Intel D945GCLF2 Dual-Core Atom

**webwiznz** · 09-19-2008, 12:57 AM

After clarifying with andrewb at SPCR that his figures exclude drives, I think we are getting somewhere on the power front. As he wasnt able to really load the board, i did a few more tests this afternoon.

You can see that the difference between idle and max CPU is 4W, conincidentally the TDP of each CPU core. However ramping up the MCH with AtiTool at 15 frames a min, added a further 6.5W.

Some pics of the meter readings at bottom.

One thing to note is the saggy 12v input from the plugpack, which flows through into the 12v rail.

The HDD used has the following specs according to:
http://209.85.173.104/search?q=cache...esturn+digital

5 VDC
Read/Write 500 mA
Idle 400 mA
Standby 45 mA
Sleep 38 mA

I found this exercise interesting for a number of reasons. With the D201gly2 board, and any other sinlge core baord i suppose it doesnt take much to 100% the CPU, a simple winrar will do it. However D945GCLF2 needs more work to get there.

One winrar (156MB psd) on d201 takes 2m33s, whereas d945 takes 2m13s.

On D945 two simultaneous rars of same file takes 2m43s.

This is quite cool. If your multitasking, thats nearing twice as fast.

On D201 a single instance of prime95 would 100% the cpu. With D945 it took 4 instances to finally get there. Kinda strange really.

Heres the screenshots.

2 x prime 95

4 x prime95

4 x prime95 (blend mode, ram intensive)

Whats up here?, same problem as the photoshop resample. Looks like RAM is the (or RAM I/O) bottleneck on this system...

AtiTool

3 x prime95+AtiTool

Idle

And the meter readings:

4 x prime95
[media]http://pl.net/~pscott/o/4xp95.jpg[/media]

AtiTool
[media]http://pl.net/~pscott/o/ati.jpg[/media]

3 x prime95 +AtiTool
[media]http://pl.net/~pscott/o/ati+3p95.jpg[/media]

Idle
[media]http://pl.net/~pscott/o/idle.jpg[/media]

S3
[media]http://pl.net/~pscott/o/s3.jpg[/media]

Off
[media]http://pl.net/~pscott/o/off.jpg[/media]

What does all that mean? Well Andrewb's data indicates that at idle the board draws 2.7 Amps on the 5V rail, (he stated that 2.8 was an average, with 2.7A being the lowest reading). Without measuring it i think we can safely add a further 1.3A to that for MCH/graphics loading, which Andrew didnt do.

The moral of the story is that when at idle the board and one notebook drive draws 20.0 Watts, and under full load, 30.0 Watts. Just for the board itself 12v rail needs are minimal, and 5v rail needs start at:

MoB idle 2.7A
Add Load 1.3A
HDD 0.5A
----------
MIN 4.5A

Add a reasonable say 1A for USB devices and/or optical, which gives a reasonable working allowance of 5.5A at 5V. However if you want a MAX max:
then add 6 USB devices at 0.5A, plus 1A for optical, for a total of:

MoB 2.7A idle
Load 1.3 A
USB 3A
Opt 1A
HDD 0.5A
----------
MAX 8.5A

Edit: Some slim opticals can draw up to 2A at 5v.

Its pretty unlikely youll need all that power at once, but if you want safe thats it. Me ill go with the 5A figure and thus the sh-6080 DC-DC boards in these mini-itx cases *squeak* in.

Edit: Wrong. My original 6A rated at 5v PSU board died after one week. See, post a page or tow forward. You will need a PSU capable of at least 8amps on 5v rail.

The BB screenshot images arent that clear, so if you want the fullsize views use these:

$ttp://pl.net/~pscott/o/excell.gif
$ttp://pl.net/~pscott/o/rar.gif
$ttp://pl.net/~pscott/o/2xrar.gif
$ttp://pl.net/~pscott/o/2xp95.gif
$ttp://pl.net/~pscott/o/4xp95.gif
$ttp://pl.net/~pscott/o/4xp95_blend.gif
$ttp://pl.net/~pscott/o/ati.gif
$ttp://pl.net/~pscott/o/3p95+ati.gif
$ttp://pl.net/~pscott/o/idle.gif

**Punky** · 09-19-2008, 01:01 AM

how tweakable is this bios? can u turn off a lot of things for faster booting?

