Would it be possible to build a AMD64-specific build of KnoppMyth? Given how inexpensive these processors have become, it would seem to be an excellent branch for KnoppMyth to pursue.

When I had my old Athlon 1800+ setup, I had given up trying to use anything but KnoppMyth because it was just too frakking easy to think of using anything else. Well, a *pop* and a failed smoke test later, and I bough an AMD64 motherboard/cpu combo with a new stick of memory for good measure. Less than $200 including shipping from NewEgg gave me a Sempron64 3000+ processor, ECS nForce3 MB and 512Mb of RAM. I know that I could just use a 32-bit version of KnoppMyth, but I just don't want to give up the extra "oomph" of 64-bit, especially given my dual-tuner HDTV/SD setup. I'm more than willing to test, help build, etc.

- Stealth Dave

This has been considered, and will most likely be looked at for R6, however it will not happen for R5.

_________________
Have a question search the forum and have a look at the KnoppMythWiki.

Xsecrets

Well stated, Xsecrets.

Has anyone on Gossamer reported compelling speed differences between A64 on 64-bit vs. 32-bit? I have not heard of any, but I don't follow the mailing lists very closely anynore.

_________________
Do you code to live, or live to code?
Search LinHES forum through Google

I'm curious about this as well. I just got hold of a EM64T CPU + mobo + DDR2 and am dying to seem them live up to their abilities. On the other hand, if the improvement is not worth it, I'd rather spend the time trying to get my Unichrome Pro integrated graphics going.

Tom's hardware Guide did a recent preview on an AMD64 with the latest AM2 chipsets and DDR2 and it was rather disappointing. Seems like you can get better performance for now with plain old DDR and the Nforce4 chipsets... Here's the article:
http://www.tomshardware.com/2006/02/21/a_look_at_amds_socket_am2_platform/

So I have the following hardware and with a vanilla KnoppMyth install, I would get to 100% CPU without xvmc. I did get the Unichrome graphics going enough to get xv but not DRI so no xvmc.

So I decided to try it out with a 64 bit system.

Hardware: (note not top end, CPU/mobo is from a Fry's special)

ECS P4M800PRO-M LGA775 mobo with Unichrome Pro graphics (VT800)
LGA 775 P4 511 2.8Mhz EM64T
1 Gig DDR2 533Mhz
1 4 Gig harddrive (enough for / & swap)

I stuck a Gigabyte heatpipe,heatsink,fan thingie on it (very quiet and effective, if a pain in the neck to install due to incorrect instructions and poor graphics) and a fairly nondescript case. The 4Gig drive was lying about (don't laugh it's enough for the / system & swap and I can NFS when I need more storage for source trees).

So I went out and got the debian installer for AMD64 (which is correct for EM64T--seems AMD was there first and their name has stuck).

I chose a base system.
After it was done and I could boot to a prompt I logged in, set up apt and installed xorg-xserver to get xorg, svn, cvs... After that I got unichrome going again (this is a multistep process). And then installed mythtv's prereqs, and then used svn to get the 19+fixes branch of mythtv, mythplugins, myththemes and configured,compiled up and installed them.
Now I had mythtv. Then I put fluxbox on it and I was ready to try it out.
I followed the instructions in the mythtv docs to create the new mysql database etc. Set up mysql.txt,ran mythtv-setup and I was ready to go.

Results?
I saw a 20% improvement. I can now play 1080i at around 80% tops if I am running a mythbackend (with 1 air2pc card) and frontend and 70% tops if I am running only a mythfrontend (and remember to kill mythbackend) to another mythtv box (it is surprisingly easy to switch back and forth by altering mysql.txt. A tip I found is to use different themes for the frontends so I can tell which backend I'm talking too).

So anyway, I'm suspicious about my numbers because 20% seems like too much gain. And I suspect I'm comparing apples to kumquats here. One key difference is I'm using plain ALSA under 64 bits but I was using ALSA's OSS emulation before. So I installed alsa-oss and did a test using it. Sure enough I lost about 3-5% of my gain by using the oss emulation. So that explains some of it (and alsa-oss has been summarily zapped from this system I might add).

Another difference is this thing is like a Yugo with a Volvo engine in it. I've spent all my time on the guts but have paid absolutely no attention to the styling. So I have to login to a text console, do startx, use flux to open a shell, type mythfrontend (I do at least have mythbackend running at boot), and then I have myth. So it's possible there's some gain in not having any prettiness, but I doubt it would be significant.

I'm not a benchmarker, so I have no real idea how to find where the improvements are (plus my partner and dogs are starting to growl over th e hours and hours I've spent with the new toy).

Anyway, that's how I spent my weekend.

Amy

my guess would be that you are getting a good gain from being on xorg with newer unicrome drivers as well. even so if it's truly a 10% decrease in CPU utilization that is significant particularly when it's from 100% to 90%.

_________________
Have a question search the forum and have a look at the KnoppMythWiki.

Xsecrets

I guess it's a misnomer to say I was on plain vanilla before too. I was on KnoppMyth R5B4 but I did install xorg-xserver (using apt which involved a removal of the original X on the system) because I thought to use the openchrome drivers I had to. So I was originally running Xorg when I did my tests on KM.

I have to say that I have a whole new appreciation for KnoppMyth after going through this process.

I would also say that anyone wanting to try out a 64 bit environment would do well to try out the 64 bit Ubuntu. In my case I had a small harddrive and therefore wanted to create a minimalist build.