Our seventh article in this season’s new quiet gaming PC build guide series is our first complete discrete liquid cooled system, featuring the VisionTek CryoVenom R9 290 Limited Edition Graphics Card. This 450W beast gets tamed to a cool, soft purr. POSTCRIPTS added on pump acoustics — 15th and 27th of Jan.
Most of the recent SPCR silent gaming PC build guides were initiated with a
clear set of goals we brought to each project: For example, a mid-tower ATX
system incorporating the quietest GTX 980 with sub-20 dBA SPL at full load.
Or a mini-ITX system suitable for LAN events with similar gaming and acoustic
performance. The components were then sought out, and once collected, assembled,
with the usual iterative fine tuning for optimal quiet and cooling.
There have been one or two exceptions where a particular product or some external
stimulus got a project started. The NCASE M1 build was spurred by the simple
fact of having one on hand: It’s a hard-to-get, unusual and expensive case,
and it was a challenge to make a powerful gaming system in such a small case.
The second was our first MicroATX gaming build, which was initiated by my personal
fascination with the Corsair Carbide Air 240 case.
This build is another exception. It was initiated by an invitation to review
an AMD Radeon R9 290 graphics card equipped with a full-cover watercooling block.
The invitation from VisionTek came just as we were planning out the entire series
of gaming builds for late fall and winter. Further discussion proved VisionTek
were willing to supply the other components for a discrete watercooling system
and a CPU watercooling block to complete the package. Could the VisionTek CryoVenom
R9 290 be the basis of a quiet, discrete, entirely water cooled system? Our
limited experience with water cooling told us the pump would be a huge challenge
to silence. But for a super-hot graphics card like the R9 290, 290X or any recent
dual-GPU top dogs from either nVidia or AMD, especially in a multi-card rig,
watercooling looked close to being the only viable option, never mind the noise
level. It was a challenge we could not resist.
That’s how this quiet watercooled gaming PC build guide project began. We had
never even looked closely at a discrete watercooled PC before, never mind build
one. Diving into the deep end seems an apt phrase here.
Since SPCR has never covered VisionTek products in the past, it seems appropriate
to introduce the company.
Visiontek was founded in 1988 and became a major supplier of peripherals and
memory products for computers in the US. According
to its web site, Visiontek currently “offers state-of-the-art, graphics
cards from AMD, memory, solid state drives, power supplies, candyboards, gaming
network cards, TV tuners, and more to the computing industry, with branded products
still found at major retailers, distributors, PC system builders, and e-commerce
sites across North America.”
VisonTek’s partner for watercooling technology is EKWB
(EK Water Blocks) of Slovenia, a renown specialist in computer watercooling
products for more than 15 years.
CRYOVENOM R9 290 LIMITED EDITION
A big portion of any build guide is a discussion of component option for the
build. In this case, only one graphics card was considered, the CryoVenom R9
290 LE which initiated the project.
The R9 290 GPU has been around for just over a year, long enough for it to
have been thoroughly tested by many web tech review sites. At stock clocks,
the R9 290 ranks a bit higher than nVidia’s GTX 780 but trades blows with the
much newer Maxwell-core GTX 970, losing by a hair more often than not. Huge
AMD price cuts since the GTX 970 and 980 introduction last fall have positioned
the 290 at some 10% lower price than the $330~380 GTX 970.
The cooling block covers all the hot elements on the R9 290 card. The
quality of machining is top notch. The back panel cover occupies two slot
spaces for all the output ports but the card itself is probably slim enough
to fit into the space of a single slot.
The PCB gets support from a full size, thick backplate made apparently
from anodized aluminum.
The full designation of the product is VisionTek
CryoVenom® R9 290 Limited Edition Graphics Card (900698). This is an
AMD Radeon R9 290 with 4GB GDDR5 factory-fitted with a full cover water block
from EKWB. The card came with an individually signed test report certifying
it to be stable with the GPU overclocked 19% to 1127 MHz, and memory overclocked
18% to 1475 MHz. OC capabilities should push this 290’s performance past the
stock 1 GHz clock of the 290X even with the latter’s 9% advantage in the number
of stream processors. It probably allows the CryoVenom 290 to edge past the
GTX 970. Of course, at such clock speeds, this card will also exceed its nominal
275W TDP and the 290W TDP of the 290X. This is where the nVidia Maxwell architecture
GPUs actually have their greatest edge: 165W TDP for the GTX 980 and just 145W
for the 970. Devotees of silent computing need no more information to know the
R9 290 is much harder to cool quietly than the GTX 980/970. Liquid cooling of
the entire card certainly looks worthwhile given the ~300W thermal load.
The CryoVenom R9 290 LE is a most impressive device to behold and examine.
The base that mates with the GPU, VRM and RAM is made of nickel-plated copper,
while the rest of the waterblock is machined from clear acrylic, with an additional
aluminum plate possibly for additional heat conduction. Liquid tubing can be
attached from either side of the card — or both sides as necessary. A ~2.8mm
thick anodized aluminum plate which covers the entire backside provides excellent
rigidity and resistance to PCB flexing.
