Shuttle XPC ST61G4 SFF barebones PC

The Shuttle XPC Zen made a big splash at SPCR earlier this year as the quietest SFF system we’d reviewed, with an external fanless PSU and no AGP slot. The Shuttle XPC ST61G4 uses a quiet fan-cooled higher power PSU and offers the AGP slot on the same basic ATI9100 chipset board, making it more suitable for power users. How about quiet-loving power users? Check our ST61G4 review to find out.

July 21, 2004 by Mike Chin

With the longest history behind them and the greatest number of product cycles, you’d expect the originator of the Small Form Factor to be making the most advanced SFF systems around. Our look at the Shuttle Zen earlier this year identified it as the quietest (and one of the smallest) SFF system we’d ever heard, with a host of refinements in technology, execution and ease of assembly for the user.

The XPC Zen achieved its impressive <30 dBA/1m noise performance with two clever but defining aspects:

• An external fanless power supply, much like a very big notebook PC PSU, and
• An absence of the AGP slot, with integrated video the only option.

The two are likely related. Leaving an AGP port in the system would open it up to VGA cards that can draw upwards of 75W, which in combination with a powerful P4, would have the capacity to overtax the passive PSU of the Zen. It was a compromise made for noise, and with the good performance of the integrated graphics of the ATI Radeon 9100 chipset in the Zen, it works quite well.

But PC enthusiasts always want it all, and quiet PC enthusiasts wanted the low noise of the Zen along with the power of an AGP video card. Shuttle has something to offers these people as well. The XPC ST61G4 under review is this offering. It combines the same ATI chipset of the Zen in a motherboard that has an AGP port and dual SATA drive support in a slightly larger case along with an internal higher power PSU with a claimed 32 dBA/1m noise level.

In short, the ST61G4 could be a Zen with the option of a powerful AGP video card. Is it? This is the primary question posed by this review.

Photo of ST61G4 courtesy of Shuttle. Why the standard PR photo? Read the text.

The ST61G4 features the matte brushed aluminum cover common to all the
Shuttle SFFs. It is the size that seems standard in the current Shuttle XPC line — a little bigger all around than the Zen, but still quite small. The most striking thing about this model is the front facia. The reason the standard PR photo from Shuttle is shown above is because that center portion is a mirror surface. It looks and acts like a glass mirror. Have you tried photographing a mirror without getting all kinds of unwanted reflections in it? You’ll see the other photos in this review.

First, the selling features (straight from Shuttle’s web site):

Second, the specifications:

VISUAL INSPECTION

As usual, a colorful package to catch the retail shopper’s attention.

Inside is a full complement of goodies, cables and accessories along with the main product. There are excellent instructions for assembly and operation, and even a soft cloth to keep the case clean.

Notice the mirrored portion of the front bezel? It’s plastic, amazingly enough.

You will have noticed earlier that there is no slot for a floppy drive. There is no place for one in the case, as the 6-in-1 card reader takes up the only available internal spot. Note the row of side vent holes, which is repeated on the other side. These act as the intake vents for the system. There are a couple of small intake holes and a tiny row of pinholes on the bottom at the front, but these can’t really be considered as air intake vents; they’re just too small to be significant.

The front panel features the following, from top, left to right:

• Optical drive bay
• Dual memory card slots
• Power switch, followed by power LED, reset switch and HDD activity LED
• Line-in, mic, headphone, 2 USB2.0 and Firewire connections

The back panel is familiar to anyone who has looked at SFF systems from Shuttle before. The center exhaust grill for the 80mm Integrated Cooling Engine (ICE — or CPU heatpipe / heatsink) is rather restrictive, probably blocking more than half of the fan’s airflow. This is much worse than the exhaust grill on the Zen. Two other items of note:

• SPDIF connectors on top right
• 40mm exhaust fan of PSU on left

Too bad they didn’t use the Zen grill.

INTERNAL DESIGN

The roominess and lack of cable clutter inside is pretty amazing given the small size of the box. Nice cable management, logical design and the ICE cooling system are the keys. Pictures tell the story nicely here.

Drive cage is a single structure held to the top frame with two screws.

With drive cage removed: Yes, that is a 40mm fan you see on the chipset HS.

Slim and thin 250W PSU.

Even though the PSU is internal, its size and shape takes up very little room, as you can see in the photo above. This is the SilentX PC40 (PC40N250EV) Power Supply, rated to deliver 250W, which is quite a lot for a SFF. It measures just 82mm(w) x 43mm(H) x 190mm(L). It has a 40mm fan on either end, one blowing in, one blowing out, in a push-pull configuration.

