[mythtv-users] Lots of confusion as to which card(s) to buy for vdpau

Michael T. Dean mtdean at thirdcontact.com
Sun Feb 1 18:32:32 UTC 2009


On 02/01/2009 01:07 PM, Allen Edwards wrote:
> On Sun, Feb 1, 2009 at 9:52 AM, Jan Ceuleers <jan.ceuleers at computer.org> wrote:
>   
>> Allen Edwards wrote:
>>     
>>> So what is the point here?
>>>       
>> (...)
>>     
>>> What I am saying here is that computing power is computing power. Put
>>>       
>> (...)
>>
>> My expectation is that the savings can be much greater. Not only can you
>> make do with a much less powerful CPU (leading to lower power consumption
>> that you estimated at 20W), many of the other components in the system can
>> also be much less powerful (or are not used as intensively), because there
>> is less data to move across buses, and between memory and GPU.
>>
>> Furthermore, I would expect that hardware that has been purpose-built for a
>> particular function (in our example the GPU) is more power-efficient than
>> when that function is implemented in general-purpose hardware (the CPU). So
>> not all types of computing power are created equally.
>>
>> I have no numbers, nor am I in a position to collect numbers. Until then, my
>> remarks above are conjecture.
> I would assume the same thing that you did but when the required house
> power of the graphics chip started to go way up I decided to collect
> some numbers.  I am just trying to open up this discussion to some
> data.  If my data is wrong, that would be great.  It is the best data
> I could find.
>
> If the 8600 graphics chip draws 40 watts more than a 8400 chip as
> Tom's Hardware's numbers suggest it is going to be very difficult to
> make that up.  I would also assume that you can make a fine myth
> system with a 45 watt dual core processor without vdpau.
>
> If vdpau worked with a 8400, that would be a different story but
> apparently is doesn't.

Definitely a good idea to look at the system as a whole.  However,
you're forgetting that some people plan to use integrated (vs discrete)
graphics processors for VDPAU (or, perhaps you were responding to
someone who's adding a discrete GPU just to use VDPAU).  Integrated
GPU's tend to be closer to the 18-34W range.  However, I'm definitely in
the, "Just wait until it's ready to use of you want stable MythTV,"
camp, so I don't think people should be rushing out to buy new hardware
to use VDPAU to save some money on power usage.  After all, if you wait
for stable 0.22, you'll probably only be waiting a few months (and
really, how much will extra could you possibly save by switching even 6
or 9 months earlier).

The best measure, however, is the power draw from the wall.  I'm just
starting to measure my Myth backends' power usage (as I'm switching to
an 80 Plus PSU, so I'm collecting before/after data).  I run 0.21-fixes
and I don't have a GPU that can do VDPAU, so I can't test that. 
However, it would be nice to get a couple of people who are doing VDPAU
to try measuring their system's power usage while playing back (say, 1
hour of HDTV) with VDPAU and without VDPAU, and report their system
configuration (integrated vs discrete GPU, GPU and processor type) as
well as with/without numbers.

I'm guessing that they'll see the power usage go down a little bit with
VDPAU versus without it on the same system, but as you mentioned, if
you're adding a 100+W discrete GPU to the system, you're probably
increasing the overall power usage while playing back HDTV and while
not.  And, since you probably spend a great deal less time watching TV
during a day than not watching TV, adding a bunch of watts 100% of the
time to save a few watts during playback is focusing on the wrong part
of the problem.

And anyone who gets a tricked-out SLI rig for it is definitely not
saving power.  :)

As always, the best way to save money on power is to use mythshutdown
and only run the system when you need it.

Mike


More information about the mythtv-users mailing list