VISIONTEK CRYOVENOM 360 LIQUID COOLING KIT
The CryoVenom R9 290 LE came together with the $360 VisionTek
CryoVenom® 360 Liquid Cooling Kit, which includes a discrete 360mm radiator
and three 120mm fans, a couple meters of tubing, a combined pump/reservoir and
lots of other hardware. It arrived in a large carton containing an impressive
wooden box and the long radiator in a separate retail box.
Impressive wood box for all the goodies except the big radiator.
VisionTek EK Supremacy CPU waterblock
A little later, we asked for and received a sample of the VisionTek
EK Supremacy CPU waterblock. This would be used in the loop along with the
VisionTek G9 290. Another nickel-plated base and acrylic assembly, this CPU
waterblock is also impressively made with excellent machining, fit and finish.
VisionTek EK Supremacy CPU waterblock. The mounting system
is rugged and precise, with large preloaded spring thumbscrews and hardware
for all current CPU sockets.
Beautifully smooth, flat nickel-plated copper base.
With no previous experience to go by, our take on the supplied watercooling
components was basic:
PC WATERCOOLING RESOURCES
A fair amount of web research was done on PC water cooling to bolster our knowledge.
One of the most detailed and comphensive we found is Martin’s
Liquid Lab, described by Martin thus:
“This site is voluntarily built upon science, engineering, testing,
reviewing, and sharing information with my fellow water cooling community…
This site is funded entirely by my donated time, my donated testing equipment,
my donated testing hardware, community donated hosting, and donated parts
to test and review that comes from manufacturers, vendors, the community,
and myself. It is entirely a donation & volunteer community based effort
for the good of the community.”
Other useful sites we drew upon include:
ASSESSING BASICS: PUMP ACOUSTICS
Even before considering any other components we felt it was important to get
a handle on any unfamilar noise-making components already on hand. There were
really only two: The three Gelid fans on the big radiator and the VisionTek
EK-SBAY DDC 3.2 PWM reservoir-pump combo. Fans are easily replaced if too noisy,
so we focused on the pump.
Neither VisionTek nor EKWB provide any acoustic data on the pump, which is
a Laing DDC-3.2. The primary specifications are reproduced here:
– Motor: Electronically commuted ball bearing motor
– Rated voltage: 12 V DC
– Power consumption: 18 W
– Maximum head pressure: up to 5.2m
– Maximum flow rate: up to 1000 L/h
– Maximum liquid temperature: 60°C
Pressure and flow rates are probably significant parameters in liquid cooling,
but we don’t know quite enough to say whether these numbers are good enough
for the ~400W GPU + CPU loop we are considering. The maximum liquid temperature
specification of 60°C gave us pause. We’ve routinely seen GPU temperatures
at >80°C at full tilt in our stress testing of both video cards and the
recent gaming systems. If the pump in this system isn’t rated for operation
with water hotter than 60°C, given that temperature in a liquid loop will
tend to equalize over time, does this mean we need to keep the GPU and CPU temperatures
below 60°C? The airflow needed for such a level of cooling could mean really
high fan RPM, and ultimately, a lot more noise than we want.
Putting this thought aside, the reservoir/pump was connected to the ASUS P8Z77-V
Pro that we’ve been using as a fan test platform for a couple years. The pump
is powered by a legacy 4-pin Molex connected directly to the PSU, and a second
lead is connected to the PWM CPU fan header of the motherboard. SpeedFan was
used to control the speed of the pump; the control arrangement worked just as
intended. The reservoir/pump was placed atop a soft piece of foam to prevent
the benchtop from resonating and adding to the pump noise.
The acoustics were not good. As we noted many times in the past, water pumps
for PC component cooling have a considerably worse acoustic signature than typical
DC axial fans used in computers. High tonality, especially in the upper registers
of the audio spectrum, combined with high vibration makes the VisionTek EK EK-SBAY
DDC 3.2 PWM reservoir-pump a very difficult component to use in a quiet PC.
Even at the slowest setting, it sounds reminiscent of a fan with ball bearings
about to fail — there’s a screechy quality that pervades, albeit at much
lower SPL (sound pressure level) than at higher speeds. We could use the elastic
suspension mechanical decoupling method that has served so well for HDD vibration
damping for years, but the screechy quality was so prevalent that even the damping
afforded by being wrapped in my hands did not change it. The absence of liquid
in the reservoir and pump may have exacerbated the noise, as water has acoustic
damping properties, but we were doubtful that it would have more than a marginal
Laing DDC-3.2 pump in VisionTek EK EK-SBAY DDC 3.2 PWM reservoir-pump
has PWM controllable speed range of 1225~4485 RPM.
see the high tonality of the acoustic signature.
When set to the lowest speed, initially, the SPL is just 18 dBA@1m, but
after 20 seconds or so, the prominent screeching indicated by the peaks
450, 700 and 900 Hz begins, and the SPL actually rises to the same level
as at full speed. There may have been bearing damage here. The only way
to stop this noise was to press down hard with both thumbs directly atop
It was decided this is not a device we can effectively damp enough to use in
our quiet PC. And while it could have been an issue specific to our sample,
we did not want to waste time trying out more samples of the same model pump.
Better to jump ship early.
NOTE: The cause of this screeching noise
was identified. It is a resonance caused by the motor spinning in the empty
water reservoir. Filling it with water completely eliminates the noise. My apologies
for any negative effects of my original report on this reservoir/pump to EK,
VisionTek and/or Laing. See the Postscript dated 15/01/2015 and followed up
27/01/2015 on the last page of this article.
