You are here

AOpen i915GMm-HFS: 2nd Gen Pentium M desktop board

AOpen i915GMm-HFS Pentium M Motherboard

June 19, 2005 by Ralf
Hutter

Product
AOpen
i915GMm-HFS
motherboard
Manufacturer
Street Price
US$280~350

The first consumer-oriented Pentium M motherboards were released late last
year by AOpen and DFI. (Editor's Note: And both boards were reviewed thoroughly by RH.) Both were based on the Intel 855G/ICH4 chipset, which
is stable and mature but doesn't come with the latest features that enthusiasts
expect on a brand new board. The rather high price point of these boards
only served to exacerbate this issue.

AOpen was not content to sit
back and see how well its initial Pentium M offering did in the marketplace. They
quickly followed up with a new board, the i915GMm-HFS,
based on the new 915G/ICH6 chipset.
Included on this board are features like PCIe x16, DDR2, SATA300, HD Audio and
a 533MHz FSB among others. The integrated graphics includes not only the typical
analog VGA out, but DVI and HDTV outputs as well. I doubt that anyone will be
complaining about the lack of features on this iteration of a Pentium
M board!

Luckily for SPCR, AOpen was kind enough to send us one of these brand
new boards to play with and we've been testing it out for the past few weeks.
Today's review will give a bit of an overview of our impressions.





AOpen i915GMm-HSF. Unusual layout for an Intel board seems to emulate BTX.

First off, here's a chart comparing the features of the AOpen i855 and i915
boards:


Comparison of AOpen i855GM and i915GM Motherboards

Feature
i855GMEm-FLS
915GMm-HFS
Processor
Socket-479m Pentium-M
Socket-479m Pentium-M
Chipset
Intel 855GME Northbridge

Intel ICH4 Southbridge
Intel 915GM Northbridge

Intel ICH6M Southbridge
FSB
400 MHz
400 MHz, 533 MHz
Memory Interface
2 x DDR333 Slots (2GB Max)
Single Channel DDR-333 or

Dual Channel DDR2-533
Memory Bandwidth
2.7 GB/sec Peak Bandwidth

(Single Channel DDR333)
2.7 GB/sec (DDR)

8.5 GB/sec (DDR2)
Graphics Interface
1 x AGP (4x)
PCI Express x16
IDE Ports

2 x Ultra ATA/100 Ports

(4 Drives)

Floppy Port

1 x Ultra ATA/100 Port

(2 Drives)

2 x SATA Ports

2 x SATAII Ports
S-ATA RAID
2 x SATA-II/150 Ports

Promise 20579 Controller

RAID-0, RAID-1, JBOD
2 x SATA-II/150 Ports

Silicon Image SATA Controller

RAID-0, RAID-1, JBOD
PCI Slots
3 x 32-bit (33 MHz) PCI
2 x PCIe (1 x1, 1 x16)

2 x 32-bit (33 MHz) PCI
USB / Firewire
Ports
6 x USB 2.0 (Intel)

2 x Firewire 400 (Agere)
8 x USB 2.0 (Intel)

2 x Firewire 400 (Agere)
Integrated
Graphics
Onboard Intel

Extreme Graphics 2

HD-15 Analog Output
Onboard Intel

Extreme Graphics 2

HD-15 Analog Output
Onboard Ethernet
2 x Marvell
Yukon GigE Ports

(32-bit PCI)
2 x Marvell
Yukon GigE Ports

(32-bit PCI)
Onboard Audio
Realtek ALC655 AC'97 Audio

5+1 Channel Analog
Intel HD Audio

7+1 Channel

And here's the complete rundown on the board's features, as given by AOpen:

AOpen
i915GMm-HFS Specifications
Processor
Intel Pentium
M CPU (Dothan)

Socket 479, 533Mhz
Chipset
Intel 915GM, ICH6-M
Super I/O
Winbond
Clock Gen
Realtek
Memory
Dual Channel Mode (DDRII
400/533MHz only)

Support : DDR333 or DDRII 533

DDR or DDRII DIMM x 2

DIMM Type : DDR or DDRII 128/256/512MB & 1GB

Max Memory : 2GB
Graphics
Integrated Intel GMA900
graphics core

PCI Express x16 (PCIe x16) slot
Audio
Azalia Codec
on-Board embedded S/PDIF (ALC880)

