fans
usually have worse acoustics and models that are designed for horizontal orientation
often employ ball bearings that produce more motor noise than sleeve/fluid bearings.
These fans have an unpleasant growl which manifests in multiple tonal peaks
in the 150~600 Hz range. Subjectively, it didn’t sound that bad inside the case,
as the total noise output wasn’t that high. That being said, it’s clear that
we should try to keep the GPU fans at as low a speed as possible.
Mounted in the bottom cage in the usual manner, the Seagate hard drive produced a tonal peak at 120 Hz, corresponding to its 7200 RPM motor speed. This peak pulsed slightly up and down and was accompanied by a strumming effect which was audible at short distances. It’s a clear sign of heavy drive vibration, the result of the drive rattling against the rest of the case.
Here we compare the system with the hard drive hard-mounted vs. suspended,
with the rest of the fans running at minimum speed. As you can see, suspension
is superb for dealing with this issue, wiping out the 120 Hz peak almost completely
and reducing the overall noise level slightly. The only drawback is the 5.25
inch drive bay is near the top of the case, a hotter position than the lower
drive cage near the floor. It’s a small sacrifice to pay for a superior sounding
machine, though.
Stress Testing: CPU-centric
| System Measurements | ||||
| System State | Idle | x264 Playback | Video Encoding | Prime95x8 |
| CPU Fan Speed | 340 RPM (minimum) | 650 RPM | ||
| CPU Temp | 26°C | 27°C | 53°C | 64°C |
| MB Temp | 32°C | 33°C | ||
| SSD Temp | 23°C | 27°C | 24°C | |
| HDD Temp | 32°C | 33°C | 32°C | |
| GPU Temp | 35°C | 38°C | ||
| SPL@1m | 15 dBA | |||
| System Power (AC) | 46W | 50W | 106W | 131W |
| System fans at 500 RPM (minimum), GPU fans off (auto). Ambient temperature: 21°C. | ||||
We began by testing the system with CPU-centric applications to see how it
performs with non-gaming tasks. First off, it’s clear that despite its size,
the Kotetsu is a solid heatsink. Even with its fan spinning at a paltry 340
RPM, the CPU temperature was held at 53°C during video encoding with TMPGEnc,
a demanding multi-threaded real world test. Prime95 is a synthetic test that
pushed it further, driving the CPU close to 70°C. At this point we felt
it appropriate to ramp up the CPU fan to 650 RPM as the fan was still so quiet
at that speed, it didn’t effect the overall noise level one bit. The CPU doesn’t
actually throttle until close to 100°C so it wasn’t a necessary step, but
as it’s essentially a free upgrade acoustically, there’s certainly no harm in
doing so.
The GPU stayed very cool throughout these tests and barely warmed up under
any of the load states. With a temperature under 40°C, its fans didn’t spin
up at all, as promised. The entire system measured 15 dBA@1m which is very quiet,
and as all the active fans were running at low speeds and the hard drive was
suspended, the sound emitted by the system was almost completely without tonality
or periodic variations; in a word, its noise was very soft, smooth and benign.
Stress Testing: GPU-centric
For our GPU-intensive states, we used the Resident Evil 6 Benchmark Tool which benchmarks for a couple of minutes, displays the result, pauses for a short interval, and repeats. We tried a few other games but this one seemed to put the most demand on the GPU. The other test was a more demanding combination of Prime95 and FurMark, an incredibly stressful utility that pushes the GPU to its limit. For this state, we ran Prime95 with only two threads instead of the maximum eight as most games run with less than 50% CPU utilization. The two combined still produces much more heat and power consumption than any PC game title.
The GPU fans kick in once graphically intensive applications heat up the GPU
sufficiently. By default, the GTX 980’s fan control system attempts to keep
the the GPU temperature in the low 70’s which is fairly aggressive. We usually
manually override the speed to allow it to reach 90°C but the card has a
fail-safe that activates at that temperature, forcing the fan to speed up, so
we raised the bar and deemed 85°C to be the maximum acceptable GPU temperature.
| System Measurements | ||||
| System State | Prime95x8 | Resident Evil 6 Benchmark | Prime95x2 + FurMark | |
| GPU Fan Speed | off (auto) | 740 RPM (minimum) | 990 RPM* | 1630 RPM (auto) |
| CPU Temp | 64°C | 50°C | 63°C | 61°C |
| MB Temp | 33°C | 43°C | 46°C | 44°C |
| SSD Temp | 24°C | 25°C | 26°C | 26°C |
| HDD Temp | 32°C | 35°C | 36°C | 36°C |
| GPU Temp | 38°C | 79°C | 85°C | 72°C |
| SPL@1m | 15 dBA | 17 dBA | 18~19 dBA | 25 dBA |
| System Power (AC) | 131W | ~252W | 291W | 290W |
| *Speed adjusted to achieve a GPU temperature of 85°C. CPU fan at 650 RPM, system fans at 500 RPM (minimum). Ambient temperature: 21°C. | ||||
Despite almost doubling the system power draw compared to a full blast of Prime95, the Resident Evil 6 test didn’t prove to be much of a challenge. We were able to slow the GPU fans to the minimum speed while keeping the GPU core under 80°C. The CPU wasn’t under much pressure so it settled at a relatively lukewarm 50°C while the MB sensor recorded a 10°C increase. The resulting noise level was 2 dB higher but this wasn’t enough to alter the rig’s general acoustic character.
