• Home
  • blog
  • Scythe Musashi Dual Fan GPU Cooler [Postcript Added]

Scythe Musashi Dual Fan GPU Cooler [Postcript Added]

blog image

The Musashi is a long, slim, dual fan GPU cooler from Scythe. We installed it on an ATI HD4870, and put it through a grueling torture test against our old favorite, the Arctic Cooling Accelero S1. POSTCRIPT added Nov 29, 2009. See page 8 for details.

Scythe Musashi

September 28, 2009 by Lawrence Lee

Product Scythe Musashi
GPU Cooler
Manufacturer Scythe
MSRP ~US$40

POSTSCRIPT: Scythe Musashi Dual Fan GPU Cooler, Updated added Nov
29, 2009 (page 8)

GPU cooling is a challenging task given the way video cards are positioned
in typical ATX systems and the fact that some modern GPUs can require more power
than quad core desktop CPUs. Graphics card circuit boards are also getting longer
and more complex with extra ICs that require cooling as well — cooling the
core alone may not be enough. With the way things are, enthusiasts no longer
flinch when they see a VGA cooler with two fans like the Scythe Musashi.


The box.

The Musashi takes its name from a famous Japanese swordsman who not only claimed
numerous victims in his various duels, but also developed a fighting technique
using two swords. The Musashi GPU cooler’s two 100 mm fans are analogous to
these swords. Like a sword, the heatsink is fairly lengthy but slim. This makes
it a good prospect for CrossFire and SLI configurations.


Package contents.

The Musashi ships with a manual fan speed controller, mounting hardware, and
a myriad of heatsinks for memory and VRMs. If your card’s GPU core sits lower
than the frame surrounding it, you can use the included copper base to make
up the difference.

Scythe Musashi: Key Features
(from the product
web page
)
Feature & Brief
Our Comment
Wider Compatibility
This product is compatible for wide variety of graphic cards, for ATI Radeon
9000 to HD4800 class, and for nVIDIA GeForce FX5900 to 9600 class.
With five sets of mounting holes, the
Musashi is compatible with many older cards, though it some cases this cooler
might be worth more than the card.

Dual Independent Fan with Controller
Equipped with dual Scythe original super thin "“Kaze Jyu Slim
100mm" fan, users can independently control each fan speed to fulfill
the demand for silent to over-clocking purposes.

A nice feature as you may wish to run
one fan faster than the other.
Optional Single Fan Mode
Single fan mode can be arranged by simply detaching 1 fan and place another
fan onto the middle of the VGA cooler to intensively cool the GPU.
The ability to use one fan is dependent
on whether your card’s VRMs require active cooling.
Air-Through Construction
Airflow can be directed to the entire graphic card to ensure the complete
cooling of the graphic card.
Like our favorite, the Accelero S1, the
Musashi has an open design allowing direct airflow across the entire PCB.
Various Memory Heatsinks Included
In order to maximize the cooling performance, the various memory heatsinks
(chips) are included in the package.
Along with 16 ramsinks, Scythe also provides
VRM heatsinks of varying size.
Scythe Musashi: Specifications
(from the product
web page
)
Model Name:MUSASHI VGA Cooler
Model #:SCVMS-1000
Manufacturer:Scythe Co., Ltd. Japan
Dimension
(W x D x Thickness):
104 x 250 x 35 mm
4.09 x 9.84 x 1.38 in
Weight:375 g
12.1 oz
Heatsink
Dimension
(W x D x Thickness):
104 x 250 x 35 mm
4.09 x 9.84 x 1.38 in
Weight:251 g
8.8 oz
Fan
Dimension
(W x D x Thickness):
100 x 100 x 12 mm
3.94 x 3.94 x 0.47 in
Speed:800 ~ 2000 rpm
Noise:12.50 ~ 29.22dBA
Flow:11.45 ~ 27.60CFM
Bearing Type Sleeve Bearing
Compatibility
ATi:Radeon 9xxx