SEARCH FOR A QUIET PUMP
Our search for a quiet pump quickly led us to this article on Building
a Quiet Water Cooled Gaming PC at Codesociety.com. In it, James (the author)
“I wanted to get the quietest pump possible, so the top contenders are
the various D5 Vario pumps, which are said to be among the quieter pumps available.
The top choice was initially the Koolance PMP-450 with the COV-RP450 top based
on this Martin’s
Liquid Lab Pump Noise Testing. This pump happens to be the same as the
Swiftech MCP655, just with a different model name. However, I found that Swiftech
also offered a PWM version of this pump, the Swiftech MCP655-PWM-DRIVE. Martin
also reviewed this pump, and found that it could be run at even slower
speeds to be even quieter than the PMP-450, giving it the noise crown. Thus,
I had to choose it.”
Martin’s review of the MCP655 PWM showed it has a speed range of 800~4800 RPM.
Physically, it seems to be a larger pump, with a power rating of 37W, although
Martin’s highest measured power draw during testing was only 21W (at 12V). Martin
went through an impressively comprehensive set of tests at different speeds
and conditions, a bit like SPCR’s fan/cooler tests where performance is measured
at multiple RPM. The resulting data for pressure, flow rate, efficiency, etc
are somewhat overwhelming. His acoustic measurement are underwhelming, however,
as the lowest SPL mentioned in any of his reviews of fans and pumps is around
32 dBA, with highs up beyond 60 dBA, suggesting a high ambient noise environment,
measuring distance closer than 1m, and a sound level meter with 30 dBA minimum
sensitivity. Still, the work Martin has done to quantify noise/performance is
extensive and impressive. With articles like “Why
Static Pressure & Max Flow Specs Are Poor Measures of Fan Performance“,
we appear to be in general agreement on fundamentals.
It’s also notable that Martin enthusiastically espouses pump decoupling:
“Pump decoupling has HUGE benefits, up to 15 dbA lower noise levels
than pumps without decoupling… In general, all of these pumps are extremely
silent pumps when decoupled. Most general users with fans over 1000 RPM would
typically have a hard time hearing these pumps at all in a normal fan noise
masking environment. Critical noise folks using ultra slow speed fans should
put extra emphasis into pump decoupling methods and undervolting or reducing
This tells us that we are definitely on the right path.
Moving on… Martin pronounced the MCP655 PWM to be an “awesome
pump” and the “King of Pump Silence (The most silent
pump tested to date)“. And how many pumps has Martin tested? It’s
hard to tell as his site navigation is a bit convoluted, but a safe bet is at
least 20. Obviously, Martin is a much better position to judge than we are,
and we’re not about to examine 20 pumps to verify his findings just for this
Swiftech MCP655 PWM
A query to our contacts at Swiftech led us to Frozen CPU, apparently the only
retailer to offer the Swiftech MCP655 PWM pump. Frozen CPU generously provded
a sample of the pump installed in a XSPC Dual 5.25″ Bay Reservoir V2. This
reservoir looks like it has higher volume capacity than the EK-SBAY.
The XSPC Dual 5.25″ Bay Reservoir V2 looks and feels nicely built,
with a brushed anodized aluminum fascia. It also comes with an LED that
can be plugged into the back to illuminate the front water level window.
The XSPC reservoir + MCP655 PWM pump was tested just like the EKWB/Laing DCC
3.2. Unfortunately, we ran into a hiccup. This pump may not have a speed sensor;
RPM could not be reported by either SpeedFan or ASUS FanXpert. The speed could
be controlled, however, with SpeedFan. Based on Martin’s reports,
here is an approximation of RPM for SpeedFan PWM % settings for this pump, along
with our measured SPL
MCP655 PWM Pump
There’s some very high frequency tonal noise centered around 17 kHz, but
The results were favorable. The overall sound was somewhat tonal but without
the screechy quality that plagued our sample Laing DCC 3.2 PWM pump. The noise
range from minimum to maximum was quite small, about 8 dBA, roughly 18~26 dBA@1m.
Vibration is definitely present, worse than any hard drive we’ve tested recently,
but since we don’t have to deal with lots of airborne noise, mechanically decoupling
the pump should take care of it. The SPL declined fairly linearly with PWM,
to around 40~35%. At that point the SPL was 18~19 dBA@1m, very modest and slowing
the pump any further had no effect on noise. All good: We found a pump we can
work with. Interestingly, the specified maximum temperature for this pump is
also 60°C, although it doesn’t specifiy whether this refers to the liquid
We hunted for a suitable large case that could house the 403 x 131 x 39 mm
Coolstream PE 360 Triple radiator. This was tougher than it first seemed.
Our short list was based on brand, discernable case details, and sample availability.
All are medium or larger ATX tower cases. We did want to avoid the really huge
cases like the Coolermaster Cosmos II — basically anything so big that
it could be a casket for a Rottweiler.
A sample of the Fractal Define R5 was on hand, and a quick look at the space
between the top edge of the motherboard and the top panel mounting place told
us the 74mm depth of our radiator + fans would be a very tight squeeze. A 2-fan
radiator would fit better, but a 3-fan radiator would make working around the
CPU area very tight. It was set aside.
