It is not the first AMD Athlon 64 SFF barebones system from Shuttle, but it is the first to use a Socket 939 motherboard. We installed a 3500+ Winchester (90nm) core in a minimalist setup with a notebook HDD, played long in the BIOS and measured, tested and listened to bring you this review of a quiet, powerful system that is almost as small as the Shuttle Zen.
March 6, 2005 by Sean Boyd and Mike
Chin
| Product | Shuttle XPC SN95G5 |
| Manufacturer | Shuttle |
| Selling Price | ~US$300 |
The XPC SN95G5 is not the first AMD Athlon 64 barebone system
from Shuttle, but it is the first one to use a Socket 939 motherboard. This
is of particular interest to the SPCR community because of the high performance
and low power consumption 90nm Winchester core AMD Athlon 64 processors available in the 939 socket. By combining top-of-the-line processor support with the well developed
nVIDIA nForce 3 Ultra chipset, Shuttle has made it possible to build a truly high-end Small Form Factor
PC. It is very small and looks really good.
Simple, sleek, and small; marginally bigger than their Zen model.
The obligatory full-color packaging with a carry handle.
Accessories include all the instructions, cords, manuals, and
software you need to be up-and-running quickly.
Manuals, cables, CDs and knick-knacks galore.
SPECIFICATIONS & FEATURES
Shuttle’s overview of the SN95G5‘s specifications.
| Model | SN95G5 | Comment | |
| Size WxDxH (mm) | 310(W) x 200(D) x 185(H) | 3.6kg = 7.92lbs Overall size at 11.5 liters is only slightly larger than the miniscule 9 liter Shuttle Zen. | |
| Chipset | AMD Athlon 64 / nVIDIA nForce 3 Ultra | ||
Memory | Type | 2x Dual Channel DDR | 2GB Max |
| DDR | 333/400 | ||
CPU | Type | Socket 939 | |
| Hyper Transport Technology | 1GHz (2000MT/s) | ||
VGA | 8X AGP Slot | No Integrated Graphics | |
S.Video | None | ||
PCI slot | x 1 | ||
PCI Express x16 | None | ||
| Sound | CHIP | AC97 5.1 | Integrated |
LAN | SPEED | 10/100/1000 | Gigabit LAN |
ATA Connector | 2 x ATA133 | ||
| SATA Connector | 2 x SATA | ||
USB 2.0 | 2(Rear) / 2(Front) / 4(Internal) | ||
IEEE1394 | 1(Rear) / 1(Front) (6/4pins) | Front panel has mini-IEEE1394 port | |
Drive bays | 5.25″ | 1 | |
| 3.5″ | 1 | ||
| HDD | 1 | ||
Power Supply | 240W | claiming 32dB @ 50cm | |
One departure from most of the previous Shuttles is the absence of onboard video.
It is simply not available as part of the nVIDIA nForce 3 Ultra chipset. Anyone who wants to build a high-performance
system will not settle for integrated-quality video anyway. On the plus side, this system’s 240W supply will be able to power
a top-of-the-line AGP video card.
EXTERNAL OVERVIEW
The SN95G5‘s designers went for a simplified,
elegant look. The drive bays and connectivity panel have been stealthed, leaving
only the power and reset buttons exposed. The power and reset buttons were not
very smooth, however. Sometimes the power/reset buttons had to be pressed repeatedly to work. The 3.5″
bay and connectivity panel have been covered by flip-down covers, while the
5.25″ optical drive bay is covered by a spring loaded cover which automatically
opens when the CD/DVD tray is ejected. We found that the 5.25″ bay cover
did not always close completely when the CD tray was retracted, but this may be an issue with the particular optical drive we used.
As with all previous Shuttles, the chassis is constructed
from aluminum. The one-piece top and side cover is a elgantly finished black
anodized aluminum. The sides and top stick to the simplified styling of the
front. The only extraneous features of the sides are a pair of recessed “Shuttle”
logos and couple of intake vents. The intake vents are on both sides and run most of the length of the case.
Center external floppy bay and connectivity panel covers open.
The hidden front panel connectors include 1/8″ headphone/microphone jacks,
two USB 2.0, and a mini IEEE1394 port. All other ports, including audio line-in
and SPDIF, are located on the rear panel, to avoid over-crowding the front.
Also, the space between the USB ports has been enlarged so that multiple USB
devices, such as USB flash drives, can be used at the same time.
The large vents are on both sides.
The rear panel: Decently open grill for the fan exhaust. Note the small PSU exhaust opening.