Radeon Xxxx
Radeon X1300
Radeon X1600
Radeon X1650
Radeon X1800
Radeon X1900
Radeon X1950

Radeon HD2600
Radeon HD2900XT

Radeon HD3650
Radeon HD3850
Radeon HD3870

Radeon HD4850
Radeon HD4870

* Not compatible with ATi RADEON 9550/9600

nVIDIA:GeForce FX5900

GeForce 6600
GeForce 6800

GeForce 7300
GeForce 7600
GeForce 7800
GeForce 7900

GeForce 8500
GeForce 8600
GeForce 8800GT
GeForce 8800GTS

GeForce 9600
GeForce 9800GTX

* Not compatible with nVIDIA GeForce 6600 AGP

PHYSICAL DETAILS

Some cards like the HD 4870
and HD 4890 require direct airflow
over the voltage regulators and a single fan doesn’t cover enough area to cool
both the VRMs and the GPU core effectively. A two fan solution seems to be optimal
for the graphics cards of today.


The Musashi is equipped with two slim Scythe Slip Stream 100
mm fans.


In classic Scythe fashion, the fins are thin (0.30 mm) and are
set a good distance (2.0 mm) apart from one another. Air passes freely
through the heatsink and over the components on the circuit board.


While most GPU coolers have heatpipes perpendicular to the board,
the Musashi’s run parallel along the entire length of its lanky body.
It is long enough to cover the entire PCB of most graphics cards.


The fans are attached with very tight-fitting wire clips.


The fan controller.

BASE & INSTALLATION

The most critical aspect of installation is for the heatsink to be securely
mounted. The more firmly it is installed, the better the contact between the
heatsink’s base and the GPU core itself. Ease of installation is also important
— a simple mounting scheme means less time spent installing, and a reduced
likelihood of screwing up.


Our victim: a Radeon HD 4870. The included memory and VRM heatsinks
are 8 mm tall and while the adhesive film is not that strong (we dislodged
a couple of them accidentally during installation), they will hold. Reusing
them a second time on a different card is probably out of the question.


The base is typical Scythe quality: nickel-plated, flat, and
with a dull finish. Installing the Musashi is very simple — in fact
it’s tool-less after you remove the stock cooler. The first step is fix
the baseplate with double sided bolts.


With the various heatsinks in place and thermal compound applied
to the core, drop the graphics card on top of the Musashi, place the backplate
over the bolts and use four nuts to secure it.


Clearance under the cooler is good — as there is room to
spare, it would have been nice if they made the memory/VRM heatsinks taller.


While we are impressed with the method, the actual mounting isn’t
that secure. The heatsink can be wiggled slightly — perhaps the bolts
could’ve been shaved down slightly to create a tighter fit.


Installed in our test system. The Musashi isn’t quite a two-slot
solution — the fan does overlap slightly with the expansion slot
beneath it. Even if it didn’t, installing a card there would cut off circulation.

TEST METHODOLOGY

Our test procedure is an in-system test, designed to determine whether the
cooler is adequate for use in a low-noise system. By adequately cooled,
we mean cooled well enough that no misbehavior related to thermal overload is
exhibited. Thermal misbehavior in a graphics card can show up in a variety of
ways, including:

  • Sudden system shutdown or reboot without warning.
  • Jaggies and other visual artifacts on the screen.
  • Motion slowing and/or screen freezing.

Any of these misbehaviors are annoying at best and dangerous at worst —
dangerous to the health and lifespan of the graphics card, and sometimes to
the system OS.

Test Platform

Measurement and Analysis Tools

  • CPUBurn
    processor stress software.
  • FurMark
    stability test to stress the GPU.
  • GPU-Z to
    monitor GPU temperatures.
  • SpeedFan
    to monitor CPU temperature.
  • A custom-built variable fan speed controller to power the system
    fan
  • PC-based spectrum analyzer:
    SpectraPlus with ACO Pacific mic and M-Audio digital
    audio interfaces.
  • Anechoic chamber
    with ambient level of 11 dBA or lower


A summary of how our video card/cooler test platform is put together can
be found here.

