Viewing page 5 of 7 pages.
The first thing I do on a new build it pop a floppy in the floppy
drive and run Memtest86.
we have a problem.
There is no floppy drive or even a floppy drive header
on this Shuttle motherboard. Drat, now I have to leap into the 21st century
and play with bootable CDs.
Oh well, fortunately there's a bootable CD image
of Memtest86 on the website so I created one and proceeded to boot into Memtest86.
This is an intensive memory testing application that uses all the available
memory (since it doesn't run in Windows it has access to all the RAM). I always
run at least 20 loops of Memtest86 on every new build to make sure there's no
problems with the memory or memory timings. There's no reason to start installing
an OS if you're running with flaky memory; you're just looking for trouble.
I ran 18 hours of Memtest86 and it passed with zero errors. We're off to
a good start!
Next up is to install the OS, Windows XP Pro SP1 in this case.
All went fine here too. No problems cropped up, and after I booted into Windows
for the first time I installed the ATI RS300 Chipset, Audio, IGP Video, and
LAN drivers from the provided ST62K driver CD. Screen resolution was set at
my LCD monitor's default of 1280 x 1024 for the duration of testing. Next,
20+ MB of wonderful Windows patches and bug fixes to make things as safe as
possible. Then I installed some basic applications including a bunch of the
usual benchmark applications.
At this point on a normal build I will run 24 hours of Prime95,
a distributed computing application that loads the processor 100% and runs a
computational program that checks the generated answers against a series of
known answers. I use this application to test the CPU and memory subsystem for
stability. I also keep track of the system and CPU temperatures while running
Prime95 as it puts a very, very heavy load on the CPU. You can see how well
your cooling system will work under a heavy load by running an extended session
I ran into a partial snag here too. I installed Motherboard
Monitor, my standard temperature, voltage and fan monitoring application.
Unfortunately no temperature sensors, voltage readings or fan speed showed up!
I wasn't too surprised as this IGP 9100 is a brand new chipset that may not be supported by many monitoring applications. I
started looking for a monitoring app that would work and I found none. I tried
Aida32, SpeedFan, MB Monitor and Sisoft Sandra but none saw any sensors. Drat.
I really like knowing what's going on in the background of my systems, especially
when the system is completely new to me, as this small Shuttle is. I inspected
the included Shuttle driver CD hoping that they had put some sort of temp monitoring
app on the CD but found nothing. Contacting Shuttle directly was also no help.
I was going to have to do all testing and benchmarking without any idea of what
temperatures I was running, unless I booted back into the BIOS and checked those
Unfortunately, the BIOS temps don't have a lot of bearing on actual
Windows temps, especially when running in Windows at full load. So I sucked
it up and ran 18 hours of Prime95 to test for system stability. It passed with
zero errors. Apparently the CPU temp never topped the 60°C threshold that
I had set in the BIOS, as I never heard the ICE cooling fan ramp up in speed.
At least that's some consolation.
The SMART temp sensor on the hard drive was visible
with any of the HDD monitoring applications that I used. The drive
usually ran around 39-40°C during normal usage and topped out around 45°C
during heavy disk access. These numbers are almost identical to what I'm used
to seeing in my regular system and is a good indication that the ST62K's system
temperature isn't running too warm.
Here's PCMark2002. The newest version requires DirectX 9, which
the IGP 9100 chipset doesn't support. Using the exact same hardware in my desktop
system gave me almost identical results in the CPU and HDD score but a significantly
higher memory score of 7522:
PCMark2002 numbers for the ST62K.
Next up is 3DMark 2001SE. The newest 3DMark also requires DirectX
9. This score is nothing too great but it's much better than the score of 3755
that I got on my desktop system with it's ATi Radeon 7500 video card:
3DMark2001SE score for the ST62K's integrated
IGP 9100 video processor.
Now we'll look at some Sisoft Sandra scores. This is version 9.73.
The math and multimedia scores on the desktop system are almost identical to
the ST62K's but again, the ST62K's memory score was significantly lower than
that of the identical same sticks of memory, running at the exact same timings
in my Intel desktop system:
Sandra Math scores for the ST62K. Not bad
and nearly identical to my reference desktop system.
Sandra memory scores. Intel
desktop system scores 4712/4682 with the exact same memory. Something's odd
Sandra multimedia score for the ST62K. Neck
and neck with my reference system.
Most of the benchmarks are right where I would expect them to
be but there's something amiss with the memory scores. They're only about 80-85%
of the memory scores I'm used to seeing on other P4 DDR400 dual-channel motherboards.
Both the BIOS POST screen and the BIOS itself report the memory running at 200MHz,
dual channel but none of the Windows-based applications like CPUID and CPU-Z
are showing any information on the memory, exactly the same situation as with
the temperature monitoring applications that I tried to use with this new chipset.
I also tried two other memory performance/benchmarking applications to check
the performance of this RAM. They are "CTIAW" and "Cachemem".
Each of these applications also showed the memory to be running at about 85%
of what I'm used to seeing. I then swapped sticks of RAM from the PC3200 Corsair
that I was using to Mushkin PC3200 Level II memory.
I got the same results with
the Mushkin RAM as I did with the Corsair. I then went into the BIOS and manually
set the memory speed to "DDR333" and re-ran my benchmark applications.
The results of these runs showed that the DDR333 setting did indeed lower my
scores about as much as it should have, against the baseline DDR400 scores.
I tried many other BIOS adjustments but was never able to make the RAM perform
as I expected it to.
Could this lower than expected memory performance be due to the
integrated video chipset? There's no way for me to tell because this board doesn't
have an AGP slot and I don't have a PCI-based video card so I can't try and
disable the onboard video and re-run the benches so I'll just have to live with
it for now. A BIOS update or new drivers from ATI may also improve the memory performance.
Electrical Efficiency and Power Draw
I also checked the power draw of the ST62K, at idle and load,
at the default Vcore of 1.525 volts. After that, I went into the BIOS and lowered
the Vcore to 1.325 volts, a level that this same CPU has previously run 100%
stable at. I ran about 18 hours of Prime95 with the Vcore set at 1.325V with
100% stability on this Shuttle system as well. On the other two systems that
I've run the CPU undervolted by the same amount, my full load temperatures ran
a full 8-10°C cooler than when I ran the same CPU at it's default
Vcore. Quiet a big temperature drop for absolutely no decrease in performance.
I'd love to be able to see the same results with this ST62K, but of course I
have no temperature monitoring ability so I'll have to leave you hanging. I
did check the power draw of the undervolted system so we can compare the two:
SYSTEM POWER DISSIPATION
Note how much difference the lower Vcore makes
in the total power draw.
This usually equates to 8-10°C cooler
CPU temperature under full load.
Note that the AC power draw overall seems low, suggesting high AC/DC conversion efficiency. The PSU also has Active PFC, judging from the Power Factor measurements, which ranged from about 0.93 at idle to 0.98 at maximum power dissipation. According to Mike, who has examined umpteen PSUs, these numbers can only be obtained with Active PFC.