7.1 Channel and above
Network
Dual Marvell Gigabit
PCI Express LAN
IDE
Integrated ATA100 and Serial ATA Controller


Silicon Image Serial ATA II Control Chipset [Support RAID 0,1]
USB
Integrated in chipset, 8 x USB2.0 ports
IEEE 1394
Agere 1394 Control Chip
Slots
PCI Express (PCIe x1) x
1

PCI Express x16 Graphics (PCIe x16) x 1

PCI x 2
Storage & Back Panel I/O
Floppy Drive Connector x 1

IDE Channel : ATA100 x 1

Serial ATA Channel x 2

Serial ATA II Channel x 2

USB Port x 4

LAN Port x 2

VGA Port x 1

D-Connector (D1, D2, D3, D4) x 1 (for Japan only)

DVI-i Connector x 1

Standard Video x 1

YPbPr Connector x 1

Line_In/SPDIF_In x 1

Speaker_Out x 1

MIC_In x 1

Center/Subwoofer x 1

Rear Surround/SPDIF_Out x 1

Side Surround x 1
On Board Connector
Front Panel x 1

CPU FAN x 1

System FAN x 1

Chassis FAN x 1

Power Temperature Connector x 1

CD_IN x 1

COM Port x 2

IEEE 1394 x 2

USB Port x 4

Printer Connector x 1
BIOS
Award PnP 4Mb Flash ROM BIOS
Form Factor / Size
micro-ATX / 244 mm x 244
mm
Software & Utility
Acrobat Reader

AOconfig utility

EzInstall utility

EzSkin utility

EzWinFlash utility

i915GMm-HFS Utility

Online eBook Manual

WinDMI utility
Accessory
CPU Cooler x 1 (Ha!)

Easy Installation Guide x 1

Enhanced Full Pictured Manual x 1

Bonus Pack CD disc x 1

Floppy Disk cable (special) x 1

80-wire IDE cable x 1

Serial ATA cable x 2

Serial ATA Power cable x 2

Back Panel I/O Shield x 1

PS/2 Keyboard Connector x 1

S/PDIF Converter x 2

The board is a micro-ATX form factor, in keeping with the all the
Pentium M motherboard releases to date. Given the very complete feature set
and the size of the board, the 915Gmn will be a prime
candidate for HTPC and SFF desktop systems.

The board itself features the AOpen standard black PCB, as well
as the fancy aluminum "AOpen" heatsink, although this time it appears
on the southbridge instead of the northbridge. The most obvious thing about
this new board is the positioning of the components on it. The CPU, NB chipset
and DIMM slots are all in non-standard locations, making this board look like
a strange cross between an AMD A64 board, and a BTX offering. Why AOpen laid
it out this way, I don't know but it doesn't seem to interfere with its installation
or operation.




NB heatsink at center of board. It's actually larger than the CPU heatsink!

The board is dominated by the big aluminum northbridge
heatsink positioned near the center of the board itself. The processor socket
is located to the right of the NB and is sandwiched in between the NB heatsink
on one side, the voltage regulators and IDE header on the opposite side, and
a pair of jumpers near it's bottom side. This doesn't leave much room for a
CPU heatsink, a fact that is clearly evident by the tiny, apparently proprietary
set heatsink mounting holes located around the periphery of the CPU socket.
The CPU socket is the same low-profile "twist-lock" notebook type
that was found on AOpen's earlier 855G board.



Close quarters around CPU socket.

DIMM slots and caps above, Vreg on right, jumpers below.

AOpen's earlier 855G board used a standard Pentium 4 heatsink
bracket, and included their own cooler that easily installed onto the P4 bracket,
which did a good job of cooling the CPU fairly quietly, especially if run at
under 12V. AOpen has made big changes to that system for their latest P-M board.
Whether due to space constraints around the socket, or perhaps a marketing-based
decision, this board includes a small 18 fin aluminum heatsink and a 40 x 10mm
fan. This combination is smaller than a lot of northbridge heatsinks that I've
seen, but perhaps AOpen considers it enough to handle the low 20-25W
output of a Pentium M CPU. The heatsink is mounted by bolting it down onto a
metal bracket that fits beneath the motherboard and correct pressure is provided
by a set of four spring-loaded screws.