Our main test consists of FurMark stability test running in conjunction with
CPUBurn to stress both the graphics card and processor simultaneously. This
combination produces more CPU/GPU stress than a typical gaming session. As our
test system has very limited airflow, our results are not indicative of a real-world
situation, but rather a worse-case scenario. If the heatsink in question can
cool the card and its components adequately in this environment it means there
is some degree of thermal headroom when employed in a more conventional manner.
GPU temperatures are recorded using GPU-Z. On our HD 4870 test card, there are
three main on-die sensors, as well as three on the voltage regulators. We average
the results of each set of sensors.

The cooler is tested at various speeds to represent a good cross-section of
its airflow and noise performance. Noise is measured and recorded with the heatsink
installed in our test system with the side panel on. Our mic is positioned at
a distance of one meter from the center of the case’s left side panel at a 45
degree angle.

Ambient conditions during testing were 23°C and 11 dBA.

TEST RESULTS

Stock Fan Testing

Fan Measurements
Fan Voltage
Fan Speed
SPL@1m
(one fan)
SPL@1m
(both fans)
12V
1770 RPM
24 dBA
27~28 dBA
9V
1350 RPM
17 dBA
21 dBA
8V
1190 RPM
14 dBA
17 dBA
7V
1020 RPM
12 dBA
14 dBA
Fans tested on heatsink blowing downward, placed
on foam in open environment.

While the fans included are rated for 2000 RPM, our samples spun slightly slower,
topping out at approximately 1800 RPM. Like all Slip Streams, the noise character
was very smooth — at high voltages, the acoustics were dominated by air
turbulence. The included fan controller had a range of 600 to 1770 RPM, approximately
equivalent to 5V~12V. We consider 9V to be quiet for one fan, 8V for both fans.

Noise Level: In System (both fans)
Fan Voltage
SPL@1m*
Side Panel Off
Side Panel On
Side Panel & System On
12V
26 dBA
22 dBA
22 dBA
9V
20 dBA
16 dBA
17 dBA
8V
17 dBA
14 dBA
15 dBA
7V
14 dBA
13 dBA
14 dBA
*measured from the side panel

Inside our P180B case, the side panel blocks out much of the noise generated
by the two fans. The noise produced was actually fairly innocuous even at 12V.
Our test system is very quiet, at under 20 dBA@1m levels, it added only 1 dB
to the system’s SPL.

Performance Testing

Test Results: Scythe Musashi
Fan Voltage
SPL@1m
Avg. GPU Temp
Avg. VRM Temp.
CPU Temp.
12V
22 dBA
77°C
130°C
58°C
9V
17 dBA
82°C
144°C
58°C
8V
15 dBA
83°C
149°C
59°C
7V
14 dBA
86°C
156°C
N/A
Grey boxes indicate test failure.

The Musashi performed fairly well, keeping the GPU core at around 80°C
at 8~12V. At 7V however the card stopped outputing an image. After some investigation
we discovered the cause: the VRMs were overheating. The entire underside of
the ATI stock cooler acts as an enormous VRM cooler while the Musashi has only
three, small, independent heatsinks. As a result the VRM temperatures were a
lot higher than the stock solution, even if the main heatsink was able to cool
the GPU core more efficiently.

Accelero S1 Results


S1 mounted with two Nexus 92 mm fans attached using plastic twist-ties.

Due to the VRM problem encountered with the Musashi, it was clear that our
previous favorite, the Accelero S1 also needed two fans to cool both the GPU
core and VRM area. We equipped the S1 with a pair of Nexus 92 mm fans and we
also tried a couple of Scythe 100 mm fans (SY1025SL12M) as they are cheaper
and more widely available. The Musashi’s memory and VRM heatsinks were reused.