Top slot for optical drive, middle occupied by card reader, and the bottom for the HDD.

A plastic clip on the bottom of the drive cage is meant to hold IDE cables flush.

INTEGRATED COOLING ENGINE (ICE)

The ST61G4 uses the familiar Shuttle ICE (Integrated Cooling Engine)
heatpipe heatsink to cool the CPU and the rest of the case.
The ICE cooling system consists of an aluminum heatsink with a swaged-in copper
base, four heat pipes and an aluminum finned radiator with an 80mm fan in front
of it. This fan acts as a case cooling exhaust fan and
it also evacuates the heat from the CPU. The fan is speed controlled by the BIOS “Smart Fan” technology
for quieter operation.

The 80mm KD1208PTSI Sunon for CPU cooling is rated at 1.9W, 40 CFM and 33 dBA at 12V.
The HS fins are visible against the interior of the back panel.

Four heatpipes connect copper-base heatsink and fins; similar if not identical to the one in the Zen.
It appears to be a powerful cooling system.

The ICE cooler is secured to the standard Intel heatsink retention
bracket by a nice retention clip that is easy to install and remove but still
holds the heatsink on firmly. Shuttle uses rubber washers to dampen vibrations
between the ICE cooler’s fan shroud and its mounting points on the rear of
the ST61G4’s chassis.

Open design of fins should allow good airflow… if not for the back case grill.

BOARD LAYOUT

The ST61G4 uses a Shuttle FT61 motherboard with the ATi 9100 IGP
chipset. The IGP 9100 competes directly with Intel’s
865G chipset but offers much better onboard video performance. The IGP 9100 consists of the RS300 northbridge and the IXP150
southbridge. The NB chip is actively cooled on this board, with a 40mm fan atop the small HS. This is not a welcome sight for those seeking low noise.

As with the Zen, most of the plug-ins are near the front left, where they are easy to access with the dive cage off. It’s also the location of the CMOS memory clear jumper. But the SATA drive connectors and the floppy drive connector are both in the back.

The optical and hard drives are installed and plugged in, along with the CPU, HSF and memory chips.
A single IDE cable was used to connect both the HDD and optical drives.

As the specifications indicate, this system is compatible with Intel Socket 478 processors on
a 400, 533, or 800MHz front-side bus, giving a wide range of CPU choices
from high-end P4s to low-end Celerons. This chipset also supports
Hyper-Threaded processors and dual-channel memory for up to 2GB of PC3200 SDRAM.
The IXP150 southbridge supports six USB 2.0 ports, four of which are external
plus the two internal headers for the integrated card reader.

There is also support for two ATA-100 P-ATA
IDE channels and two Serial ATA channels; the latter is not included in the Zen. SATA is provided by the Silicon Image 3512 chip, with Raid 0/1 function and support up to 1.2Gb/s (150MB/S).

The Radeon 9100 Integrated Graphics Processor supports DirectX
8.1 and utilizes a 128-bit DDR memory bus running at 300MHz. This should be powerful enough for HTPC’s, business
systems and non-hardcore gaming. The 9100 IGP has single VGA and S-Video outputs,
but no built-in DVI output or multi monitor support. That’s what the AGP slot is for.

The AGP slot is located closer to the outer edge of the board than the single PCI slot, due to the need for VGA cards to clear the drives and cages midway up the length of the case. This is a standard location for AGP slots in SFF systems. It means that dual-width cooling VGA systems such as the Zalman heatpipe coolers or the Arctic Cooling VGA Silencer simply cannot be used.

NOISE-RELATED BIOS CONTROLS

Line the Zen, the ST61G4 BIOS has five fan settings, starting with Smart Fan mode
that allows the threshold temperature to be set manually. If set to 60°C, for example, the fan stays at minimum speed and does not ramp up until this CPU temperature is reached. There are four others, as shown in the screen below: Ultra Low, Low, Mid, and Full.

The second relevant BIOS setting for noise control is Vcore, the core voltage of the CPU. Overclockers raise Vcore to keep the CPU stable while it’s running higher than the rated speed. The price is a big increase in CPU heat because power dissipation varies with the square of the voltage, which is an exponential relationship between voltage and heat.

PC silence enthusiasts know this relationship well and use judicious drops in Vcore to make the CPU run much cooler. Almost every CPU I’ve ever used has been perfectly stable with 0.1V less than standard voltage, and the temperature drop is usually 2~3°C. So how undervoltable is the ST61G4? Very underclockable, in 0.125V increments all the way down below 0.9V, which is way below what’s feasible with any 478 socket CPU that will run on this board. This feature alone helps makes the ST61G4 worthy of a silent tweaker’s attention.

