Asus P5N7A-VM: Geforce 9300 IGP

nVidia has been the laggard in chipsets for Intel processors, but their GeForce 9300 may help them turn things around. We checked out the 9300-based Asus P5N7A-VM, and we liked what we saw.

Dec 1, 2008 by Lawrence Lee

For awhile there it seemed like nVidia was stuck in the bargain-basement chipset
market when it came to the Intel platform. Budget users looking for an Intel-chipset had to choose between aging motherboards based on the 945G/GC chipset, which
had spotty CPU support, and the G31 boards, which had limited features
and a relatively high price-tag. nVidia’s GeForce 7 boards, by comparison, were
a bargain, delivering plenty of features at affordable prices, though at the
cost of having only a single memory controller.

When more advanced integrated chipsets like Intel’s G33 and G35 were released,
Intel pretty much had free-reign, despite offering poor 3D performance. Their
only real competition came from nVidia whose GeForce 8, for whatever reason,
never materialized in LGA775 form. For Core 2 users seeking a top-end
motherboard, Intel was the only game in town and they paid a premium for it. Now, Intel
finally has a little competition —
their latest G45 chipset may have a worthy opponent in nVidia’s GeForce 9300
chipset. We have an Asus P5N7A-VM GeForce
9300 motherboard on the test bench today and we’re hoping it can break Intel’s
virtual monopoly.

The star of GeForce 9300 is its integrated graphics processor. It
is nVidia’s first DirectX 10 capable IGP and comes complete with Shader Model 4.0
support as well as 16 shader processors. It also supports Hybrid Power/SLI,
CUDA, and PhysX just like the discrete versions of the GeForce 9 series.

LAYOUT

A board’s layout is important in several regards. The positioning of components
can dictate compatibility with other products (mainly third party heatsinks)
and also ease of installation. Poorly placed power connectors can also disrupt
airflow and thus make the system more thermally challenging.

The PCB layout is pretty clean with all the major power and interface connectors
on the edges. The CPU socket is clear of any obstructions — the capacitors
and chokes in the vicinity are very short. The board has 5 SATA ports (6 if
you count the eSATA connection on the back pane), a single IDE channel sitting
on its side, and 3 fan headers.

The board has a single chipset heatsink held on with 4 hooks. It measures 34mm
tall from the PCB surface. There is no cooling provided for the VRMs,
but all the board’s capacitors are of the solid-state variety.

The chipset cooler has 6 stubby horizontal fins on both sides and 5 vertical fins on the inside, all with slits to increase the overall surface
area. Strangely, beneath the Asus face-plate sits a solitary heatpipe shaped
like a racetrack. It seems completely out of place, both aesthetically and in
terms of heatsink design.

None of the space on the rear I/O shield is put to waste — it is stocked
with S/PDIF, HDMI, DVI, and eSATA connectors. This is also the first time we’ve
seen a DisplayPort on the back of a motherboard. The only common feature the
P5N7A-VM is missing is FireWire.

TEST METHODOLOGY

Test Setup:

• Intel
  Core 2 Duo E7200 processor – 2.53GHz, 1066FSB, 3MB L2 cache, 45nm,
  65W
• Corsair
  XMS2 memory 1GB, DDR2-800
• Western
  Digital Scorpio notebook hard drive – 320GB, 5400RPM, 8MB cache,
  SATA
• Asus
  BC-1205PT Blu Ray drive – SATA
• Seasonic
  SS-400ET mATX power supply
• Arctic Cooling Alpine
  7 Pro CPU cooler
• Microsoft
  Windows Vista SP1 operating system – Home Premium, 32-bit
• nVidia
  Forceware 20.08 graphics driver
• QuickTime
  Alternative 1.81 codec package

Measurement and Analysis Tools

• Cyberlink
  PowerDVD 8 to play video.
• Prime95
  25.6 processor stress software for loading the CPU.
• ATITool
  0.27 Beta 4 artifact scanner to stress the integrated GPU.
• CPU-Z
  to monitor CPU frequency and voltage.
• SpeedFan
  4.35, used to monitor temperature and fan speeds.
• Seasonic
  Power Angel AC power meter, used to measure the power consumption
  of the system.
• Custom-built, four-channel variable DC power supply, used to regulate
  the CPU fan speed.