AOpen heatsink package. Insulated steel bracket goes underneath the mobo.



Yup, this CPU HS actually seems to have less cooling surface area than the
HS on the NB chip.

The DIMM slots are located above the processor and northbridge,
as in the newer AMD A64 boards. I'm not too sure of the engineering behind this
layout, but it certainly doesn't mean that Intel/Aopen have moved the memory
controller on die like the A64. The blue set of DIMM slots supports up to 2GB
of DDR333 memory, the black set supports up to 2GB of the new DDR2-533 memory,
with a theoretical bandwidth almost three times greater than that of the DDR333
slots. The DDR2 memory does operate at higher latencies than typical DDR memory,
and whether the Pentium M can even use all the extra bandwidth provided by the
DDR2 is another issue. It's nice for the end-user to have a choice though. As
DDR2 becomes more mature and penetrates more widely into the marketplace, the
inclusion of the DDR2 slots could be a nice -- even necessary -- feature.



DIMM slots. Black for Dual Channel DDR2, blue for good 'ol DDR.

The upper left corner of the board sports a standard 20-pin ATX
connector, as well as the AUX12V power connector. Below the
ATX connector lies the sole P-ATA header on the board. The 915GM chipset only
supports one standard UATA channel, which limits the user to no more than two
non-SATA devices. This may not be too much of an issue though since the ICH6M
southbridge includes support for two SATA-1 devices, and the integrated Silicon
Image controller provides a pair of SATA-II ports, which can support a RAID 0 or
1 configuration, or be used as JBOD to add another two SATA devices to the
system. The SATA-II ports also support NCQ.



SATA and SATAII ports and color coded front I/O connectors, plus southbridge
heatsink.

Continuing clockwise around the periphery of the board, we find
the I/O header for the Power and Reset switches, and the power and HDD activity
LEDs. These are color coded, just like on the previous AOpen Pentium M board; it's a feature that I really appreciate. Next up are a pair of internal Firewire
headers, run by an Agere 1394 chip, and two pair of internal USB 2.0 headers,
for a total of four internal USB 2.0 channels.



Internal headers. From L to R: FDD, COM 1 & 2, Parallel, 2 x USB 2.0 and
2 x Firewire.

Next up on our little tour will be the section of the board reserved
for the Luddites out there in Computerland. There's an internal Parallel Port
header, a set of COM port headers and a vestigial FDD header. The floppy header
is the standard 34 pin connection, but it's been squeezed into a proprietary
width, due to space constraints on the board itself. AOpen has included a custom
size FDD cable along with all the other cables and goodies, but if this cable
ever got lost or broken, you'd be out of luck. Last, but not least in the Luddite
section is an internal PS/2 header. The rear I/O panel is devoid of both PS/2
keyboard and mouse connectors, but AOpen has kindly provided an internal PS/2
keyboard header, and a PCI bracket to cater to those consumers (like yours truly)
that are still using the centuries-old PS/2 keyboards.



PS/2 keyboard adapter plugs into internal header on board.

AOpen has provided both full and half size PCI covers to mount this.

The rear I/O header shows off many of the extra features of the
915GMm board. In place of the PS/2 connectors is a set of six analog
and S/PDIF in/out digital ports for the board. These ports are 7.1 channel HD
Audio connected through the integrated Realtek ALC880 CODEC. AOpen includes
two optical S/PDIF adapters with the motherboard, so the motherboard will be
able to connect directly to digital speakers. Inhabiting the previous location
of the Serial and Parallel ports are the video output connectors,
one being the standard analog VGA connector, and the other being a DVI-I digital-out
connector. Next in line are an S-Video TV out and a set of YPbPr (HDTV) connectors.
All of these connectors are fed from the integrated Intel GMA900 graphics core.
Intel's GMA900 is DirectX 9.0 compatible, and supports DVI connections up to
1600 x 1200.

Intel touts the GMA900 chip thus:

"Graphics designed with media in mind. Incredible
visual quality with native 16x9 formats for wide screen flat panels, motion
compensation for smooth DVD playback and support for popular HDTV display formats.
Outstanding overall media experience when combined with Intel® High Definition
Audio (Intel® HD Audio)."