Accelero S1 (2 x Nexus 92mm)
Fan Voltage
SPL@1m
Avg. GPU Temp
Avg. VRM Temp.
CPU Temp.
12V
16 dBA
79°C
141°C
57°C
10V
15 dBA
82°C
151°C
58°C
Accelero S1 (2 x Scythe 100 mm)
10V
17 dBA
77°C

139°C

57°C
9V
16 dBA
80°C
146°C
57°C
8V
15 dBA
80°C
154°C
58°C

The S1 was able to cool our test card adequately and quietly using both sets
of fans. The Nexus fans, despite their slight size disadvantage delivered better
VRM temperatures at equivalent noise levels however. GPU core temperatures were
about the same.

Comparison: Average GPU / VRM Temperature
SPL@1m
Scythe Musashi
Arctic Cooling Accelero S1
2 x Nexus 92mm
2 x Scythe 100 mm
17 dBA
82°C / 144°C
77°C / 139°C
16 dBA
79°C / 141°C
80°C / 146°C
15 dBA
83°C / 149°C
82°C / 151°C
80°C / 154°C

The Accelero S1 installed with two full-sized 25 thick fans is a four-slot
solution, taking up almost twice the space of the Musashi. Despite the size
advantage, the Musashi holds its own. At the 15 dBA level, its performance is
equivalent to the S1/Nexus combination. At 17 dBA, it performs 5°C worse
than the S1 using Scythe fans.

MP3 SOUND 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.

These recordings are intended to give you an idea of how the product sounds
in actual use — one meter is a reasonable typical distance between a computer
or computer component and your ear. The recording contains stretches of ambient
noise that you can use to judge the relative loudness of the subject. Be aware
that very quiet subjects may not be audible — if we couldn’t hear it from
one meter, chances are we couldn’t record it either!

The recording starts with 10 seconds of room ambiance, followed by 10 seconds
of the VGA test system without a video card installed, and then the actual product’s
noise at various levels. 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.

FINAL THOUGHTS

The Scythe Musashi is a fairly impressive product overall. While it is fairly
long, its height and weight are on the smallish side. The fact that it can keep
up with the much larger combination of an Accelero
S1
and two full-sized fans is nothing short of miraculous. The fan selection
and fin design are both optimized for quiet cooling and for the most part, the
Musashi delivers just that. It will deliver better temperatures with less noise
than the stock solutions on most graphics cards. Installation is also extremely
easy, though the final mount is not as secure as we would have liked.

Unfortunately the hottest card it supports, the HD
4870
, requires heavy duty VRM cooling, and that’s where the Musashi
falters. Sure, our test system and methodology is more demanding than actual
gaming, but when we see VRM temperatures head north of 150°C and then cause
the graphics card to destabilize (as it did with the Musashi’s fans set to 7V),
we get more than a little uncomfortable. This is unfortunate because the slimness
of the Musashi makes it attractive for dual video card configurations for which
4870’s are commonly used. The included heatsinks are simply too small, though
to be fair, most other third party GPU coolers suffer from the same problem.

Zalman has tried to address the issue by selling stand-alone VRM heatsinks,
the ZM-RHS70
and ZM-RHS90,
designed to be used in conjunction with their VF1000
VGA cooler. Unfortunately both are too tall to fit under a Musashi or Accelero.
Thermalright also has a solution — awkward heatpipe coolers that extend
well past the PCB vertically, the VRM-R1/R2.
They are rather large and expensive though. A DIY solution may be the most feasible
option for some.

Scythe Musashi
PROS

* Easy to install
* Slim
* Very good GPU cooling
* Excellent fans
* Fan controller included

CONS

* VRM heatsinks too small
* Mount could be more secure

Our thanks to Scythe
for the Musashi sample used in this review.