Our main test procedure is designed to determine the overall system power
consumption at various states (measured using a Seasonic Power Angel), and to
test the integrated graphics’ proficiency at playing high definition video.
Standard HD-DVD and Blu Ray discs can be encoded in three different codecs: MPEG-2, H.264 / AVC and VC-1. MPEG-2 has been around for years
and is not demanding on modern system resources. H.264 and VC-1 encoded videos,
on the other hand, are extremely stressful due to the complexity of their
compression schemes and will not play smoothly (or at all) on slower PCs,
especially those with antiquated video subsystems.

We use a variety of H.264/VC-1 clips encoded for playback on the PC. The clips
are played with PowerDVD 8 and a CPU usage graph is created by the Windows Task
Manger for analysis to determine the approximate mean and peak CPU usage. High
CPU usage is indicative of poor video decoding ability on the part of the integrated
graphics subsystem. If the video (and / or audio) skips or freezes, we conclude
the board’s IGP (in conjunction with the processor) is inadequate to decompress
the clip properly.

Video memory was set to 128MB, SpeedStep was enabled, and the following OS features / services
were disabled to prevent spikes in CPU or hard drive usage that are typical
of fresh Vista installations:

• Windows Sidebar
• Indexing
• Superfetch

Video Test Suite

1080p | 24fps | ~10mbps

H.264: Rush Hour 3 Trailer 1 is encoded in H.264 with Apple Quicktime.

1080p | 24fps | ~7.5mbps

WMV3: Coral Reef Adventure trailer is encoded in VC-1 using the WMV3 codec (commonly recognized by the moniker, “HD WMV”).

720p | 60fps | ~12mbps

VC-1: Microsoft Flight Simulator X trailer is encoded in VC-1 using the Windows Media Video 9 Advanced Profile (aka WVC1) codec — a much more demanding implementation of VC-1.

 

1920×1080 | 24fps | ~19mbps

VC-1: Drag Race is a recording of a scene from network television re-encoded with TMPGEnc using the WVC1 codec.

 

1920×1080 | 24fps | ~33mbps

Blu-ray: Disturbia is a short section of the Blu-ray version of Disturbia, the motion picture, played directly off the Blu-ray disc. It is encoded in H.264/AVC.

TEST RESULTS

Our test system was fairly basic, featuring a Core 2 Duo E7200 (45nm, 65W TDP) cooled
by an Arctic Cooling Alpine 7 Pro, and a single stick of Corsair
memory. The heatsink fan is connected to an external DC fan controller so that
the fan’s power draw does not come into play. The rest of our test platform consists of an Asus Blu-ray drive,
a 5400RPM notebook hard drive, and an OEM Seasonic 400W power supply. The operating
system used is Vista Home Premium SP1 (32-bit).

We tested the board with the CPU at stock settings with SpeedStep enabled and
underclocked to 1.2GHz with the minimum stable voltage which turned out to be
0.93125V. It should be noted that our CPU is stable at 0.85000V on other boards.

At stock settings, the system idled at 41W, drew between 45W and 54W during
video playback, and pulled a modest 68W at full load. Stressing the IGP did not significantly
increase power consumption. The GeForce 9300 graphics chip did not have any
problems with our test suite, passing it with flying colors. The CPU usage was
relatively low during video playback, and  the CPU stayed at 1.6GHz (thanks to SpeedStep) during the majority of playback except in the case of the Flight
Simulator and Drag Race clips when it increased
to its full 2.53 GHz frequency. We did not encounter any playback problems
with our Blu-ray
title, HDCP-related or otherwise.