A NZXT Source S530 case sample came in time for this article. It is a fairly
plain, straightforward case with lots of venting, but like the Define R5, there
is not enough space above the top edge of the motherboard for comfortable installation
of the radiator and fans. Like the R5, a-2 fan radiator would work well on top,
but it’s just too tight for a 3-fan radaitor.
Both Corsair Obsidian 750D and Graphite 760T look like they would have been
very suitable cases for the system being planned, with plenty of room for the
radiator and its fans. Neither of the Corsairs were actually requested early
enough for samples to come in time for this article, unfortunately.
Not nearly as big as the Phanteks Enthoo Prime which we reviewed in late 2013,
the Luxe looked like worthy contender. A sample did arrive in time, and the
Phanteks Enthoo Luxe case proved not only suitable for a 3-fan radiator on top,
but also extremely nicely built with some excellent features. All in all, it
is a very nice case. The Enthoo Luxe became the case of choice for our first
watercooled gaming rig.
PHANTEKS ENTHOO LUXE
The Enthoo Luxe is a 70 liter windowed case with extensive support for
liquid cooling. The white painted finish is attractive.
It’s obvious from the back panel that the exhaust fan position is adjustable,
The power button is on top, with reset switch and other I/O on a hidden
Some of the exterior is made of plastic; all of the interior is fairly
The backside of the motherboard tray is inset some 2.5cm from the right
View with drive cages removed. A large 200 x 25mm intake fan lurks in
This photo shows the inside of the front panel with the dust filter removed.
Since a big portion of this project involved many components entirely new to
us, we decided to use known and tried components as much as possible for the
rest of the system. There was no point in complicating the build any more than
it already was.
CPU: Intel i5-4790K
Intel’s superior energy efficiency takes us back to their Haswell processors
time and time again. Socket LGA1150 processors and boards represent the best
current balance price/performance for gaming. The latest top dog among socket
LGA1150 processors is the i5-4790K, a titch quicker than the i7-4770K
which we reviewed a while back, and generally sells for about the same price.
If you’re on a budget, consider than a similarly clocked dual core provides
about the same performance with most games, as they still don’t take full advantage
of multiple cores.
ATX MOTHERBOARD: ASUS Z97-PRO
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 well-reviewed
ASUS Z97-PRO has been solid for our
other ATX builds this season. It gets pressed into service again here.
OS DRIVE: Crucial MX100 512GB SSD
The Crucial MX100 512GB remains a standout choice for its combination of performance,
reliability and low price. Our
review confirmed the consensus of many other tech reviews. We’ve included
it in several other gaming builds; it’s reprised here again.
Crucial MX100 512GB remains a standout SSD.
As in many of our other gaming builds, we did not include a large-storage HDD
for the sake of simplicity and brevity, but in the big Phanteks Enthoo Luxe
case, it is easy to add one or more quiet HDDs with hardly any effect on overall
noise. The lower HDD cage is the best spot, or if you’re into the lowest possible
noise, elastic suspension in the optical drive cage, as
we demonstrated most recently in our ATX Gaming build. Our standard recommendation
is to stay away from 7200 RPM and faster spinning drives. The performance advantage
over the much quieter, lower vibration 5400 and 5900 RPM drives is really small
if your OS, programs and frequently accessed data is all on the SSD. If you
have a need for a scratch disk, virtually any decent 240GB SSD will fit the
The 7200 RPM WD Se is faster, pricier and noisier than the Red.
This system is likely to be the most power demanding of all the gaming builds
we’ve assembled in the past couple of months. All the other systems have topped
out thus far at under 300W AC or roughly 270W DC, given the use of ~90% efficiency
80 Plus Gold or Platinum PSUs. This build is likely to push a hundred watts
higher. And if the CrossFire build we’re considering comes to fruition, we have
to be prepared for upwards of 600W, due to the ~300W power demand of an AMD
R9 290. While we haven’t reviewed lots of higher power PSUs recently, SPCR keeps
close tabs on developments in PSUs and have a good handle on the best options
for low noise power delivery. A handful in the 800W~1000W will do the job equally
well. All three of the modesl listed below feature semi-passive fans that start
spinning only at higher load, extremely high energy efficiency 80 Plus Platinum
rating, modular cabling, strong reputation for quality, and emphasis on very
Corsair HX850i promises super quiet efficiency.
It was a bit of eenie meenie minie mo… and in the end, the Corsair HX850i
was chosen for broad availability and price.
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
has been solid for us, and it sports lower profile heatspreaders that don’t
get in the way of big heatsinks.
Installation began outside the case, with mounting of the VisionTek/EK Supremacy
heatblock atop the CPU in the Asus Z97-Pro motherboard. This was a cinch, as
the parts are very well made and easy to use.
Radiator positioning was dictated by the mounting holes available on the radiator
frame, the space in the case, and the screw holes in the mounting flange of
the top panel. It seemed best to keep the tubing away from the back where it
could interfere with access to fan headers and the AUX12V power connector on
the motherboard. So the radiator was mounted on the bottom side of the mounting
flange, with screws going down into it from the top, the tubing valves in front.