The rear panel includes a serial port, an IEEE1394 port, a PS/2
keyboard and mouse socket, two USB 2.0 ports, a RJ-45 gigabit LAN socket, 6-channel
audio out jacks, audio line-in jack, SPDIF in/out ports, and a coaxial audio
port. Also the CMOS reset button is accessible via a pinhole in the rear panel,
which saves the user from having to open the case up in the event of an overclocking-gone-wrong
scenario. On the right side of the rear panel are the slot covers for one AGP
and one PCI card.
The underside. The majority of vents are located beneath
the PSU.
Rubber feet are fairly soft.
Fairly soft rubber feet are located on the underside to minimize the transfer of vibration.
Near the front two feet are two threaded holes which will accept the included
taller metal feet. We suspect that the metal feet have two functions: to
enable more airflow under the case, and to look good. Unfortunatly, these metal
feet are not as soft and do not have the same vibration reduction properties.
Front metal feet to provide greater clearance?
INTERIOR
The interior of this Shuttle is much like most Shuttles: Well thought out,
with all the cables routed neatly in order to maintain good airflow and to simplify
installation.
A bird’s-eye view.
The center of the case is quite wide open, allowing for lots of
airflow. The northbridge cooling fan was not as loud as expected. All of the fans in the SN95G5 are 4-wire PWM controlled fans, but the headers can accept
3-pin connectors as well.
A side view of the interior.
The 5.25″, 3.5″, and HD drives are best mounted in the
one-piece drive cage prior to the installation of the drive cage into the chassis. This
makes the installation of these components much easier. This is different from previous Shuttle drive cages, which have often been 2-piece affairs with the main 3.5″ drive bay at the bottom in a removable tray. The Shuttle Zen is a good example of this.
An open, easy-access chassis design.
SOCKET 939 HEATSINK / FAN MODULE
The Shuttle SN95G5 utilizes another variant of the simple-yet-brilliant I.C.E. (Integrated
Cooling Engine) which carries heat away from the CPU via heatpipe. At the opposite
end of the heatpipe is an array of shrouded fins. A 92mm fan which is attached
to the shroud exhausts the CPU heat as well as serving the purpose of a general
exhaust case fan. It is a simple and sensible design which works very well..
The I.C.E. (Integrated Cooling Engine) Module.
The copper base of the I.C.E. module is just about as flat and smooth as one could want. A 92mm fan is used. The wire grill on the inside and the hex-pattern grill on the outside both restrict airflow a bit. Truly noise-conscious users will want to remove these grills.
Smooth and shiny copper base.
The 92mm fan is rated at a whopping 0.56A at 12V. This fan can move a lot of air — and make a lot of noise.
The fan is equipped with a 4-pin plug which is said to allow for improved RPM / PWM control.
The three fan headers on the board are all 4-pin. These are now virtually standard on Intel socket 775 boards.
Note that 3-pin connectors can also be used.
THE 240W PSU
Although physically quite small, the 240W power supply ratings show that it has 16A on the 12V line (192W), enough juice to power any CPU in the AMD64 lineup, along with a high-end
AGP video card and a hard drive. To keep things simple for international distribution, they’ve made it a universal input device that runs on any AC from 100V to 240V.
PSU as seen from the outside.
The air is drawn into the PSU from the side facing the interior
of the case.
A close-up of the PSU label.
The power supply contains a 60mm fan.
When we were listening to the sound of the other fans, we attempted to stop
the PSU fan from spinning by sticking a plastic cable tie into the spinning
fan fan blades. It’s a standard SPCR lab technique and usually works well with no harm done to the fan or PSU when used judiciously. Good idea, right? No! The PSU immediately shut down to protect
itself when the fan blades stopped spinning. This is a pretty good safety feature; if the PSU fan fails, you will not burn up the whole power
supply or risk any other part of the PC. But it’s probably not an easy fan to replace.
SYSTEM SETUP
The following components were installed in the Shuttle SN95G5:
- AMD64 3500+ S939 processor (2.2GHz Winchester core, 1 GHz HTT)
- 2 x 512 mb OCZ DD SDRAM PC3700 Dual Channel Gold
- Matrox G550 AGP video card
- Samsung 40G notebook HDD – nestled on soft foam in 3.5″ HDD bay.