Next on the rear I/O panel is a pair of Gigabit Ethernet ports
which are connected by a Marvell Yukon chipset. Last, but not least are yet
another four USB 2.0 connectors. These are external connectors, and when combined
with the four internal connectors give this board an impressive eight USB 2.0
connectors. You'd really have to try to run out of USB connectivity on this
board.



Rear I/O with no Legacy ports at all.

L-R: HD Audio, VGA, DVI, HDTV, S-Video, 2 x GigE LAN, 4 x USB 2.0.

If the integrated Intel GMA900 graphics chipset isn't your cup
of tea, AOpen and the Intel ICH6M chipset have provided a PCI Express x16 graphics
card slot for more powerful video choices, as well as one PCI Express x1 slot
and two good 'ol 32-bit PCI slots.



Top to bottom: PCIe x 16, PCIe x 1, 32bit PCI x 2.

AOpen has included a complete set of accessories along with the 915GMm. There's
the usual manual and driver disc, along with a RAID driver floppy and rear I/O
plate. Also included is a set of 2 SATA cables, an ATA100 ribbon cable and the
proprietary FDD cable. A set of Molex-to-SATA power adapters, the PS/2 keyboard
adapter and a pair of S/PDIF adapters round out the accessory kit for this board.



Accessory kit included with the board.

So there's the kit, now let's assemble it and see how it works. As a bit of
a Luddite myself, I'm a bit concerned about having only one P-ATA channel, and
no PS/2 ports, but for this review I'll jump firmly into the 21st century and
see how I fare.

TEST SETUP

Since the low power Pentium M CPU should let us run a very cool,
and therefore very quiet system, we're going to utilize as much quiet hardware
as possible. If our goal is to build a fast and extremely quiet system, the
ancillary hardware shouldn't drown out the components we're using to cool the
CPU. Our test system will be based on the following hardware, all of which is
ridiculously quiet:

TEST PLATFORM & PROCEDURE

The Pentium M test platform is an open system not enclosed in a case.

Intel Pentium M 755 - 2.0GHz Dothan core - TDP is 21W, MP
is 26.8W.

Mushkin PC3200 Level II - 2 x 512MB DDRAM @ 2-2-2-5, 333MHz

Samsung MP0402H 2.5" 5400rpm, 8MB cache notebook
hard drive, decoupled with Sorbothane blocks

Seasonic SuperSilencer 300W (rev A1) PSU modded with 5V Panaflo M1A


Arctic Silver Ceramique Thermal Compound

Two-level metal platform with rubber damping feet. Motherboard on top; other
components below.

CPUBurn processor stress software

Speedfan v2.24 software to track CPU temperature and fan
speed

Seasonic Power
Angel
power monitor used to measure system power usage

The board was tested using the quietest components I could round
up. I used the onboard graphics, mostly through the DVI output, but I also used
the VGA out for a while too. The very quiet 2.5" Samsung HDD was used for
the vast majority of the testing, but just for grins I also decided to play
with an SATA drive for a while. I used a Samsung SP1614C for that part of the
testing. It was set up with a fresh OS install on it.

As per standard practice, prior to all testing, the stock AOpen
heatsink was installed on the test system as per the manufacturer's instructions.
No thermal tests were run until system stability was assured by running the
Prime95 Torture Test (v23.8)
for 8 hours and Memtest86 (v3.20) for
at least 24 hours. During the thermal testing phase, each test was run for 30
minutes even though all temperatures generally stabilized within 15 to 20 minutes.

Each thermal test was repeated three times on consecutive mornings to check
to the consistency of the results. All results were within 1-2°C of each
other and the average readings are included in the charts.

Ambient temperature was measured at 71°F (21°C) over the entire series
of tests. No tests were run unless the ambient temperature was at that reference
level.

* All temperatures in degrees Celsius.

* Diode: Reading from Pentium M 755 CPU diode via Motherboard Monitor.
Diode was calibrated on each board using the standard SPCR
CPU Diode Calibration method
.

* Temp Rise refers to the difference between ambient temperature and
the diode reading. .

* °C/W refers to the °C of temperature rise per watt of heat
dissipated by the CPU.

SYSTEM BUILD

As motherboards become more full featured, building a system seems to take
less and less time. This latest AOpen board was no exception. Using the onboard
video meant that I could skip the usual video card (as well as its heat), so
the build consisted of the processor, a couple of sticks of memory, the HDD
and a floppy drive. Oh, and the Legacy PS/2 keyboard adapter.