SYSTEM SETUP & TESTING

The following components were chosen:

• Intel P4-2.8C (800MHz bus w/ Hyperthreading)
• 2 x 512MB OCZ DDR PC-3500 Platinum memory
• Hitachi 180GXP 120G ( 8MB cache) hard drive
• Plextor CDRW PX-W1610A

No AGP VGA card was used initially, because I did not have the right class of card for this case: Something like a fanless ATI 9600, SE, or XT card. Or maybe a Matrox 650. In any case, you’ll see from the results below that noise limits are reached by the ST61G4 without the extra heat load of an outboard video card. (There! The cat’s out of the bag!)

The P4-2.8 is really a bit hot for a SFF system, but it is about the slowest P4 Intel officially makes these days, so it’s appropriate. The processor was perfectly stable at 1.4V (instead of the usual 1.5V) in this setup, so this setting was used throughout testing. Lower heat is a “free” benefit when you have a board with undervoltable Vcore. Both the memory and hard drive were chosen for speed and, in the case of the Hitachi drive, relatively low noise. It was such a simple assembly that there’s nothing to say about. It took a little over half an hour, a testament to Shuttle’s excellent human engineering. Installing Windows XP on the machine took much longer. (NOTE: Installation with a SATA drive is more complex because the controller is not native to the chipset and apparently demands the use of a floppy drive.)

Note that there are five discrete sources of noise in the ST61G4:

• 1 – 80mm ICE cooling fan
• 2 – 40mm fans in PSU, separated by the 190mm (7.5″) length of the PSU, with one fan facing the outside and the other embedded on the inside end.
• 1 – 40mm fan on the Northbridge chip heatsink, running at ~4600 RPM
• 1 – Hard Drive (the Hitachi 180GXP in this case)

The black CD drive looks pretty good against the silver.
In the mirrored panel, you can see reflections of the cloth the Shuttle is sitting on.

There are no performance benchmarks presented for this machine. The rationale is simple: They will be identical to the results obtained with the Shuttle Zen, aside from effects of the different components used. Please check in the Zen review for details.

But what about performance with an AGP video card? Well, it will depend entirely on the video card, won’t it? And whatever video card is used, you can be virtually guaranteed to see an increase in CPU temperatures and PSU fan noise reported below.

What’s far more interesting for us are the noise and cooling performance. Although not available publicly, we managed to obtain a utility that enables the CPU temperature diode output to be monitored in Windows on the ST61G4. (A nondisclosure agreement binds us to secrecy.) This utility was used to obtain the CPU temperature readings below.

Other test tools used:

• CPUBurn processor stress software utility, two instances run simultaneously to obtain 100% processor load.
• Seasonic Power Angel AC Power Meter (a new product; will be covered soon.)
• B&K model 1613 sound level meter. This 20+ years professional caliber SLM has a dynamic range that spans over 140 dB. and can measure accurately to below 10 dBA.

The measurements for the full load CPUBurn results were made after 30 minutes in that state.

AT POWER ON – With the BIOS fan setting are set for Ultra Low (my preferred setting) or Smart Fan, the noise at turn on was modest, but not really quiet by SPCR standards. 34 dBA/1m is 4 dBA above the level we generally consider quiet. Most of the noise emanated from the 80mm fan, which has a rather annoying acoustic signature combining both low noise humming as well as some high pitched whining. This was at a speed of 2200 RPM. Despite my initial qualms about the 40mm fans in the Shuttle SilentX PC40 Power Supply, the PSU was not a significant source of noise. Neither was the >4500 RPM 40mm fan on the NB chip heatsink.

IDLE / LOW LOAD – Because the fan was set at Ultra Low, the 80mm fan noise did not change at all. However, over time, the external PSU fan became more identifiable as a distinct source of noise even though the overall level did not really increase. As expected, it is a higher pitched sound, but under light load, not particularly intrusive. The 80mm fan is more annoying.

AT FULL LOAD – The noise of the PSU fans became more intrusive, but again the overall level was not that changed. The 80mm fan is still more annoying.

Because of the good headroom in the CPU temperatures, it seemed safe to make a small change that anyone could make to reduce the noise: A Zalman Fanmate1 fan speed controller was inserted between the 80mm fan and the output header on the motherboard. The Fanmate1 was simply reduced to minimum speed, at which point, the fan dropped to 1600 RPM. The tests were run again.

