Amithlon Setup Guide

[shaf]

a note extra bb patches, do no install the bb2 shell update "Shellupdate.lha" it corrupts the kernel.

[slartibartfass]

Hello everybody, long time no see.

@huepper: flatpanel=1 didnt change anything - doesnt matter much.

my issues playing h264 videos still the same.
tried:
overclocking ram from 400 to 433
change from gf4ti to gf fx5900
FastCopy from Amithlon CD since Ramithlon didnt work
but still the same

Yesterday i tried another mobo: ASRock AM2NF3-VSTA with AM2 Socket and Athlon64 4200 Dual-Core CPU and DDR2 Memory.
This is sooo much slower than K8N Neo2 + Athlon64 4000 + ddr1 Memory!
Does amithlon only use one core??

I have no more Ideas. Perhaps I have to recode all my h264 videos ...

Has anyone of you guys tips or hints for me??
Thank you!

Slartibartfass

[SnkBitten]

When you switch video cards, did you attach the FX5900 in Picasso96mode to allow the hardware acceleration modes to be presented in ScreenMode?

If you don't, you are just using VESA "built-in" modes. Make sure Picasso96mode has the 'nvidia-1' board attached.

can you provide a small sample video and what software you are using for playback so others can test on their systems?

[slartibartfass]

The gf4ti and fx5900 are both in p96mode and fully HW accelerated.

For video playback I use ffplay m68k v0.91 from aminet.

All the videos are HD recordings from onlinetvrecorder.com.If you dont use OTR, we have to find other ways ...

[slartibartfass]

... now im even more confused.
in screenmode I selected builtin just to compare and everything is the same as nvidia-1. is this vesa??
just a moment - ill change graphics card to gf6 (unsupported>vesa?) right now. to be continued ...

... okay tested gf6, then set my gf4ti in VESA mode (VESA-1 in screenmode).
scrolling through shell and directorys is much slower then, drawing of workbench windows too. But video - always the same. Vesa or not. seems that ffplay doesnt use hw acceleration.

[slartibartfass]

... tried skipframe option in ffplay - a/v sync is better but video is badly artifacted.
... tried to decrease screen resolution to 1024 - same as at bigger resolutions

@all: Can anyone play onlinetvrecorder HD videos on amithlon???

[SnkBitten]

Built-in and Vesa are the same, only a named device in picasso96mode like nvidia-1 or one of the other brands would be accelerated. Vesa-1 in p96mode is not accelerated.

I haven't done any video playback test or use yet so not sure if the player needs to be aware of hw acceleration.

[shaf]

getting an nVidia 6800gs AGP Card with 256M Ram working from the 310 Kernel I needed to deduct the Video Ram from total Memory the mem=768M switch didn't work but I got Small working using the mem=704M switch.

[croupier]

Managed to make a bootable USB key with Amithlon installation files (as on Amithlon CD)....but have a question regarding smallird.gz and bigird.gz files....anyway to update/modify them in order to add programs (like Poseidon) to the partition 'amithlon_hd'?

When booting Amithlon on machines without a CD unit you are left only with the pre-installed partition and no way to install WB3.9 as USB ports are inaccessible from Amithlon until Poseidon is up and running....

BTW...with my USB stick and Kernel4 I managed to boot Amithlon on my ASUS EeePC and Toshiba NB100......now I just need to find a way how to install WB3.9 and updates....to start testing what else works on these two small laptops......(would like to avoid the procedure of setting up the HD on other CD equipped machines and then mount the HD inside my 2 mentioned laptops)...

[SnkBitten]

You can try this..

Can you run hdtoolbox and see your drive that you want to install OS 3.9? Create it, format it and copy amithlon_hd to it. It's a stripped down 3.9 (no BoingBags installed) but it may allow you install Poseidon. Or....perhaps do the above and copy the .iso to another partition and mount the .iso in AmigaOS and install over "your" amithlon_hd partition.

You can extract the amithlon_hd .img and mount it in WinUAE and then try to add Poseidon and copy it back in to the bigird.gz. Though I haven't tried booting WinUAE off that hd image. I briefly played around with mounting it in WinUAE and then formatting it and installing OS 3.9 and patching..but it's not large enough to hold the full OS3.9.

