Mini ITX Build

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
I'm considering building a mini ITX system for use as a Windows 7 Media  
Center.  I have a couple of Silicon Dust HDHomeRun Primes running, so it  
would have to be able to stream HD video.

Perhaps it's backwards, but I found a case I like:

It only has a 90W power supply, however, and I'm afraid that it won't be  
able to power the rest of my system:

ASRock H77M-ITX LGA 1155 Mini ITX Intel Motherboard

Intel Pentium G2020 2.9GHz 55W Dual-Core Processor Intel HD Graphics

Kingston HyperX Blu 4GB (2 x 2GB) DDR3 Desktop Memory

Some 5400rpm Laptop Drive I have lying around until I buy an SSD


Q1) Are my fears founded that 90W won't be enough to power this thing?

Q2) Would this system be fleet enough to stream HD video?

I really like that I can attach this box to a TV's VESA mount, but may  
have to give that idea up.  There is also this barebones system, but  
it's sort of spendy:

Any suggestions you have along the lines of this project would be  

Re: Mini ITX Build

On 11/3/2013 2:39 PM, Grinder wrote:
Quoted text here. Click to load it

Edited to add this:

Q3) Do you think this thing would be able to stream HD video?

Re: Mini ITX Build

Grinder wrote:
Quoted text here. Click to load it

To build up a mini-ITX, you have to find characterization
information to "reduce the risk" of screwing up the build.

Now, the first problem, shows up in the reviews for your
computer case. One review, claims the case uses enough plastic,
that his build had an RFI (radio frequency interference) problem.
The external adapter is rated 90W, and supplies
power to an open-faced DC-DC converter inside the Antec case.
It's not a Pico PSU, and instead is a custom build of some
sort. The internal power supply is only rated 80W, as stamped
on the unit itself. So that's a 90W adapter, feeding an 80W

    "total power 80w max
     12v 5A max
      5v 6A max
    3.3v 5A max
    -12v 0.1A max
    5vsb 1.5A max
    Combined 3.3V + 5v power = 45w max"

   [Combined power all rails less than 80W ???]

That means Vcore can use at most 60W. At 90% VCore efficiency,
that gives you 54W at the processor VCC pins. I think that
still might be enough, as your other load would be the
cooling fan.

I found in a review of the G2020 processor, someone
claims he measured 40W (out of the 55W TDP rating).
TDP power ratings work in "classes". The processor
at the top of the class, comes closest to hitting
the TDP. Processors with lower clock rates, won't
be exactly running at the TDP. The top one in the
G2020 class was a 3.3Ghz processor still claiming
to draw 55W. So your processor will be less than
the top of the line one. That's where your
margin comes from.

To give two examples, I have two 65W Core2 processors.
One measures 36W on the 12V rail, the other measures
43W on the 12V rail, while running Prime95 Torture test.
(They were 2.6GHz and 3GHz respectively, and different
processor generations.)

When I've done the measurement in the past, I found
the older P4 processors (Northwood), the measured power
was bang on the TDP value. So those were riding on the
edge. (One actually went over.) These processors seem
to be a little bit shy of the limit. And that's why
the 54W limit above isn't a killer.

"High tech gambling", is when you attempt to build
one of these things, with no "Plan B" in place.
For example, would your build be ruined if you
had to find another case for it ? If the case was
3" wider or 2" taller, would that kill the deal ?

The problem with your 80W supply above,
is if the motherboard happened to draw 6A from
the 3.3V rail, that's only 20W, but it would
still cause the power supply to switch off.
Each rail is a limitation. The mini-ITX motherboard
designer, has to be very careful, to "balance" the
loads. For example, if I was designing it, I'd
run the CPU Vcore off 12V (as usual). I'd run
the VDimm switcher off the 5V rail. I'd run
the chipset off the 3.3V rail. With the
objective being, to balance the current flow.
On a regular motherboard, it wouldn't matter,
and I could wire everything to 12V and not
give a rats ass.

When the mini-ITX standard came out, they should
have switched to pure 12V power, and done a single
rail design, to make these designs easier to build up.
That would have made things much less risky for the
system builder.

Since there is no margin for error on these
systems, you need a plan B if you screw up.
Bigger case, bigger power supply, adding a
video card to compensate for bad processor, etc.

(For example, as a Plan B, this site offers a
160W Pico PSU, needing a pretty beefy wall adapter.
Now, would this fit in your computer case ? Would
this blow the budget ?)


The Gigabyte product will make a nice DVD player
core. The processor appears to be BGA, and that
means it is soldered to the motherboard. This slightly
limits your Plan B options, if you screw up.

Maybe if you can find a review for it, they'll
quantify how capable it was with higher bitrate
video content.