AT POWER ON – As the table above indicates, the overall noise dropped by 3 dBA, but subjectively, it sounded like a bigger drop than that. It was borderline quiet, but much more acceptable than before. The 80mm fan was still the biggest noise source. Again, the noise from the 40mm fans was there but not really identifiable above the noise of the 80mm fan.

IDLE / LOW LOAD – Because the fan was set at Ultra Low, the 80mm fan noise did not change at all. However, over time, the external PSU fan became more identifiable as a distinct source of noise even though the overall level did not really increase much.

AT FULL LOAD – The noise of the PSU fans became the most prominent part of the noise but the 80mm fan still could be heard. It was not that quiet.

TIME TO SWAP THE FAN

By now, it’s obvious that the 80mm Sunon fan is a noisy, nasty fan. Why Shuttle could not have stayed with something like the much quieter Bi-Sonic fan in the Zen is a mystery. It was time to see if anything else could be done.

The ICE fan shroud was removed and examined. The Sunon fan is a standard 1″ thick 80mm fan, which means any one of several quiet fans on had could be easily swapped in its place. A Nexus 80mm “Real Silent Fan” was installed in place, and the wire grill over the original fan remained off: Anything to improve the airflow. The Nexus fan is only rated for 20 CFM and 0.15A at 12V, and just 17.6 dBA/1m (which may be optimistic but…). It is one of the smoothest and quietest fans I’ve examined, though the price is pretty low airflow, especially compared to the 40 CFM rating of the Sunon. It does have RPM monitoring, which is useful here.

Nexus “Real Silent Fan”
in place of the nasty Sunon.

The wire grill remained off, unlike the one over the BI-Sonic fan from the Zen.

A quick test with the Nexus fan in place showed that the 4500 RPM 40mm fan of the NB heatsink would be the next bottleneck. So the Zalman fanmate1 was pressed into service again, and the speed dial rotated until the noise of the NB heatsink fan dropped to the ambient level of the case set by the PSU fans and Hitachi hard drive idle noise. The end result was much better.

AT POWER ON – The improvement in noise was dramatic. This was close to the level of the Zen, perhaps a little higher, like the AOpen XC Cube. The overall noise dropped by at least 3 dBA and sound like more because of the much smoother quality of the noise. In fact, the 80mm fan no longer was the predominant noise source, even though a small degree of turbulence could be heard through that restrictive fan grill. No single noise source was identifiable as the defining one. It could definitely be considered quiet. Note that the CPU temperature did not change despite the lower airflow fan.

IDLE / LOW LOAD – The external PSU fan became more identifiable as a distinct source of noise even though the overall level did not really increase much. The distinctive humminess of the aluminum case identified in previous aluminum SFF systems became evident. Previously this effect had been masked by fan noise.

AT FULL LOAD – The noise of the 40mm PSU fans became more prominent again. It is a higher pitched noise that’s more intrusive than a good 80mm fan would be. The overall noise is probably still acceptable for many users, however.

Note that the CPU temperature actually stayed the same as with the double airflow capacity stock fan. This suggests that the effective airflow at the heatsink fins is set not by the speed of the fan but by the grill, which restricts airflow beyond a certain fairly low level. It would take a lot higher RPM (and noise) to force more air through that grill.

TEST #4: GF4800 AGP + NEXUS FAN

The final question that will be asked is what will happen to PSU noise when an AGP video card is used and the total load increased significantly higher than the 110W AC seen at max load with just the integrated video. So an nVidia GeForce 4800 card was pressed into service, the same one used in the AOpen XC Cube EZ65 review. Its noisy fans were disconnected when checking the PSU noise levels.

The GF4800 card takes up almost all the available room; any longer and it would not fit.
With the cover on, the VGA card fans get jammed up against the side panel and the added impedance makes them whine louder with greater turbulence noise.

TEST #4 ANALYSIS: The results of this test are promising. With a 40W (35%) increase in AC power consumption, the PSU became only marginally louder, and the CPU temperature remained safe if a little high. The thermal and acoustic results here are fairly indicative of performance with some variant of a fanless ATI 9600 or similar fanless midrange AGP card. It’s what I’d call marginally quiet but well below the level of typical PCs or SFF systems that are similarly equipped.

NOTE: Throughout the testing, the Hitachi 180GXP hard drive’s idle noise was below the noise threshold of the overall system. Seeks could be heard, but were well damped and not particularly annoying in the context of the overall noise. The head reset noise Hitachi drives are known for was not heard during the testing.

• A quieter main 80mm fan
• Removal of the grill over the 80mm fan
• A quieter NB chip cooling fan (this fan can probably be dispensed with altogether if a larger HS was used)