**wywywywy** · 09-19-2008, 03:03 AM

Has anyone got a problem with the SATA controller on this board?

When I use my IDE drive, everything is ok.

When I add an SATA drive to SATA1 channel, the PC would freeze for a few minutes every time I access the drive.

When I add it to SATA2 channel instead, the drive appears as PIO mode instead of DMA5, and is a looooooot of CPU time! Not to mention very slow transfer rate.

This is driving me mad. Any ideas?

Thanks.

**wywywywy** · 09-19-2008, 07:13 PM

An observation by the way - Speedstep doesn't seem to work, although we don't desperately need it.

**Woofnstuff** · 09-20-2008, 12:39 AM

my sata works fine..

the issue i have is the s-video output just comes up in the device manager as unknown device, after reloading the drivers off the cd, and done a full reinstall of windows, a different version even ... im stumped.

and as there is no listing on intel's web site for the board i can get any other drivers

EDIT:
just reinstalled of a different version of windows and it works i must have removed something important when i used nlite

**wywywywy** · 09-21-2008, 04:00 PM

Found out what caused the SATA issue.

It seems like this board (don't know whether it is design issue, BIOS, OS, or driver) is not compatible with SATA drives that has NCQ, when used together with an IDE drive.

Changing the SATA mode from Native to Legacy in the BIOS, thus disabling NCQ and other SATA specific features, has solved the problem.

Another observation - if you change the fan to another quiet one such as the Sycthe Mini Kaze (the original one is pretty noisy), then don't use automatic fan speed control. It doesn't seem to provide enough voltage to start up low-rpm fans. I now set my fan speed to constant 90% on a Mini Kaze (about 2000rpm) and it is very cool and quiet.

**the_gooner** · 09-21-2008, 04:22 PM

Sorry Noob Question but what is NCQ?
going to wait for this board to come down in price before buying
see what other issues there is ?
got another board on its way but looking at having this as a media centre in the house

**crooper** · 09-22-2008, 12:26 PM

webwiznz, thanks for the detailed wattage findings. Nice to see that dual core seems to actually make a difference.

What kind of case is that in your pictures?

**crooper** · 09-22-2008, 12:28 PM

Originally Posted by the_gooner

Sorry Noob Question but what is NCQ?

Native Command Queuing

"a technology designed to increase performance of SATA hard disks under certain situations by allowing the individual hard disk to internally optimize the order in which received read and write commands are executed"

http://en.wikipedia.org/wiki/Native_Command_Queuing

**webwiznz** · 09-22-2008, 11:10 PM

crooper: its a morex/procase cs 3688

http://www.procase.co.nz/pc/cs3688.htm

But, watch out for the PSU. For the record the DC-DC baord in this case is now officially dead after 1 week of use, and the DC-Dc board RMAed.

I gather they make a variant of this case with a 150W PSU.

http://www.morex.com.tw/products/pro...l.php?fd_id=91

Which is rated at 8 amps on 5V rail. The 5v power transistors on the Dc-DC baord in the Noah case are much bigger.

The atom board has been moved to a Noah case.

http://www.procase.co.nz/pc/noah.html

Which is rated at 7A on 5v. But the Noah case is certainly roomier to work in with better airflow and a pretty quiet system fan included.

To be honest im not even sure about the Pico120
Min Max
5V 6A 8A
5VSB 1.5A 2A
3.3V 6A 8A
-12V 0.1A 0.1A
12V 6A 8A

And its not certain that the Pico will clear the drive tray, it might be tight.

M2 is rated at what 8A on 5v. But id be even happier with dsatx at 12A on 5v. I guess we will be waiting a while for the manufacturers to produce PSUs for these baords, or more likely its us who will be waiting for Intel to abandon the 945 NB.

The message is still: this board had big 5V needs, choose your PSUs carefully.