Installing the graphics card was just a matter of plugging it into the slot
and tightening a couple of screws, but the reservoir/pump posed more of a challenge.
It could be mounted as intended, in two optical bays, as shown below.
In previous discussion, I mentioned that this reservoir/pump has fairly high
vibration, and would be best mechanically decoupled for lowest noise. A check
was done with just the pump powered up by the power supply (whose fan does not
spin at such lower load) and controlled via PWM in SpeedFan from an external
test bench system.
The SPL with the pump at 35% PWM speed was 33~34 dBA@1m (!!!), characterized
by a nasty droning tonal hum as well as higher frequency overtones. That’s an
increase of 15 dBA over the 18~19 dBA pump-on-foam measurement I obtained
earlier. Subjectively, this is about three times louder
compared to the mechanically decoupled sound, and completely unacceptable for
a quiet PC build.
Thus ensued a study of placements of where the big XSPC dual-bay reservoir/pump
could be suspended with elastic string. It was important to get this right,
as the tubing needed to be cut to length for the particular positions of the
water cooled components. It would be a major pain to adjust this after the tubing
is cut, fitted and liquid filled. In the end, the chosen location is in the
upper half of the airflow path of the front intake fan, where the upper HDD
cage normally goes. Both HDD cages were removed for ease of installation and
would remain out, as no HDD was planned.
Tubing runs were simulated multiple times before any cutting was done, then
cut and installed one by one, to ensure that the tubing did not kink anywhere.
The VisionTek/EK seal compression fittings proved to be very secure and easy
to use. All the watercooling parts have ports marked for in and out flow, and
this flow directionality was followed to the letter. The flow goes: Pump to
radiator to CPU to GPU and back to pump.
Fitting the tubing.
With all the tubing and connections triple-checked, the 100ml of blood-red
liquid from VisionTek/EK was mixed with 900ml of distilled water as instructed.
When mixed, the liquid color turned bright red. The XSPC reservoir was filled,
then the cap closed, and the pump turned on by itself without the PC powered
up. The liquid began showing up in the tubing. The liquid level in the reservoir
naturally dropped quite a bit, but then the pump seemd unable to push the liquid
any further. The reservoir was topped up with more liquid, and the process repeated.
The system kept sucking up more liquid over several repetitions of this process,
until a total of about 700ml was used. At that point, the liquid was moving
through the system quite smoothly with some occasional bubbling. It was left
running in this state for about an hour, at the end of which time the reservoir
was topped up a bit again.
It seems that any air bubbles trapped in the system eventually get released
in the reservoir. Within a couple of hours of system power on, there was hardly
any bubbling noise. Keeping the pump running continuously seems the best way
to prevent bubbling noises if they bother you. If you prefer your system powered
down or to sleep when it’s not being used, you’ll need to add something like
a 30W 12V DC power supply to keep that pump running.
The red contrasts nicely against the black and metallic colors
of the Asus Z97-Pro and the white of the Phanteks Enthoo Luxe case.
Measurement and Analysis Tools
The three fan models and the pump were analyzed after the system was assembled,
plugging in their leads to our Asus P8Z77-V Pro fan testing platform and using
FanXpert 2 or SpeedFan. So the fans and the pump were measured installed as
shown in the photos on previous pages.
The two 140 x 25 mm Phanteks fans supplied with the case are also 3-pin
The VisionTek/EK/Gelid 120 x 25mm fans that came with the radiator has
The fans were measured for noise at various speeds. Most important is the speeds
at which each of these fans becomes quiet enough, remembering that there are
three of the 120mm fans, which generally means about a 4~5 dBA@1m increase over
just one fan. Those 120mm fans are also on the radiator, whose airflow impedance
will cause some additional noise from turbulence, especially as RPM is increased.
RPM vs SPL – Noise Makers in WC Gaming PC #1
% = setting of fan speed in SpeedFan
SPL = Sound Pressure Level in dBA@1m
In the table, I highlighted the most pertinent fan speeds and SPL. It is somewhere
in this range that the fans and pump will need to be set in order for the system
to be near 20 dBA@1m total SPL. Interestingly, suspended in the closed case,
the pump is audibly and measurably quieter than when sitting atop soft foam
on the test bench. Obviously, the case is blocking some of the airborne noise;
it sounds less tonal, with fewer high frequency components.
CONFIGURATION OF FANS
There is no question that the top fans on the radiator will be used for exhaust.
Blowing the radiator heat down into the case would be silly. So what about the
front and back panel fans? Since the radiator is where most of the heat in the
system will end up, the main task of any fans not directly on the radiator should
to ensure as much cooler air to the radiator as possible.
A front case fan is almost always configured as an intake, as it is here. Unfortunately,
this 200mm fan isn’t particularly quiet at any speed above ~500 RPM. Our sample
has a chuggy, clickety sound through most of its speed range reminiscent of
some antiquated industrial machine, and a major droning whoosh at full speed,
even though it spins at only 760 RPM. At <500 RPM, it’s decently quiet, but
the airflow is very modest. This fan was plugged into a motherboard header that
AIDA64 identified as Chassis #4.