POSTSCRIPT, Nov 29, 2009: Scythe Musashi Dual Fan GPU Cooler,
Updated (overleaf, on page 8)

* * *

Articles of Related Interest
HIS
Radeon HD 4890 Turbo Edition

Asus Radeon HD 4870 Matrix
Thermaltake Duorb VGA Cooler:
Are Two Orbs Better Than One?

Xigmatek Battle-Axe: First Direct-Touch
Heatpipe VGA Cooler

Arctic Cooling Accelero S2
VGA Cooler + Turbo Module

Arctic Cooling Accelero S1 VGA
Cooler

* * *

Discuss
this article in the SPCR forums.

POSTSCRIPT: Scythe Musashi Dual Fan GPU Cooler, Updated

Nov 29, 2009 by Lawrence Lee

After originally reviewing the Scythe Musashi, members of our forum mentioned
that the issue of the voltage regulators overheating could be solved by utilizing
the stock cooling plate. The fan and GPU heatsink can be removed on the reference
ATI stock cooler of the 4870/4890, and the metal plate underneath which cools
the memory chips and voltage regulators can be left on and used in conjunction
with third party GPU coolers.

Not only was this an excellent suggestion to solve the VRM problem, but it
would also improve our testing methodology. It’s a more convenient option as
we would not need to remove the seperate memory and VRM heatsinks if there are
any clearance issues — the stock cooling plate is as low profile as it
gets. In addition, it allows us to use a hotter card, the HD 4890, which we
originally abandoned in favor of the HD 4870 because its VRMs became so hot,
we could not cool it quietly with any of the heatsinks we tested.

So without further ado, we present the updated test results for the Zalman
VF1000, Accelero S1, and Scythe Musashi using the HD 4890 stock cooling plate.

Updated Test Results

Test Results: HIS 4890 Turbo Stock Cooler
Fan Voltage
SPL@1m
Avg. Core Temp
Avg. VRM Temp
Auto
26 dBA
82°C
94°C
Ambient temperature: 23°C
Ambient Noise Level: 11 dBA
(12 dBA with the test system on)

First up, for reference we tested the HD 4890 stock cooler. The system produced
26 dBA@1m when placed on load, with core and VRM temperatures registering at
82°C and 94°C respectively.

Test Results: Zalman VF1000 LED
Fan Voltage
SPL@1m
Avg. Core Temp
Avg. VRM Temp
12V
22 dBA
91°C
121°C
10V
22 dBA
96°C
133°C
9V
20 dBA
100°C
139°C
8V
19 dBA
102°C
145°C
7V
17 dBA
106°C
152°C
Ambient temperature: 23°C
Ambient Noise Level: 11 dBA
(12 dBA with the test system on)

We were previously unable to re-test the VF1000 at all on the HD 4870 due to
VRM cooling issues, but with the stock cooling plate, it wasn’t an issue at
least until the fan voltage was dialed down to 7V. As one would expect from
a single fan cooler, the VF1000 struggles to keep the HD 4890 cool at noise
levels of 20 dBA@1m and below, with the core temperature exceeding 100°C.
VRM cooling was particularly poor, almost reaching the failing point of 155~160°C
when the fan was set to 7V.

The VF1000’s acoustics are rough, due to the poor undertones of the fan utilized.
It is only quiet at 7V but close up, the character is dreadful. You may notice
that we measured the same noise level with the fan at both 12V and 10V — this
is not a typo. At 10V, the pitch of the fan increases dramatically resulting
in a similar A-weighted SPL as the fan running at 12V.

Accelero S1 (2 x Nexus 92mm 1500rpm)
Fan Voltage
SPL@1m
Avg. Core Temp
Avg. VRM Temp
12V
16 dBA
85°C
114°C
10V
15 dBA
91°C
128°C
Accelero S1 (2 x Scythe 100mm 1500rpm)
12V
21 dBA
93°C
116°C
11V
19 dBA
94°C
117°C
10V
17 dBA
96°C
123°C
9V
16 dBA
97°C
126°C
8V
15 dBA
98°C
132°C
Ambient temperature: 23°C
Ambient Noise Level: 11 dBA
(12 dBA with the test system on)

The Accelero S1 does a much better job, but cooling a HD 4890 is still a serious
challenge. With a pair of Scythe 100mm 1500rpm fans strapped to it, core temperatures
were in the 90’s. VRM temperatures were 20~30°C better than the VF1000 though,
so the advantage of having a second fan is clear.

