[mythtv-users] MythTV backend running on NVIDIA Ka-el, 35W replaced by 1W

Raymond Wagner raymond at wagnerrp.com
Tue Mar 8 22:00:16 UTC 2011

On 3/8/2011 15:40, mythtv wrote:
> I'm sure there are others who have their backend on 24/7.  Is anyone else
> (esp the devs) intrigued by the idea of replacing a 35W CPU  (+ additional
> system power) with a 1W CPU with higher performance?
> http://blogs.nvidia.com/2011/02/tegra-roadmap-revealed-next-chip-worlds-first-quadcore-mobile-processor/

Take a close look at those pictures.  The ARM builds are done using 
heavily optimized settings and GCC 4.4.1.  The Intel builds are done 
using typical settings and GCC 3.4.6.  Just building with the (much 
much) newer 4.4.1 puts the Core 2 ahead of the ARM, and by using 
similarly optimized settings, the Core 2 is a good 50% more powerful 
than the ARM.  Also understand that this test is only performing integer 
math, and the ARM platform has traditionally had pathetic FPUs.

So 66% of the performance at only 1W consumption, that's pretty good, 
right?  Well you're still not getting the whole story.  That is one 
measurement using a synthetic benchmark, which as clearly shown can be 
falsified, and it is done using a quad core part.  The scheduler is 
single threaded.  The independent backend and MySQL bits are done 
sequentially, and not in parallel.  The backend code is not parallel, 
and the sql calls are not something that can be broken into multiple 
threads well by the MySQL server.  On a single threaded workload, the 
ARM is now only 33% the performance of the Core 2.

Commercial flagging is going to be a bit different, because the decoding 
and detection can be handled in independent threads.  Video decoding in 
North America is still either going to be MPEG2, which is single 
threaded, or H264 out of an HDPVR, which is single sliced and thus still 
single threaded.  The T7200 at 2.0GHz won't quite be capable of handling 
full bitrate HDPVR output in real time, so the ARM at less than half the 
performance per core won't come close.  If you intended to live with the 
scheduler constraints on an under powered backend, you would still want 
to have a separate machine (maybe your frontends) do your video 
processing for you.

Now lets look at power consumption.  The Core 2 is rated at 34W TDP, but 
that's both cores at full speed, plus heavy cache use, it's absolute 
worst case scenario.  More realistically since this is largely single 
threaded, it's going to be closer to 20-25W while running the 
scheduler.  When it's finished in 1/3 the time of the ARM, it will drop 
back to low power mode, and being a laptop part, it will be well under 
10W.  10W run non-stop, at average North American utility rates, equates 
to around $10/yr in power consumption.  Even the desktop processors can 
be downclocked when idle such that the entire system will run under 25W 
at the wall.  Do understand that a significant portion of the power 
consumption is going to be from the attached tuners, hard drives, and 
STBs (if you need analog capture), which are going to be the same 
regardless of what CPU you're using, and will likely end up consuming 
far more than that 10W idle power of a mobile Core 2.

Let's take this a bit further.  This T7200 chip they're comparing 
against was a release part.  It's one of the original 65nm processors 
released in mid-2006 when the Core 2 line was first launched.  They're 
comparing their brand new not-yet-available processor, to one that's 
nearly five years old, and several generations passed on both the 
microarchitecture, and fabrication techniques.  I doubt they're even 
still available for purchase.  I'd like to see the comparison between a 
modern dual core Sandy Bridge part, with a 17W TDP and turbo speed of 
2.7GHz, or quad core part with a 45W TDP and turbo speed of 3.4GHz 
(turbo being where one or more cores shut down to allow others to run at 
higher speed in the same power envelope).  It will be even more 
interesting the AMD Bulldozer parts due out in a few months, where the 
cores physically have a trench dug around them, with gating to allow 
whole chunks of the chip to be completely powered down, and even the 
high end eight and sixteen core parts are expected to have an idle 
consumption under 10W

> I was surprised that I couldn't find any discussion of the backend running
> on ARM but then again ARM CPUs have never had this kind of horsepower
> before.

There have been, but they've all come to the same conclusion that ARM is 
not sufficient high performance to recommend for a backend.  Individual 
users claimed it was 'good enough', but they had limited channel count, 
with one or few tuners, and were willing to put up with the minute or 
longer scheduler runs.

More information about the mythtv-users mailing list