This board has native support (via jumpers) for both the earlier 400MHz processors
and the newer 533MHz versions. For this review I used my existing 400MHz,
2.0GHz M755 CPU. Its default speed is 100MHz with a multiplier of 20. While
experimenting with earlier boards, I found that it will also run fine at 133MHz
with a multiplier of 15 (also 2.0GHz), which meant that I could run it on this
new board at those same settings. I ran it both ways throughout my testing and
experienced no glitches or instability either way.

This board uses the same low-profile CPU socket of as all the previous Pentium
M boards that I've used. The heatsink came with a pre
applied Thermal Interface pad which I removed and replaced with Arctic Silver Ceramique.
Mounting the heatsink was very easy using the supplied bracket (which is positioned
underneath the motherboard) and the set of four spring loaded screws.

The low profile mPGA479 socket was a pleasure to use. It's a ZIF socket, just
like any other modern socket, but instead of locking via a lever on the side,
it locks through the camming action of rotating a little screw 180°. I found
this method even easier than using the typical locking lever.

The build went fine with the system system POSTing with no problems on the
first boot. Upon booting into the BIOS for the first time I noticed, with some
concern, that the CPU temp was being reported as 47°C. This was almost 20°C
hotter than I've been used to seeing on my other Pentium M setups. I touched
the heatsink and found that it was fairly warm (for just idling) so I made a
mental not to keep my eye on this.

I flashed the BIOS to the latest revision
(1.04 at the time) and then installed Windows XP SP2. After Windows was installed
I loaded the latest Intel chipset drivers and then installed the NIC drivers
from the supplied driver disk. I then connected to the internet and installed
all the latest OS hotfixes from Microsoft. At this point I did a minor amount
of GUI configuration and then installed Speedfan
so I could check the odd CPU temperature from within Windows.

Speedfan showed
idle an temp of 45°C. A calibrated finger check of the heatsink also showed
that it was fairly warm, considering it was just idling. I pulled the heatsink
so that I could recheck my mounting procedure, but found nothing wrong. A very
careful reinstallation gave me identical load temps. I then uninstalled Speedfan
and loaded the AOpen monitoring software. It also reported CPU idle temps in
the mid 40°C range. I then ran the Thermal Calibration procedure detailed
in this SilentPCReview
article
. Test results showed that temps were actually being underreported
by about 3°C! At this point during a typical system build I normally run
stability tests to verify that everything's actually working correctly before
proceeding on with the rest of the testing process.

So it was with more than
a bit of trepidation that I fired up Prime95 to test for stability within the
OS itself. CPU temps rapidly shot up into the high 60°C range, and finally
stabilized at around 72°C! I wasn't thrilled to see my $430 CPU running
this hot, but a check with Throttlewatch
v2.0
showed no throttling, so I let it run for 24 hours. After passing Prime95
with (hot, but) flying colors, I ram Memtest86 for another 24 hours. Everything
seemed stable, but darn hot, especially for a CPU that puts out just 20-25W
at full load. Just for grins I lowered the default Vcore from 1.324V to 1.100V.
This served to lower the Prime95 load temps to a chilly 58°C, a range that
I felt much more comfortable with.

CPU temperature not withstanding, everything on the board worked fine.
All the Firewire and USB 2.0 ports worked great, at full speed using just the
included slipstreamed sp2 drivers. DVI output was nice and crisp on my DVI supported
LCD monitor, VGA was softer, as is typical on a DVI LCD monitor. The dual GigE
LAN worked flawlessly with the supplied AOpen drives., and fortunately for me,
the legacy PS/2 keyboard convertor also worked great.

SYSTEM TESTING

So, how well will this board work as the centerpiece of a super quiet desktop
or HTPC setup? The high idle and load temps that were noted during stability
testing weren't too comforting, but now it was time to lean on this board, and
possibly do a bit of tweaking to optimize things.

First up, we'll test the stock HSF at it's default speed of 12V. After various
experiments with the stock heatsink during initial testing, it ended up being
installed in it's out-of-the-box configuration using Arctic Silver's Ceramique,
which was "burned in" for a full 24 hours prior to thermal testing.
I wanted to give the heatsink every chance in the world to perform it's best.