Paired with Nexus 92mm fans, gave us an almost unreal increase in performance.
At the 16 dBA level it defeated the Scythe 100mm fans by more than 10°C
in both core and VRM cooling. At 15 dBA, the results were closer, but the dominance
of the Nexus fans continued. The acoustics of both sets of fans are beyond reproach.

Scythe Musashi
Fan Voltage
SPL@1m
Avg. Core Temp
Avg. VRM Temp
12V
23 dBA
85°C
109°C
10V
19 dBA
90°C
117°C
9V
17 dBA
93°C
125°C
8V
15 dBA
97°C
131°C
7V
14 dBA
102°C
142°C
Ambient temperature: 23°C
Ambient Noise Level: 11 dBA
(12 dBA with the test system on)

The Scythe Musashi delivered a slight improvement over the S1 paired with Scythe
fans. Its pair of slim 100mm fans are also very smooth with a benign character.

Comparisons

Test Results @ 19 dBA
GPU Cooler
SPL@1m
Avg. Core Temp
Avg. VRM Temp
Scythe Musashi @ 10V
19 dBA
90°C
117°C
Accelero S1 @ 11V
(2 x Scythe 100mm)
19 dBA
94°C
117°C
Zalman VF1000 @ 8V
19 dBA
102°C
145°C

At the 19 dBA level, the Musashi pulls slightly ahead of the S1 with Scythe
fans. The VF1000 which is deficient in acoustic quality to begin with, also
falls behind in performance, particularly if you look at the VRM temperatures.

Test Results @ 15 dBA
GPU Cooler
SPL@1m
Avg. Core Temp
Avg. VRM Temp
Accelero S1 @ 10V
(2 x Nexus 92mm)
15 dBA
91°C
128°C
Scythe Musashi @ 8V
15 dBA
97°C
131°C
Accelero S1 @ 8V
(2 x Scythe 100mm)
15 dBA
98°C
132°C

With limited airflow producing 15 dBA of noise, the S1 paired with Nexus fans
is an easy winner, beating the Musashi by 6°C in core temperature and 3°C
in VRM temperature. The S1 paired with Scythe fans performs similarly to the
Musashi — it would seem the Musashi is preferable with higher speed fans.

Final Thoughts

The Scythe Musashi acquited itself quite nicely, equaling the performance of
the Accelero S1 with Scythe 100mm fans with low airflow, and pulling ahead slightly
with high airflow, all while taking up less physical space. It cannot compete
with the S1 paired with smaller 92mm Nexus fans though — that combination
amazingly dominated the field. The superiority of Nexus fans continues to amaze
us.

The Zalman VF1000 isn’t in the same league as either of the aforementioned
heatsinks, but managed to keep the HD 4890 stable, though at fairly high temperatures.
With only a single fan, it’s out of its element. To cool today’s hottest graphics
cards, two fans is really the way to go.

It should be noted that even with the ATI stock cooling plate, VRM temperatures
of all the competitors were much higher than the 4890 stock cooler. Still, with
a good dual fan heatsink, you can run a HD 4890 fairly quietly while keeping
the VRMs about 30°C from their failure point, even in a restricted airflow
environment.

Fortunately, the release of the HD 5000 series has brought overall power consumption
down. If you plan on using one of these new cards, with the exception
of the 5870 or dual GPU 5970, its unlikely high temperatures will be even close
to a problem with a cooler of the Musashi’s caliber.

Discuss
this article in the SPCR forums.

Leave a Comment

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