- Microsoft Windows XP Pro SP2 fully updated
The Samsung notebook drive is non-standard. This barebones system is designed to accept a standard 3.5″ HDD, but in our experience, even the quietest 3.5″ desktop HDD makes too much noise, particularly vibration-induced noise, to be recommended for a quiet SFF system. The main issue is placement so close to the user. SFF are meant to be placed on top of the desk, within reach of the user for easy access to the optical drive and front panel ports. In such proximity, it is almost impossible to obtain a level of noise SPCR considers quiet (<30 dBA/1m) without resorting to the very quietest HDD in an elastic suspension.
Samsung notebook HDD nestled in 3.5″ bay; top picture from side/back, bottom picture from front.
If our standard recommended quiet Samsung P series 7200 rpm 3.5″ HDD was mounted normally in this system, the overall noise level would move up to around 30 dBA/1m or slightly higher. There would also be a level of hum from HDD vibrations that would be too high for us to ignore. This is true of ALL small form factor systems we know of.
This begs the question, “Can a 3.5″ HDD be soft-suspended in the Shuttle SN95G5?” The answer: Probably not without modifying the drive cage or eliminating the optical drive to use that space for a suspension. There is simply not enough clearance beneath the HDD cage, which is the place we’ve been able to suspend 3.5″ HDDs in some other SFF PCs.
Very little clearance to suspend a HDD.
BIOS
A BIOS update was available for the SN95G5 at the Shuttle website. It was dated mid-Dec 2004. It is our general policy to use the most recent BIOS available at the time of testing, so WinBios was downloaded and used to flash the board to the new BIOS.
There are a variety of user-configurable options, including Power Management, which seemed the obvious place to look for a switch for AMD’s Cool’n’Quiet technology.
The power management options are related to the conventional suspend mode; there are no references to Cool’n’Quiet in the BIOS. It turns out that CnQ is automatically supported by default.
Surely, by now, everyone is aware of Cool’n’Quiet, but for the record, here is a quick summary:
Cool’n’Quiet dynamically varies the CPU clock speed and voltage so that it is not under heavy load, the power draw and heat disspation is minimal, but makes the full power of the CPU available instantly whenever the load goes up. The changes happen so fast, it is impossible to notice when the processor speeds up or slows down. The exact frequencies and voltages of the reduced speed and voltage vary a bit depending on the type of A64 you have. Earlier versions of the A64 went down to 800 MHz @ 1.3V when CnQ kicked in. Newer versions idle at 1000MHz @ 1.1V. The aim is to reduce power consumption and heat when the CPU does not have a demanding load. It is the same technology that has been used for a number of years in both AMD and Intel mobile processors to stretch battery life in laptops, under the names Speed Step and Power Now! Now with CPU thermal management so high on every PC maker’s priorities, it makes perfect sense to bring the technology to desktop systems. It was implemented in the very first Athlon 64s AMD brought to market in 2003.
To enable CnQ, one has to have a CnQ supported motherboard, then download and install the CPU driver from AMD’s web site and set the Power Option Properties in Windows to Minimal Power Management. There are several versions of the CPU driver, for various operating systems.
It is probably confusing for some users that the term Cool’n’Quiet is used to describe only the Windows ME and Windows 2000 version. The Windows XP version is called AMD Athlon 64 Processor Driver for Windows XP.
The CPU’s core voltage had a wide range of settings from as low
as 0.800V to as high as 1.700V in 0.025V steps. This allows for some serious
manual undervolting, as an alternative to Cool’n’Quiet..
BIOS: CPU Vcore Adjustment.
CPU Fan Control
Another interesting feature of the BIOS is the CPU Fan Control, which is found under PC Health Status. It has several settings:
- Ultra
Low (1000 rpm) - Low (1600 rpm)
- Mid (2400 rpm)
- Full (3600 rpm)
- Smart Fan (~830 rpm default)
Interestingly, the control is selectable between three fan headers: Fan 1, Fan 2, and Fan 3. One last setting selects both Fan 1 and Fan 3. Fan 3 is the header to which the small northbridge HS fan is connected, so Fan 1 + 3 was selected to keep the small fan running more slowly. However, the northbridge HS fan never changed speed, always running 3100~3200 rpm regardless of the fan control setting.
Smart Fan is the most interesting setting.
The user chooses a setpoint called the “CPU Temp Tag” from a range
of 30 to 60°C. This setpoint is where the CPU fan will begin to ramp up
from its minimum speed of ~830 rpm. A chart on the side of the screen
indicates the fan’s speed setting at specific temperture increases beyond
the setpoint.