At 1.2GHz and 0.93125V, power consumption was 2W less at idle, about 4W during
video playback (except during the Rush Hour clip) and approximately 15W lower
on full load. Despite the extremely low clock speed (none of Intel’s desktop
dual core processors are less than 1.6GHz), our entire video test suite played
smoothly without any problems.

Few boards we have tested can claim the functionality of the P5N7A-VM — the one that
comes closest is the Asus
P5Q-EM. They are both from the same manufacturer, and they have similar feature-sets
and prices. It the closest we can get to a direct nVidia
to Intel chipset comparison without using reference boards.

At stock speeds, the P5N7A-VM is more power efficient across the board. It
had
a savings of 3W at idle, 8W during H.264 playback, and 7-8W fully loaded. It required fewer CPU cycles
to play our test clips — a testament to the strength of nVidia’s GeForce 9300
video subsystem. Overall, it’s the most efficient mATX LGA775
board we’ve ever come across.

The P5N7A-VM could not run our E7200 processor stably at 1.2 GHz with any less
than 0.93125V (the P5Q-EM required only 0.85000V). As such, the
P5N7A-VM’s superior power efficiency all but evaporated when compared in our
underclocked 1.2 GHz test. There was only a 1-2W difference between the two
boards in this state — not enough to make a meaningful difference — though it
should be noted that, even with the higher CPU voltage, the P5N7A-VM was still
the more efficient of the two.

Fan Control

When it comes to customizable control, SpeedFan is our application of choice.
If properly supported, it can be configured to raise/lower multiple fan speeds
to designated limits when any specified temperature threshold is breached.

SpeedFan duplicates most of the functionality of Asus’ PC Probe application,
except for the +5V reading. To enable fan control, set PWM modes 2 &
4 in the Advanced menu to “Manual PWM Control.” Speed02 and Speed04
provide full control for the CPU and CHA fan headers respectively.

Q-Fan

We tested the board’s Q-Fan feature by connecting fans to the CPU and CHA fan
headers, but connecting the CPU cooler’s fan to a variable DC fan controller,
so we could lower it as necessary. We then proceeded to stress the processor
using Prime95 and monitored the fan speeds using SpeedFan. The fans used were
a Scythe Kama 92mm PWM fan (maximum 2560 RPM) and a Scythe Kama Flow 80mm (maximum
1550 RPM).

We found that once the trigger temperature is reached the CPU fan speed increases
over a span of 8-9°C from between 600 and 2600 RPM no matter which setting
was used. The only difference seemed to be the trigger temperature with “Performance”
starting at 54°C, “Optimal” at 62°C and “Silent”
at 67°C. The CHA fan speeds on the other hand varied very little. It should
be noted that we actually had to turn the CPU cooling fan off before the CPU
temperature got high enough to make the fans ramp up. Either our CPU was running
especially cool, or the fan control just wasn’t aggressive enough. We did not
experience any instability due to overheating.

Overclocking

With effective third party heatsinks and the many low power
CPUs available on the market, overclocking can improve performance without
compromising the noise level of a silent PC.

A simple overclocking investigation was conducted with the CPU multiplier set
to 6x and RAM to its lowest speed with an extra 0.3V. The CPU frequency was
increased in increments of 5 MHz / 10 MHz until the system failed a 5 minute
stress test consisting of Prime95 running simultaneously with ATITool 3DView, failed to boot,
or showed other signs of instability. We then maximized the multiplier and increased
the CPU voltage to a stable level for our final overclock.

The highest stable FSB we achieved was 1480MHz, yielding a CPU speed of
3.33GHz, or 39% above stock. At higher frequencies, the system froze when ATITool
was loaded. Note that this result was obtained without adjusting the chipset
voltage, so it’s possible the board can overclock further with a little extra
juice. These are simply the out-of-the-box results.