[croupier]

Thanks!

How do you extract the amithlon_hd image of the partition from bigird?

And how do you move Poseidon files and/or .iso image to another partition (on a system without a CD on an IDE port)? Amithlon initially recognizes (AFAIK) only the amithlon_hd partition from bigird.gz and the CD (if present)...not counting the Amiga partitions created in FreeDOS that are anyway mty/formated only...

[SnkBitten]

Sorry, wasn't thinking fully. I was copying the files to the CD image to create the .iso and forgetting that you don't see all that, you just see the amithlon_hd with no way to access the full .iso (mounted, CD or USB) without the files that are on it.

Use linux, follow the guide I put up for modifying the amithlon kickstart and replace smallird.gz with bigird.gz where needed.

If you follow the guide, the amithlon_hd is in /aim/newdisk/uaefiles and is named "diskfile_haage". Add that as a hardfile to WinUAE (or whatever UAE you use) and try modifying it and copying it back. Maybe you could try just adding the Poseidon and ArakAttack files to it and saving it out.

[croupier]

Thanks!

Will try it out...

[SnkBitten]

There is very little free space available. You could possibly clean out a few unnecessary files to free a little more space.

I haven't tried copying it to a slightly larger hdf and pushing that into the bigird.gz.

@SnkBitten - I have a hypothesis about the 1GB RAM issue. I suspect the kernel is 32 bit non-PAE. Hypothetically merging all the changes into a PAE linux kernel should support booting Amithlon on systems with more physical hardware in RAM, do you think?

[ptyerman]

Originally Posted by esc

@SnkBitten - I have a hypothesis about the 1GB RAM issue. I suspect the kernel is 32 bit non-PAE. Hypothetically merging all the changes into a PAE linux kernel should support booting Amithlon on systems with more physical hardware in RAM, do you think?

Anything is possible I suppose, but 32 bit non-PAE kernels usually support up-to 3GB of RAM.

[shaf]

A few extra notes for the GA-K8N939SC Board. Onboard Audio works with Kernel 3.10 but not Onboard Lan. With Kernel4 (IDE) Onboard Audio Doesn't Work but Onboard Lan Does.
I will attempt to recompile the Kernel4 to fix the audio problem.

[ptyerman]

It's great that work still goes on to support more Lan chips. I never bother with the onboard Lan myself though, I just bung a Realtek 8139 card in there, they are compatible with just about anything and available dirt cheap on eBay and other places. Only useful if you have a free PCI slot though.

[SnkBitten]

Originally Posted by esc

@SnkBitten - I have a hypothesis about the 1GB RAM issue. I suspect the kernel is 32 bit non-PAE. Hypothetically merging all the changes into a PAE linux kernel should support booting Amithlon on systems with more physical hardware in RAM, do you think?

Here are some of Bernd Meyer's notes on the Kernel and supporting only 1 GB.

I am thinking about it. However, short of making Amithlon a kernel thread rather than a user space program (which sounds like a spectacularly bad idea, because then all sorts of APIs no longer work), there isn't much I can do about it.

The linux kernel by default is limited to 1GB physical RAM, giving user space processes 3GB of virtual address space to play with. That 1GB limit can be changed to 2GB. However, doing so reduces the user space address space to 2GB.... And in Amithlon, I need to cram the following into that address space:

a) *all* of physical memory
b) 1.5GB of PCI space (some PCI BIOSes are really really crap)
c) several megabyte of executable and its data and stack

That ain't gonna work with a 2/2 split; It hardly fits in a 1/3 split.

So at least for the foreseeable future (or until the advent of the AMD 64 bit CPUs... no promises!), 1GB is the max you will see on Amithlon.

Bernie

So if basing off a 64 bit kernel would allow greater "system" memory but still providing 1 GB "amithlon" memory is possible.....it may be. I tried compiling initially on CentOS 3 x86_64 but was getting errors so jumped to CentOS i386.

[Methanoid]

I'm really impressed with how you've taken to Amithlon and seeing what can be done with it... Keep it up if you can!