The Prime95 + FurMark test was much more demanding, drawing close to 300W from the wall. We managed to maintain a stable GPU temperature of 85°C with the GTX 980’s fans spinning just shy of 1000 RPM, for a total noise level of 18~19 dBA. Interestingly, the CPU ran almost as hot as it did during the full Prime95 test, even though only two cores were being put to work. The hot air coming off the graphics card made the CPU area as hot as it would be under a full CPU load.
When we reverted the GPU fans to automatic control, the fan speed kicked up
by more than 600 RPM, resulting in a bit of extra cooling for the CPU and MB,
while the GPU core itself received 13°C of relief. However, the noise level
increased by 6~7 dB, which is substantial when you consider a 10 dB difference
is equivalent to twice the perceived volume.
We also experimented with the speed of the case fans, but the results frankly
weren’t worth presenting. Ramping up the case fans resulted in slightly lower
temperatures but it also drove the noise level up as well so we were unable
to achieve a net positive result. It’s almost a shame to have such a versatile
fan control system at our disposal when the case fans seem to work optimally
at minimum speed all the time.
Dialing back the GPU fan speed not only produced a quieter machine but the subjective quality of the sound improved as well. It mostly affected the higher frequency noise at 800 Hz and above, which was reduced to levels similar to the system at idle when the GPU fans were turned off completely. There was also less tonality in the mid-range as well, between 400 and 700 Hz.
The GTX 980 fans emit a harsher sound than the other fans in the system but
if they’re running slowly enough, they’re difficult to pick out. At 990 RPM
they were barely discernible from the rest of the system, but don’t take our
word for it. Listen to the audio recordings on the next page to hear exactly
what the system sounded like.
Postscript, 20 Nov: Custom GPU Fan Automation
Once we located the user fan profile option screen in the ASUS Strix GPU Tweak
utility, the fan curve we ended up with was for the fan to start spinning at
35% (the minimum speed of the fans other than stopped) when the GPU reaches
59°C, then 38% fan speed at 83°C, and finally, 59% at 99°C —
this last being a simple fail-safe, as our experiments had determined the GPU
was unlikely to go past 90°C in our test conditions. These settings worked
perfectly, stopping the Strix fans at modest loads and kicking up just enough
to keep the GPU under 85~87°C. (The odd selection points are not exactly
what we intended but the very small graphic interface is difficult to manipulate
precisely; we were shooting for 35%-60°C, 38%-85°C and 60%-100°C.)
Stress Testing: Front Airflow Effects
| System Measurements: Prime95x2 + FurMark | |||
| Front Bezel Configuration | Stock | Open (no filter, grate, door) | |
| GPU Fan Speed | 990 RPM | 740 RPM (min) | |
| CPU Temp | 63°C | 52°C | 60°C |
| MB Temp | 46°C | 40°C | 38°C |
| SSD Temp | 26°C | 26°C | 26°C |
| HDD Temp | 36°C | 36°C | 36°C |
| GPU Temp | 85°C | 75°C | 83°C |
| SPL@1m | 18~19 dBA | 20~21 dBA | 19 dBA |
| System Power (AC) | 291W | 289W | 289W |
| CPU fan at 650 RPM, system fans at 500 RPM (minimum). Ambient temperature: 21°C. | |||
Though we did manage to achieve an adequately cooled system emitting less than
20 dBA@1m using the Define R4, we were curious as to how much the front intake
design affected the overall performance. Opening up the front bezel by removing
the filter and grill, and popping open the door produced a substantial improvement
in cooling. Both the CPU and GPU temperatures dropped by 10°C and the MB
temperature improved by 6°C.
This allowed more noise to escape, though, even with the door partially blocking
the sound path to the microphone (diagonally left from the front). Slowing the
GPU fans to the minimum 740 RPM helped make up this difference somewhat while
still holding onto a slight cooling advantage. Many cases omit the door completely
in favor of a well-ventilated front, and these results suggests there is some
validity to that strategy.
AUDIO RECORDINGS
These recordings were 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 WAV files to these MP3s. They represent
a quick snapshot of what we heard during the review.
Each recording starts with ambient noise, then 5~10 second segments of product
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 comparing all the sound files.
GTX 980 + Define R4 System, idle, fans at minimum speed
— hard drive hard-mounted (15~16 dBA@1m)
— hard drive suspended (15 dBA@1m)
GTX 980 + Define R4 System, hard drive suspended
— idle, CPU and system fans at minimum speed, GPU fans off (15 dBA@1m)
— Resident Evil 6 Benchmark, CPU, system, and GPU fans at minimum (17
dBA@1m)
— Prime95 x2 + FurMark, CPU at 650 RPM, system fans at minimum, GPU fans
at 990 RPM (18~19 dBA@1m)
— Prime95 x2 + FurMark, CPU at 650 RPM, system fans at minimum, GPU fans
at 1630 RPM (25 dBA@1m)
EVEN QUIETER?