The heatsink comes stock with a 40mm x 10mm transparent blue fan screwed to
the top of the aluminum fins. I was told by sources at AOpen that this fan was
chosen for its quiet sonic signature, and not for performance. Testing
showed that to be easy to believe, as it was pretty quiet for a 40mm
fan, especially at 12V. It's probably the quietest 40mm fan I've ever heard,
with only a very slight motor hum and a quiet, and rather low pitched whine.
It actually sounded more like an 80mm fan than a typical 40mm fan. This low
noise comes at the expense of airflow though. There's certainly not much, and
the lack of airflow combined with the very small size of the heatsink adds up
to a very hot running 20W CPU!

CPUBurn led to full load
temps of 79°C in the stock 12V configuration, a situation that didn't bode
well for further experiments with undervolting the stock fan. Undervolting the
CPU from the default 1.324V down to 1.10V helped somewhat, but I'm not sure
that the average user would want to do what it takes to find a stable, lower
voltage setting to operate at. Just for fun, I briefly tried it at 7V, with
the CPU at default Vcore, but shut it down after it reached the mid 80°C
range within less than 15 seconds of firing up CPUBurn. I fully realize that
modern CPUs can probably run around 70°C without lots of long-term effects,
but I'm not willing to risk my expensive processor in the name of science, so
I kind of put the brakes on things and went to "plan B". Here's the
temperatures with the stock heatsink:

Temperatures
with Stock Heatsink and Fan
CPUBurn

Load Temp.

°C
rise
°C/W (TDP)
°C/W (MP)
Fan @ 12V, default Vcore
79°C
58°C
2.76
2.16
Fan @ 12V, 1.100V
58°C
37°C
1.76
1.38
Fan @ 7V, default Vcore
85+°C
too much
N/A
N/A

°C rise refers to the rise in temperature over the ambient at load.

°C/W - TDP calculations: Intel's TDP of 21W was used.

°C/W - MP calculations: CPUHeat
& CPUMSR Projects'
estimate of 26.8W was used.

In this case, "Plan B" consisted of looking for alternatives to the
either the fan or the heatsink, or both. The quietest 40mm fans I have on hand
are louder than the AOpen fan, so that idea was out. I measured the mounting
hole pattern on the motherboard so that I could look for a replacement heatsink
that would hopefully bolt right on with a minimum of hassle. I was thinking
that perhaps a northbridge heatsink, or perhaps a 60mm type of heatsink would
work. Twas not to be, as the hole pattern measured 52mm x 52mm square, a non-standard
size for any heatsink I could find. I even tried using a Thermalright NB-1C,
the slick little all-copper NB heatsink with all the adjustable mounting methods.
It wasn't even close to being able to fit.

Next on my list: heatsinks for embedded
motherboards. A fair amount of Googling yielded some promising results, among
them some nifty looking all-copper sinks from Coolermaster, of all people. The
same Google search also yeilded the fact that AOpen happened to have an online support
forum, whose motherboard section was full of people having the exact same heat
issues with their 915GMm's and the stock heatsink. One of the posters had just
ordered the Coolermaster embedded heatsinks, so I waited and
let him be the guinea pig on that one. Well, his heatsink arrived, an it didn't
fit the AOpen hole pattern. Apparently AOpen created a hole pattern that was
entirely unique to them. Great for selling their own heatsinks, but a complete pain for
anyone trying to upgrade to a better heatsink.

Never one to throw in the towel prematurely, and having a Nexus modded Zalman
7000AlCu sitting unused on the test bench, I started measuring things to see
if it might be a candidate for the stock heatsink's replacement. The Zalman/Nexus
performed superbly on my earlier Pentium M board, even when run completely passive
so I figured it might be the ticket for this board as well. When I measured
the mounting hole pattern on the Zalman, it came real close to matching the
hole pattern on the AOpen 915 board, if it was mounted diagonally across the
socket. When I mocked up the Zalman, I found that it would need some minor grinding
of it's fins in about three places to clear the various obstacles on the board.
I would also need to to make a set of risers approximately 0.560" high to
mate the Zalman to the correct height need to work with the Socket 479 and the
heatspreaderless Pentium M processor. All this was starting to seem like a bit
too much work to fall under the guise of a "minor mod" so I decided
not to pursue the time, expense and Zalman grindage that would be required,
and went to "plan C".