The back panel fan is usually used for exhaust, but since there are three 120mm
fans blowing air out on top, it is probably better as an intake. The front fan
isn’t going to be drawing in that much air, so a little help from the back would
be a good thing. This Phanteks 140×25 fan is also not not that quiet, unfortunately,
with a decidedly whiny quality at speeds above ~900 RPM. It’s not until below
800 RPM that the sound gets innocuous. This fan was plugged into the motherboard
header that AIDA64 identified as Chassis #2.
The 120mm Gelid fans on the radiator are no champs, either. They have a distinct
midrange tonality at higher speeds. Like the Phanteks 140mm, it’s only below
800 Hz that they sound OK. These three fans were plugged into the fan hub in
the Enthoo Luxe case, and the hub was plugged into the motherboard header that
AIDA64 identified as Chassis #3.
Yup, these fans are going to be a major challenge to keep quiet enough for
a system I’d be happy with. But feeling the pressure of time, I preferred to
try my luck with them and swap them out only if absoluetly necessary.
SpeedFan was used to set all the fan and pump speeds, and Asus FanXpert was
uninstalled entirely to avoid any possibility of conflict between the two programs.
For details on how to set up SpeedFan properly, please see Larry Lee’s SpeedFan:
A Guide to Universal Motherboard Fan Control.
THERMAL & ACOUSTIC PERFORMANCE
Config #1: Radiator Fans at Full Speed
To begin, let’s check system temperatures and noise with the radiator fans
at full blast, and the pump at 100% speed. The front fan is at 600 RPM, the
back fan at 720. Furmark is running at maximum stress/power level, and the CPU
is stressed Prime95x4 on torture test. Room temperature was around 21°C
when this test started, and idle temperatures of both CPU and GPU were absurdly
low, near ambient.
All the temperature sensors started rising when the tests loads were applied,
but slowly. In fact, it took the better part of an hour for temperature to stabilize.
AC power draw began at about 435W and slowly climbed with temperature to 445W
maximum. Given the 92%+ efficiency of the Corsair HX850i PSU, this translates
to ~410W DC. The PSU fan may or may not have been running; it was simply not
audible as a discrete source. The image capture from AIDA64 below shows all
the temperature sensor details, but to summarize, the CPU hit 55°C, and
the GPU reached 51°C. By then the room was about 3~4°C hotter, so the
temperature rise was just 30°C and 26°C.
You have to remember that our test load is a constant high load, higher than
any game can reach (by a margin of at least 10%, according to many sources and
by our own experience). Even the most demanding computer games don’t demand
unvarying power, as the power demand goes up and down depending on the action,
and even the skill of the player(s). What this tells us is that under actual
use loads, this system will never see the kind of temperatures reported here,
for any of the fan/pump speed settings. (Assuming your room temperatures are
not typical of subtropical summer.)
The SPL of the system with the radiator fans at full speed is 34 dBA@1m.
SPCR regulars would not call this quiet but many mainstream PC gamers
would consider it perfectly OK. Temperatures are amazingly low.
Best Low Noise Configuration
It took most of a whole day of iterative adjustments, and another couple of
vent grill/fan mods, but I finally managed to get the overall noise way down
with a smooth acoustic signature, and still keep component temperatures to modest
levels. After 51 minutes of extreme system load, both CPU and GPU sensors stabilized
at 66°C and the measured SPL was just 18 dBA@1m.
Just 18 dBA@1m SPL for a 445W system with GPU and CPU temperatures
at only 66°C!
To achieve this result, the pump was set to 35%, the three 120mm radiator fans
were dialed down to 660 RPM, the back fan to 500 RPM, and the front fan to just
330 RPM. The front fan doesn’t always start immediately at this speed but usually
comes to life within a minute or so. It’s not a critical fan; turning it off
all together didn’t raise temperatures by more than a few degrees. To be safe,
it is better set to around 380 RPM. Any increase in noise is under 1 dBA and
Then there are the vent grill mods. These were part and parcel of the work
done to achieve the final results above.
MORE FUN WITH GRILLS & DUST FILTERS
I made a few comments earlier about the mutiple layer airflow impedances on
the vent openings. The one that bothered most me was the top grill/vent. It
has three layers, including a fine vinyl bug-screen type middle layer, completely
unnecessary as this is an exhaust vent. If you read my article Quiet MicroATX
Gamer #1, you’ll remember that the cooling/noise challenge of the Corsair
240 Air case used for that build was solved eventually by the removal
of dust filters. I decided to examine what effect the top vent grill of
the Enthoo Luxe has on airflow.
First, I measured the airflow from the front top fan set at full speed, with
the grill cover on: 190 feet per minute. Then I popped the grill off
and repeated the test, taking care to position the anenometer vane at the same
point and distance from the fan: 460 feet per minute. Wow! That’s 2.4
times more airflow without the grill. Now, if you’re a regular at SPCR you know
that we don’t take much stock in airflow, it’s a misleading specification, much
like horsepower by itself for engines. The relationship between cooling effectiveness
and airflow is far from linear. But still, a 2.4 times difference is too much
to ignore. So, off came the bug screen filter.