A general rule of thumb, was you needed 1.5GHz
to play a DVD or do the most economical video
formats. And that ruled out some of the boards
with VIA processors, as they just didn't have the
power for the job. I expect the two Intel ones
you're looking at, are above that threshold, and
would be better than what VIA was offering in
their processor line. With a VIA system, the video
card is the critical part (i.e. buying a good video
card with acceleration options for DXVA, to compensate for
the CPU, then hoping the CPU can do the finishing
steps in playback without bogging down).

Some video card acceleration blocks, they claim to
do the entire decoding operation, but they actually
cheat and still require some processor assistance.
And that's why you can't use a "zero Hertz processor"
and expect to see smooth video. So even with the
latest low power video card, the CPU still does
some of the work.

All we can really say about the modern video acceleration
features, is it is "better than IDCT". That was a
feature added to assist decoding, without
really providing a significant boost to
performance. You couldn't fix a weak CPU,
purely with IDCT. As it was only a small
fraction of the whole decoding process.

To give another example, purely for illustration.
I had a system with a P4 in it. I played video
windowed, at native resolution for the video.
Then I asked the player to go "full screen".
That requires pixmap scaling. My video card
didn't have a scaler. It took 40% CPU just to
scale the picture to fill the screen. When I
tested with a more recent version of video card,
one with a scaler, CPU usage dropped back down
to 10-15% or so. And this is why, having a little
extra margin on CPU MHz, pays off. You never know
when some missing feature somewhere, is going to
make the CPU have to make up the difference.

I've found quite a range of CPU requirements
for video players. I can have one player that
plays a movie and uses 10% CPU. Then, using the
same movie, select a second player, have it
hit 100% CPU, and I'm dropping frames. You can
waste a fair amount of time, just narrowing down
decent software to use.


Re: Mini ITX Build

On 11/3/2013 12:39 PM, Grinder wrote:

I have a Asus mobo with intel D525 atom and Nvidia ion2.
Quoted text here. Click to load it

It streams video just fine. The power goes up to about 45 watts when the  
Nvidia chip cuts in. I use win7 pro on it. I run XBMC, so I can't  
comment on Win 7 Media center.

Now mine is the old 40nm atom. Even if you could find an old D525, I'd  
go for newer technology. But nothing beefier than the atom.

I have the older Silicon Dust receiver. Good stuff.

While I hate paying the Microsoft tax, if you want to be able to stream  
from the maximum number of sources, you really need windows. You can  
probably use linux, but it will be more work.

Re: Mini ITX Build

On 11/4/2013 9:25 PM, miso wrote:
Quoted text here. Click to load it

Good to know.

Quoted text here. Click to load it

Maybe it's just odd verbiage, but it looks like Windows 7/8 is the best  
client for the SiliconDust cable card device because of DRM issues.

Re: Mini ITX Build

On 11/5/2013 3:50 AM, Grinder wrote:
Quoted text here. Click to load it

It is that damn MS Silverlight. MS has given up on it, but it is still  
used. If you want maximum compatibility, you need to build the HTPC on  

When it comes to HTPC, think small. It is hard for people who build  
their own PCs to go lightweight. At least I have that problem. But when  
I have a PC dedicated to a task, I try to limit the hardware  
accordingly. You don't need much of a SSD for a HTPC since you will not  
be loading a lot of software. I keep the ripped DVDs on a usb drive  
outside the box.

I have a BeagleBoard XM I got for a project that got sidetracked. Under  
Ubuntu, it did a nice job on HDMI. I didn't do any serious analysis on  
the quality, but it looked fine. I'm not a fan of Ubuntu, but they do a  
decent job on Arm.

Remember Roku does their streaming on a freakin' dongle.

I went through a few Asus SBCs reviews on Newegg yesterday when I did my  
post. Lots of complaints in the reviews. But that was exactly the case  
with the Asus Atom I bought. Eventually the Newegg customers figured out  
the problems. Literally, I ignored the Asus ram recommendation for the  
suggestions in the reviews.  Other than a bad fan that I had to replace,  
the computer has been very reliable.

I'd trawl the Newegg website for Intel Atom and AMD Fusion mobos, and  
see which ones have the least complaints. It is time consuming to pick  
the right mobo, but an I3 or higher is just beyond what you need for a HTPC.

Note XBMC runs on windows, linux, and Mac. I wouldn't rule out a Mac  
Mini for HTPC, but I don't like how everything is soldered on the mobo.

I'm not so sure I'd be messing with windows media center since MS has  
abandoned that as well. Yeah it is still there, but they do nothing with  

There are a number of smartphone apps to run XBMC remotely, though a  
wireless keyboard is fine if you don't want the security risk.

Site Timeline