Plan C consisted of carefully positioning the Zalman/Nexus heatsink on top
of the CPU die without any sort of retention mechanism, other than gravity and
the stickyness of the Ceramique. Obviously, this would never be something
to consider for normal use, but by now I was getting desperate to try something
other than the stock heatsink on this board. After carefully placing the
Zalman 7000 onto the CPU die, I plugged the heatsink fan + Zalman Fanmate (set
to 5V) into the motherboard fan header and fired it up, while keeping my finger
firmly planted on the PSU power switch. As it turned out my fears were unfounded
as booting into the BIOS showed the CPU running at a chilly 30°C, just about
15°C cooler than with the stock HSF, and almost exactly the same temperatures
I was getting with the same CPU on the earlier AOpen 855G board. Booting into
Windows got me idle temps of 26°C, and CPUBurn load temps of 46°C, all
this with the 92mm Nexus fan in the Zalman 7000 running at an inaudible 5V.
Now this is more like it! But unfortunately, the machinations required to properly
and safely mount this heatsink on this board are way beyond the
average enthusiast.

Zalman
7000AlCu (modded), default Vcore (1.324V)*
CPUBurn

Load Temp.

°C
rise

°C/W (TDP)

°C/W (MP)

Zalman 7000/92mm Nexus
@ 5V
46°C
25°C
1.19
.92

*heatsink not bolted onto CPU, only perched precariously on
top.

So where does this leave us? Maybe "plan D" which would involve the
stock heatsink with a decent fan mounted above it on a Zalman FB123 bracket? I doubt that would work very well, as the heatsink itself seems to
be too ineffective to do its job sufficiently. How about "plan E",
having a machine shop drill the 52mm hole pattern into the base of something
like the old (but very effective) Alpha PAL6035? Maybe, but it still sounds
like too much work to me.

I could go on brainstorming forever, but I'm not going to. If this is the heatsink
AOpen wants to use on this board, fine, so be it.

FINAL THOUGHTS

AOpen is one of the very few companies that embraced the Pentium M for the desktop, with two Pentium M boards released over
the past six months, and a third, full ATX board that should be available
soon. The first Aopen Pentium M board was based on the older, less full-featured,
but stable and proven 855GM chipset. This second release is very modern, and full featured. That,
plus the m-ATX form factor seems like it would make the perfect solution for
an Intel-based quiet HTPC/multimedia system... and it almost is.

"Almost" because of a potentially fatal flaw in an otherwise
very nice board. The deficient heatsink: I'm
no engineer, but heatsinks can now do a great job of cooling
a 100+ watt processor. It's beyond me how someone can design a heatsink
incapable of keeping a 20W processor cool. Yes, Intel and others that
claim to know this stuff will throw around statements like "70°C
is a perfectly safe temperature for the CPU", and "if it doesn't
crash, it's stable", but the sort of person (including most SPCR readers)
that would be putting a system together using this board probably would be
bothered by such temperatures. And if it's not bad enough that the heatsink
is sub-par, there's no way to replace it with any aftermarket
heatsink because of the oddball form factor.

So where does that leave this motherboard? Fatally
flawed might be a bit too strong. It's actually a great board except for the silly heatsink mounting option. I suppose
if you could choose to ignore the high CPU temperature, or never tax the
CPU with a heavy load, it might be perfectly usable. In fact, the little 40mm
fan isn't too bad, is nearly inaudible across the room,
and its tone isn't too obnoxious even from as close as a few feet away. That may be
fine for some folks, but not for me. I'll pass on this one unless AOpen, or
some third-party vendor comes up with a decent heatsink for it.

PROS



* Four different video outputs

* Very stable

* 533MHz FSB support

* DDR2 support

* 2 SATA ports and 2 SATAII ports

* A plethora of USB 2.0 and Firewire ports

* HD Audio onboard
CONS



* Non-standard heatsink mounting

* Heatsink can't keep CPU cool under load

* No PS/2 ports

* Proprietary FDD port

Much thanks to AOpen
for the opportunity
to review this motherboard.

* * *

Discuss this this article in the SPCR Forums.

POSTSCRIPT - AUG. 7, 2005: AN IMPROVED HEATSINK

In response to overwhelming criticism from reviewers and
users, AOpen has introduced an improved heatsink/fan for its i915GMm-HFS
board. It is now standard on all shipping boards. The best news: Owners
saddled with the original heatsink can contact AOpen for a replacement.
See RH's review of the replacement HSF on the next page.