6803 Anemometer is an expensive ($845 today) precision
instrument able to read as little as 40 FPM air velocity to as high as
7800 FPM with 1% accuracy.
Without the dust filter in the grill, air speed improved to 270 FPM. That’s
much better than 190 FPM, but still a long way from 460 FPM. So the metal grill
had to go. This was slightly more work: Some tabs had to be bent back. Off came
the steel perforated grill. All that’s left is the plastic inner frame with
its elongated hex holes. With that plastic “grill”, air speed went
up to 400 FPM. A loss of 60 FPM seems acceptable… and the hex grill does offer
a bit of protection for the fans.
This change to the top vent grill improved cooling by at least 5°C —
the improvement varied a bit with fan speed. It is an essential part of the
system tuning that allowed the best low noise config acoustic/thermal results
Final state of top grill for optimal noise/cooling.
Finally, I removed the dust filter from the front panel. This improved airflow
through the front by at least 25%. Yes, it is an intake, and if the system resides
on the floor, then you’ll probably want to keep this dust filter on, speed up
the fan a touch and clean the filter often.
FINAL BUILD PARTS LIST
The core system parts come to around $1500, not exactly a budget build, but
the quality and performance of the components is commensurate with price. You
could shave about $250 by spending less on the CPU, motherboard, case and PSU
without affecting gaming performance or acoustics. A smaller PSU, however, might
impinge on upgrading to CrossFire with a second R9 290.
The liquid cooling components are pricey, totalling about $400, and this doesn’t
include the $100 value full-cover EK waterblock that comes preinstalled on the
CryoVenom R9 290 Limited Edition (which at $360 with full warranty and no
installation hassles, is a steal, considering a stock 290 currently runs around
$300). It’s about what you will pay, regardless of exact brands and components,
if you seek similar quality and performance, unless you hit a fire sale. All
of these components are easily transferrable to your next system or upgrade.
With a bit of care and maintenance, they should last many years.
Care and maintenance — it actually begins before assembly with ensuring
the liquid path is clean and free of debris, using distilled water and coolant
to retard any biological growth over time. Like a car engine, the coolant should
be flushed, the system cleaned and replenished periodically. Standard advice
is to do this every six months. Some say a complete breakdown and rebuild should
be done once a year… though that sounds really OC to me. YMMV. Of course,
basic things like leaks and coolant level need to be checked routinely as well.
With good parts and careful initial preparation and installation, liquid cooling
isn’t that much more work than air-only cooling, where dust cleaing is the only
real maintenance that is required. (Check the water cooling resources linked
on the third page of this article for detailed information on maintenance and
do you own web searches.)
SPCR’s Quiet Liquid Cooled Gaming PC Component List
CryoVenom R9 290 Limited Edition
|Intel Core i5-4790k|| |
|ASUS Z97-Pro|| |
|Kingston HyperX 4GBx2 1866Mhz DDR3 RAM|| |
HyperX Savage 8GB Kit (2x4GB) 1600MHz DDR3
|Crucial MX100 512GB SSD|| |
840 EVO 1TB
Samsung 850 Pro 512TB
|Phanteks Enthoo Luxe|| |
|Phanteks Enthoo Pro |
Corsair Obsidian 750D
Corsair Graphite 760T
|Corsair HX850i|| |
|Seasonic Snow Silent 1050W Platinum |
bequiet! Dark Power Pro 10 850W Platinum
Liquid Cooling Components
EK Supremacy Universal Nickel-Plated CPU waterblock
|Swiftek Apogee XL|
|XSPC Dual 5.25″ Bay Reservoir |
with Swiftech MCP655 PWM Pump
|Swiftech MCP655 PWM Pump with |
|VisionTek EK CoolStream 360 |
|Other quality >360mm radiator|
|Compression fittings (8), tubing |
(1.5m), 3 high quality 120mm fans, coolant concentrate, etc.
CryoVenom360 Liquid Cooling Kit – $320
Western Digital Red 4TB
Western Digital Green 4TB
Seagate NAS HDD 4TB
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.
SPCR’s first liquid cooled gaming PC build guide is also the first discrete
full-bore liquid cooled system we’ve assembled. This PC building project began
in late October and did not end till the start of January. The process of picking
the right components for the system was critical and added to the total time
of the project. Armed with a pump reputed to be the quietest in the PC water
cooling scene, top notch water blocks from VisionTek EK on the CPU and the Radeon
R9 290 GPU, a huge 40cm long 3-fan radiator and a bag full of SPCR tricks, we
managed to assemble a powerful gaming system that stays cooler than any other
gaming system we’ve built before, despite being one of the most power hungry.
The system stays at a constant, smooth and quiet 18 dBA@1m under any real world
loads. All this despite a system power profile of 445W, which is a pretty hefty
A little LED lighting for geek appeal.
The system was set up for constant fan and pump speeds, which explains the
unvarying acoustic level. Many users actually prefer constant low noise to a
system that is quieter at idle but ramps up to a higher level at load. It’s
the change in sound that many people get more bothered by once overall noise
is at a quiet level. If we had more time to spend on this project, we could
probably have devised a modern automated cooling system that varies fan (and
pump) speed in response to thermal conditions, but with enough hysteresis to
Logical and efficient airflow design, low impedance in/out vents, and precise
fan control are all essential parts of this build. Probably the single most
important element for this low noise system is the mechanical decoupling of
the water pump by elastic cord. The 15 dBA@1m improvement wrought by our time-proven
elastic suspension over conventional hard mounting of the pump is impossible
to obtain in any other part of the system. Without this single technique, the
vibrations of the pump would force the noise floor of the system up to 34 dBA@1m.