Note that once the the CPU reaches 80°C, the CPU fan will
be spinning at a loud 3600 rpm regardless of which fan speed control setting
has been chosen. As an aside, when the
fan is running at the top speed of >3600 rpm, it moves a whopping 50 cubic feet per minute. Given that the total volume of the case is less than half a cubic foot, that’s
enough airflow to exhaust the entire case in half a second! On power up, the CPU fan actually starts at maximum speed for a couple of seconds before slowing down to the BIOS selected speed. How loud was this couple of seconds? Loud: 50 dBA at one meter.
Smart Fan Settings: full throttle at 80°C.
Four settings to choose from.
You get to choose when the CPU fan ramps up!
TESTING
No performance benchmarks were run on this system. There are umpteen web sites that document the performance of nVidia nForce3 chipset 939 boards, and the variance is no more than ~3% between the fastest and slowest boards in any group. We don’t think this is at all significant in any practical application. Rather than bog down the review with another bank of artificial behnchmarks, we will just state that this system was as fast any any we’ve seen in the lab save for a few large file operations related to the slightly slower speed of the 5400 rpm notebook drive compared to a 7200 rpm desktop. We have no reason to believe this system falls outside that ~3% performance variance mentioned above.
Power, Temps and Acoustics
The total AC power consumption within the Windows XP environment was measured using a Seasonic Power
Angel AC power meter. The ambient temperature was 21°C during testing.
With the CPU Fan Speed Control set to “Ultra Low“, the CPU temperature
never climbed above ~52°C. This is a remarkably low load temperature, considering the sheer computing power of the A64-3500+ processor. (Admittedly, the CPU temp reading system was not calibrated.) It is a testament to both the cool running 90nm Winchester core of this A64 model as well as the efficiency of the Shuttle ICE cooling system.
There is plenty of cooling headroom for either a hotter CPU or higher ambient temperatures.
With the CPU Fan Speed Control set to “Smart Fan” and the “CPU Temp Tag” set to 60°C, the results were virtually identical. This was because the CPU never reached 60°C, the trigger point we set for the fan to start speeding up. Although the idle fan speed for Smart Fan was slightly lower than Ultra Low while in the BIOS, measurements and careful listening in Windows showed no differences. Our Windows RPM monitoring was spotty; one assumes the fan sped up a bit.
| CPU Fan Settings & Noise
|
| Activity | AC Power Max | CPU Temp | SPL* |
Cool n Quiet | 51W | 36°C |
27 dBA@1m
Idle
64W
38°C
27 dBA@1m
Full Load: CPUBurn
105W
52°C
30 dBA@1m
* Measured 1m from the front bezel with the fan control set to Smart Fan or Ultra Low.
With
the fan speed control set to “Ultra Low” or “Smart Fan“, the noise level was
fairly low, especially considering the amount of processing power in the system.
At all other fan control settings, the CPU cooling fan noise exceeded our requirements.
At idle, the noise was a combination of the 92mm CPU cooling fan, the 60mm PSU fan and the 40mm northbridge fan. No one particular noise source dominated. The notebook HDD was too quiet to be a significant factor.
The 40mm northbridge fan had the highest pitch at idle and may also have the highest turbulence noise of the three fans by a small margin. When the case was closed however, it was less clearly identifiable. The 92mm CPU cooling fan had some bearing clicking but it was subdued. At 1000 rpm or lower, its noise was fairly benign. As for the PSU fan, it became significant as a noise source only when the system was loaded hard for a while.
After >20 minutes of CPUBurn, the PSU fan could be more clearly identified, having ramped up in speed as the PSU became warmer. Neither of the other fans sped up at all in our test environment at either Ultra Low or Smart Fan settings. Keep in mind that a warmer environment will likely have an impact on CPU temperature and/or noise, particularly the PSU fan, which controlled by a preset internal thermistor.
The large vents on either side are a concern with a
louder HDD or a fan-cooled video card. Assuming placement on the top of a desk, the vents allow the noise a direct path to the seated user. As with almost all SFF systems, there
is no room for any aftermarket video card cooling device that requires an extra slot space.
Audio Recordings of the Shuttle XPC SN95G5 as tested:
- MP3 sound
recording of SN95G5 system in idle: 27 dBA/1m – As with most cases, there is some vibration conducted to the case from the HDD and the fans, and this effect can be heard by simply pressing on the sides of the case with one’s hands. This recording is shows the effect. The first 7 seconds are in idle, the next 7 seconds are with the case pressed tightly between my hands. The last portion reverts to the case sitting free. You should be able to hear it easily. With the SLM meter, no differences in sound pressure level can be seen when set to dBA from a meter away. In any case, one way to reduce the noise is to place a soft cover book on top of it, and then a heavy weight of some kind. A large encyclopedia volume from the old days might work just perfectly. - MP3 sound
recording of SN95G5 system at maximum load: 30 dBA/1m
Comparatives: Note that this AOpen was tested with a hotter Intel P4-2.8 Prescott and a louder 3.5″ Samsung HDD.