Mike Chin



Editor

POSTCRIPT - AUG. 7, 2005: AN IMPROVED HEATSINK

SPCR's review of AOpen's i915GMm-HFS Pentium M motherboard found it
to be a nice board hobbled by a sub-par heatsink. Not only
was it incapable of cooling a 20+ watt Pentium M processor, its proprietary
mounting design precluded easy replacement by any known aftermarket heatsink.

While testing this board, I had
several conversations with AOpen's tech people. They were aware of this issue
and claimed that a redesigned heatsink was in the works. Since nothing tangible
materialized by the review deadline, the review was published using the original
heatsink. Several weeks later a small package from AOpen arrived at the Hutter
Labs loading docks. Inspection of the contents revealed an aluminum heatsink
and translucent blue plastic fan that bore a strong resemblance to the original
915GMm HSF on steroids.

TECHNICAL DETAILS

This redesigned heatsink seems to be nothing more than a taller version of
the original heatsink. It uses a slightly different mounting style for the fan,
which may slightly help the fan's airflow but it uses the exact same fan, so
any additional cooling power will be coming from the increased surface area
of the fins.



New vs. old. At least 4x the cooling surface area.

Curiously, AOpen has rotated the foam spacer on the base by 45°,
which has apparently caused interference with the CPU die on several systems.



New vs. old. Why has the spacer been rotated 45°?

The new fan is the exact same 50mm x 10mm fan as supplied with the original
Lilliputian heatsink. At 12V it spins at 3400 rpm, but its sonic signature
is surprisingly benign. As decent as it sounds, hopefully the increased cooling
ability of the new heatsink will let us undervolt this fan to see how much better
it sounds when it's running slower than stock.



New vs. old.

Well, enough with the boring technical details, let's see if this
thing can turn a sow's ear into a silk purse.

THERMAL AND SONIC TESTING:

All thermal testing was done using the exact same setup and conditions as in
the original 915GMm review. The original eensy-beensy heatsink was also re-tested
for reference.

Happily, there's not a lot to write about here. The new heatsink performed
much, much better than the original heatsink. As unprepossessing as it looks,
it even had enough headroom to run the fan undervolted down to an inaudible
5V, as long as the CPU was undervolted to 1.100V. Cooling performance at 7V
was darn good, and the fan was just barely audible from 2-3 feet away. And like
the fan on the original heatsink, its sonic signature was more like that of
a quiet 80mm fan (think Panaflo 80L) than a tiny little 50mm fan.

Temperatures
with New Heatsink / Fan
CPUBurn

Load Temp.

°C
rise
°C/W

(TDP)
°C/W

(MP)
Fan @ 12V, default Vcore
61°C
40°C
1.90
1.54
Fan @ 12V, 1.100V
49°C
28°C
1.94
1.56
Fan @ 7V, 1.100V
55°C
34°C
2.36
1.90
Fan @ 5V, 1.100V
62°C
41°C
2.85
2.30
°C rise refers to the rise in temperature over the ambient at
load.


For default Vcore (1.324V) °C/W - TDP calculations:
Intel's TDP of 21W was used.

°C/W - MP calculations: CPUHeat
& CPUMSR Projects'
estimate of 26.8W was used.


For 1.100V °C/W - TDP calculations: a TDP calculated
value of 14.4W was used.

°C/W - MP calculations: A calculated
MP value of 17.9W was used.

CONCLUSION

The original review called the 915GMm a "great board, except for the
silly heatsink..." This new heatsink removes any concerns I had about
this board. It provides enough cooling to run the stock fan at nearly inaudible
settings, as long as the CPU is undervolted a bit. My Pentium M sample runs stable
at 1.100V, and every Pentium M that I've heard of will run stable between 1.1
to 1.2V so quiet cooling should no longer be an issue for this board.

So, SPCR is in possession of a working heatsink for this board, what about
everyone else? AOpen tells me that all the newer 915GMm boards have started
shipping with the improved heatsink, and owners that are saddled with the original
heatsink can contact AOpen for a replacement. How can you beat that?

Thanks to AOpen for
the revised heatsink.

* * *

Discuss this article in the SPCR forums.

Google

www SPCR