Most other noise producing components you could add would barely be heard over
Another interesting lesson is that low liquid flow rate need not be an obstacle
to good cooling, even with very low speed fans, as long as the airflow for those
fans is as unrestricted as possible. The pump PWM speed was set to just 35%
(which apparently results in about 1000 rpm; no pump speed sensor, so this is
based on Mark’s report) for the final configuration.
One thought about the position of the pump in the liquid loop I built: It receives
the liquid after coursing through the CPU and GPU heatblocks, then pushes the
liquid on to the radiator. Given that the pump should be kept running cooler
than 60°C, a better place in the loop might have been between the outflow
from the radiator and the inflow to the CPU/GPU. That position in the loop is
where the water temperature might be the coolest. (Unless of course, some better
informed, more experienced reader assures me that the liquid temperature in
the loop always reaches equilibrium.)
Swapping to quieter smoother sounding fans right from the start might have
resulted in a slight quieter or cooler system, but the end result is excellent,
and we’re probably talking about maybe one dB, perhaps two at the most. But
nicer fans might be worthwhile if you’re interested in lower temperatures or
live in an environment consistently hotter than SPCR’s moderate 20~27°C
range through the seasons. I regret not checking whether the sample Laing DDC-3.2
pump in the VisionTek EK-SBAY reservoir/pump was damaged. I will check this
out and retest if damage was responsible for the poor pump acoustics.
I can only marvel at the VisionTek
EK water blocks received for this project. They are so professionally designed
and executed with such high quality materials and workmanship that they make
most air cooled products seem rough and unfinished. If all liquid cooled products
exhibit such high standards, little wonder that a contingent of PC DIY enthusiasts
remain dedicated to liquid cooling. As a cooling solution for hot GPUs, VisionTek
EK products are certainly great options. As long as the motherboard remains
frozen in a form factor never meant for video cards that exceed the power and
thermal profile of CPUs, liquid cooling will continue to attract.
My sincere thanks to the other sample suppliers, without whose products this
project could not have been completed: The Phanteks
Enthoo Luxe is a phenomenally well-designed and sturdy chassis for a high power
system, especially if a large water radiator is in the works, my quibbles about
dust filters and grill covers notwithstanding. The Corsair
HX850i did a fantastic job of simply delivering the power without making even
a peep; in this case, it never became an identifiable source of noise. Much
thanks to Swiftech and
for contributing the marvelous MCP655 PWM pump and XSPC Dual 5.25″ Bay
Reservoir V2. Hats off to Intel
for the CPU, ASUS for
the Z97-PRO, Kingston
for the RAM, and Crucial
for the MX100 SSD.
Coming soon: A Crossfire version of this liquid
cooled system featuring an additional Radeon R9 290X.
* * *
NOTE: Overleaf, you’ll find two detailed
Postscripts on pump acoustics. The first is on the cause of the screeching noise
of the VisionTek/EK/Laing reservoir/pump that swayed me to seek out an alternative
pump. It turns out to be a resonance caused by the motor spinning in the empty
water reservoir. Filling it with water completely eliminates the noise. The
second Postscript is a comparison of acoustic noise between the pumps discussed
in this article. My apologies for any negative effects of my original report
on this reservoir/pump to EK, VisionTek and/or Laing.
Articles of Related Interest
Gaming Build Guide #3: BitFenix Prodigy Edition
Quiet Mini-ITX Gaming Build Guide
#2: NCASE M1 Edition
Quiet Mini-ITX Gamer
Quiet ATX Gamer, R5 Version
SPCR’s Quiet ATX
Gaming Build Guide
Silent Mid Gaming PC Build Guide
Basics & Recommendations
* * *
This is a followup in two parts to the question of pump noise first raised
by the screeching noise exhibited by the EK-SBAY DDC 3.2 PWM in dry testing.
The first part explores the cause of that noise; the second part presents acoustic
measurements for the pump in wet operation, and compares it to the Swiftech
MCP655 PWM Pump used in the build guide.
|POSTSCRIPT ONE – 15 Jan 2015: EK-SBAY DDC 3.2 PWM |
I solved the mystery of that screeching noise. It took 15 mins; should
1. Power up the EK-SBAY DDC 3.2 PWM pump/reservoir again to confirm nasty
So the screeching is an acoustic resonance
Finally, having examined the electric motor & the pump attached to
Later, on the same day…
Prompted by more forum comments, I took the pump off again and took another,
The pump, removed from the housing, without the cover over
When it is running normally, the metallic cavity in which
|POSTSCRIPT TWO – 27 Jan 2015: |
EK-SBAY DDC 3.2 PWM vs XSPC Reservoir + Swiftech MCP655
It took a while, but the extra compression fittings and
As I suspected, there was no untoward bearing damage that
So let’s get it out there now: To say I have egg on my face
Not only is the EK-SBAY DDC 3.2 PWM much quieter
The reported PWM control range of the Laing DDC
Compare this with the measured SPL obtained for the Swiftech MCP655 PWM
Now keep in mind that the MCP655 PWM Pump was measured while
Suffice it to say that…