It was noted earlier that the 14 dBA@1m noise level of the stock Fractal case
fans could be lowered by upgrading to quieter fans, but with the Seagate Barracuda
HDD matching the minimum noise of the fans, there is not much to be gained.
However, if you opt for an all-SSD setup with no mechanical drives at all, then
it is certainly more worthwhile to consider that fan swap. Keep in mind that
at 15 dBA@1m or just below in idle/low load, the system is already quiet enough
that you really have to be in a super-quiet environment for any further reductions
in noise to be even audible; ie, you’re bumping against the ambient noise floor
of most living spaces. But to be completely thorough, let’s consider options.
The most obvious candidate among the 140mm fans we’ve reviewed is the Antec
TrueQuiet 140
This is a superbly quiet fan at minimum speed (500~550 RPM), which made
too little noise for us to actually measure at 1m in our 10~11 dBA anechoic
chamber, like the be quiet! Straight Power 10 PSU. Two of these fans will
bring the total SPL to 12~13 dBA. This would then become the baseline noise
of the system, given that there are no other noise sources at idle or low load,
other than the fan on the Scythe Kotetsu, which reaches only 12 dBA@1m in this
activity state. There might be a few degrees price to pay in cooling at high
gaming load, as our testing showed the True Quiet 140 to have only average cooling
capability for its RPM. However, there is still cooling headroom in the system
to play with, so this is a perfectly viable option for the uber-silencer. The
Antec fan has other winning qualities as well: It costs as little as $12~13
each and seems widely distributed all over the world.
FINAL THOUGHTS
For most of the past decade, the high-end GPU proved to be an implacable obstacle
to silencing a PC. The high power requirement, small inefficient fans, and the
undesirable video card location in most cases all combined to make for a perfect
storm of piping hot, screaming loud operation. It’s still a problem, but with
the latest technology you don’t need a huge aftermarket heatsink or a AIO liquid
cooler to tame the beast. Sound component selection will get you most of the
way, and fine fan control can carry you to the finish line where peace and quiet
awaits.
We have to give kudos to NVIDIA for their highly energy efficient new GPU architecture.
The GTX 980 is a marvel, the fastest single GPU model on the market with a TDP
of just 165W. If we had proceeded with this build a few months ago, before the
GTX 980’s release, it would have been a lot tougher to pull off. Instead, the
actual process was relatively hassle-free. The laudable fan control system built
into the ASUS Z97-PRO was
almost superfluous, the Scythe Kotetsu
barely had to break a sweat, and we’re pretty sure the be
quiet! Straight Power 10‘s fan never ramped up from its minimum speed.
All we really had to do was slow the CPU and case fans, suspend the hard drive,
and relax the GPU’s automated fan control.
Incidentally, the system we built is quiet enough to qualify as an SPCR
Certified Silent PC, meeting the required noise levels of 15 dBA@1m
or lower at idle, and 20 dBA@1m or lower on full load. Admittedly, this system
was not tested at the demanding 30°C ambient temperature the SPCR Certification
program calls for, but at a typical temperate room temperature of 20~23°C.
For those with a bit more noise tolerance, there’s plenty of cooling headroom
to squeeze more performance through overclocking. Both the CPU and motherboard
are well equipped for this and according to various review sites, the Strix
GTX 980 can handle an extra 200 MHz (about a 16% gain). On the flip side, for
those seeking an even quieter experience from this powerful gaming rig, the
SSD-only storage option with a swap to Antec
TrueQuiet 140 fans is a good option to get down to ~13 dBA@1m SPL at
low load, which is an amazing level of quiet you have to really work hard to
hear.
The total cost of our build comes out to ~US$1,600, and if you choose some
of the more value-conscious alternatives we presented, like the GTX 970, you
can cut that by about US$400 while still maintaining a high quality gaming experience.
We hope you gained some insight with regard to assembing quiet gaming systems,
and we’d like to remind everyone that this is only the first in what hopefully
will be many quiet and silent gaming build guides moving forward. Soon, we’ll
tackle smaller form factors and experiment with more extreme configurations
with dual graphics cards, possibly with aftermarket cooling, both air-cooled
heatsinks and AIO liquid units.
Many thanks to Intel,
ASUS, Kingston,
Seagate, Scythe,
Fractal Design,
and be quiet! for
sponsoring the components in this build guide.
* * *
Articles of Related Interest
Quiet ATX Gamer, R5 Version – Refresh
of the build described here with Fractal R5 case
ASUS Z97-PRO LGA1150 Motherboard
Crucial MX100 512GB
& Samsung 850 Pro 256GB SSDs
Silent Mid Gaming PC Build Guide
Recommended Power Supplies
Recommended
Hard Drives
Case
Basics & Recommendations
Recommended Heatsinks
* * *