Cooling

To test the cooling on the board, we lowered the CPU cooling fan’s voltage
to 6V to reduce the amount of top-down airflow the nearby components received.
We then stressed the system with Prime95 and ATITool and whipped our our handy
IR thermometer to check the results.

After about 20 minutes of load, the MOSFETs and chokes around the CPU socket
were in the 55°C to 65°C range. While this may sound high, we’ve seen
VRM temperatures of 90°C and higher — these results are mild in comparison.
The hottest point on the chipset heatsink meanwhile was only 58°C, also
relatively cool. Also keep in mind this is on an open testbed with very little
airflow. In a case with proper cooling, these temperatures should be lower.

3D Performance

To get a rough estimate of how well the P5N7A-VM’s onboard video plays games,
we ran 3DMark05/06. As synthetic benchmarks they have limited value, but they
should give you a rough idea of how well it performs.

GeForce 9300 scored about four times higher than GeForce 7050. In fact this
is the first time an IGP has managed to match the 3DMark scores of the discrete
HD 3450 graphics card in our lab. It bested both Intel’s GMA X4500
and AMD’s HD 3200 and, more impressively, it did it without higher power consumption.
We’re not willing to call it the fastest IGP ever though — we have yet to test
the GeForce 9400, and we were never able to get any 3DMark results for the HD 3300
chipset. Reports from other web sites seem to indicate that these are even faster GPUs.
The limits continue to
be pushed.

HDMI Output

The board’s HDMI output worked flawlessly with our Asus
MK241H LCD monitor at 1920×1200 resolution. There was no overscanning
or other video issues — it looked identical to the DVI output. The audio functioned
perfectly as well, sending stereo sound to the monitor’s built-in speakers
without requiring any fiddling with the sound settings in software.

FINAL THOUGHTS

On paper, the Asus P5N7A-VM is very similar to the
P5Q-EM,
but, head-to-head in the real world, the GeForce 9300 board
came out on top in almost every way. The P5N7A-VM pulled less power from the
wall (the lowest we’ve ever recorded for a board of its type), was more efficient
at handling high defintion video playback, and delivered much better 3D performance.

While the P5Q-EM provides more elaborate CPU / chipset / memory BIOS
options, the P5N7A-VM actually overclocked better out of the box without
modifying any voltages. It also allows adjustment of the IGP’s clock and shader
frequencies. Unfortunately, it could not undervolt as well, failing to POST when
the CPU voltage was set below 0.93125V. The P5Q-EM, on the other hand, could be
set to its minimum 0.850V, but the difference in voltage
only allowed the P5Q-EM to pull even in terms of overall power consumption when our
E7200 CPU was underclocked.

The feature lists of the two boards are almost identical. The P5N7A-VM adds a DisplayPort output and eSATA
on the back panel, while the P5Q-EM has FireWire instead (eSATA is offered via an adapter). The P5N7A-VM
is also a bit cheaper, though still fairly expensive at $130.

The Asus P5N7A-VM is probably not a good choice for a simple file server or
light-use machine given its price and feature-list, but it would be perfect
for an HTPC or any situation where you would need the processing power of an Intel
Core 2 processor in a small package. We’re happy to see that nVidia not only
gave Intel a run for their money, but solidly beat them in most areas. It hit
all the right buttons with us: features, power consumption, video playback,
and especially overall efficiency. The P5Q-EM is a good board, but the P5N7A-VM is simply
better. nVidia: Welcome back — you’ve been missed.

Our thanks to ASUSTeK
for the Asus P5N7A-VM sample.

* * *

Articles of Related Interest

Asus P5Q-EM G45 mATX motherboard

Asus
M3A78-T: AMD’s IGP Gets Another Boost
Intel
DG45FC: Loaded LGA775 Mini-ITX Board
Intel
DG35EC: G35 mainstream mATX board
Intel
D945GCLF m-ITX: Atom For The Desktop

Zotac
NF610i-ITX: A Compact Core 2 Solution

* * *

Discuss
this article in the SPCR forums.