- MP3 sound recording of AOpen EX915 system at maximum load: 35 dBA/1m
- MP3 sound recording of AOpen EX915 system in idle with Samsung hard drive suspended: 28.5 dBA/1m
| SPCR MP3s: HOW TO LISTEN & COMPARE The recordings above were made with a high resolution studio quality digital recording system. The microphone is 3″ from the edge of the fan frame at a 45° angle, facing the intake side of the fan to avoid direct wind noise. The ambient noise during all recordings is 20 dBA or lower. A quick and simple way to use these recordings for valid listening comparisons is to play the quietest recording on only one speaker (or a pair of headphones) and set the volume so it is just barely audible a meter away. You must turn off any special sound effects, and set equalizer / tone controls to neutral or flat. Don’t touch the volume setting afterwards, and use the same one speaker when you listen to any of the other files. The end result will be reasonably close to the actual recorded sound levels. Here is a recording of a very quiet sound that is barely audible from 1 meter away even in a super quiet room. For full details on how to calibrate the playback level of your sound system to get the most valid listening comparison, please see the yellow text box entitled Listen to the Fans on page 3 of the article SPCR’s Test / Sound Lab: A Short Tour. |
Making it Quieter
We did not have the time to try it, but it seems clear that the CPU cooling system has enough headroom, at least with our A64-3500+ Winchester core, to accomodate a swap for a quieter, slower fan like the Nexus 92. Even if the fan controller ends up pushing it to a higher speed than with the stock fan, the smoothness of the Nexus would lower the overall noise. Couple this with cutting out the back grill and you’d have significantly lowered noise. Perhaps as much as 2-3 dBA/1m.
The other simple acoustic improvement would be to slow down the small northbridge HS fan. A Zalman Fanmate or similar would work fine here. Dropping the voltage down from 12V to ~8V would probably be enough to make a difference. With the 92mm fan swap, we’re probably looking about ~4 dBA/1m improvement overall, and the PSU fan would then be the limiting factor. With that one, I’d say leave it alone.
CONCLUSIONS
The XPC SN95G5 is a nice AMD A64-939 addition to the Shuttle SFF lineup. It packs a lot of performance in a very small, elegant package, and runs fairly quietly, at least with our choice of processor. The CPU cooling system appears to have plenty of headroom for hotter processors, although probably not at as low a noise level. The fan controls in the BIOS are very good, but for silent PC enthusiasts, only two settings are worth using.
The lack of onboard video may or may not be an issue. We can hardly fault Shuttle as this is simply a characterstic of the nForce3 chipset. It would be preferred
for the northbridge to be passively cooled; still, the northbridge fan was
not really that noisy. The BIOS is a full featured enthusiast-type, with great CPU Vcore and clock adjustments, as well as the flexible Fan Control section.
The high energy efficiency of the Winchester core A64 and the Cool’n’Quiet software showed their merits in the low 51W AC power draw at idle. This is probably not much more than 30~35W DC. Power consumption of our system maxed out at a paltry 105 Watts AC, so the 240W DC output rated PSU seems perfectly adequate. Assuming even a high 75% AC/DC conversion efficiency in the PSU, this is less than 80W DC. It seems
reasonable that even with a high end video card and a 3.5″ hard drive,
total DC power draw would probably not go much past 200W. It is for this reason that an external
PSU, like that of the Shuttle ST62K
Zen, might be a viable option. But Shuttle has to be prepared for all varieties of the 939-pin A64s; the ones prior to the Winchester core were definitely hotter and more power hungry.
The small size and all-integrated drive bay makes it very difficult to suspend a standard desktop HDD in this PC. For performance as quiet as we measured and heard, a quiet <20 dBA/1m notebook drive is pretty much mandatory.
The acoustic output on our test sample was quite modest considering the amount of computing power packed into
this small package, but sound levels comparable to SPCR’s
modified Shuttle Zen are not possible with the SN95G2. The limitation is primary in the PSU, whose fan is probably not easy to swap out. For those who seek inaudibility in a low ambient space, there are better options. All in all, the SN95G5 is a full-featured, small, powerful and attractive SFF platform that should be quiet enough for most people if set up correctly with the right components.
Much thanks to Shuttle
for providing the XPC
SN95G5 sample.
* * *
