Amithlon Setup Guide

[SnkBitten]

Voodoo cards aren't supported in Amithlon so no warp3d support either.

You can use the Wazp3d stuff though with decent results.

[SnkBitten]

I have a new module file for the following network cards (mostly motherboard), it doesn't build with the kernel so had to be manipulated a little to get it to 'make' for kernel4.

RTL8111B/RTL8168B/RTL8111/RTL8168
RTL8111C/RTL8111CP/RTL8111D(L)
RTL8168C/RTL8111DP/RTL8111E
RTL8168E/RTL8111F/RTL8411
RTL8111G/RTL8111GUS/RTL8411B(N)
RTL8118AS

It's based off this, no changes to the driver just massaged the make file.
version 1.07
http://www.realtek.com.tw/Downloads/...3&SortByDesc=1


Download it here for your Amithlon system.
http://amithlon.snkbitten.com/files/...rtl8111nic.lha

Extract the files to ram: then copy the r1000.o file to your devs: pcidrivers/net/ folder and copy the other three files to your S: directory.

It will overwrite your pci_modules, vendors.txt and vendors_pci.txt files in S: so you may want to back those up. The new ones were updated and should give more useful info as well when doing a pcilist command as well.

[SnkBitten]

Originally Posted by jdog320

Also whenever I play a sound in VBox, there's like a 1 sec delay before it happens. I can notice it and it's very annoying.

You can try this, some info from Bernie:

Just a quick note --- in the year-and-a-half since I actually wrote the
contrib3b sound drivers, processors have gotten a lot faster... Which,
strangely enough, means that the contrib3b sound drivers are actually
working worse than when they were written.

The reason is that they contain a heuristic that adjusts buffer sizes
according to processor speed (need slightly larger buffers on slower
processors, as a rule of thumb), and that given the generational leap
in processors since mid-2002, that heuristic is now coming up with
somewhat silly results (*tiny* buffers).

Fortunately, you can override those silly results. Simply set a couple
of AmigaOS environment variables:

AHIFragSize : Gives the *size* of each buffer. Minimum 32,
maximum 8192. Default is 512, and you might want
to increase it to, say, 2048, if simply changing
the number of buffers doesn't help (or doesn't
work). Probably a good idea to keep it a power of 2,
i.e. one of 32, 64, 128, 256, 512, 1024, 2048, 4096 or
8192.

AHIFragCount : Gives the *number* of such-sized buffers the driver
uses. Whenever one gets empty, a request is sent to
AHI to please refill it --- while the other buffers
get played. The more buffers, the better the chance
that request gets serviced in time... but also the
higher the time between an app sending sound data
and that data actually being played.
The default value on all reasonably modern machines
is 3; You can increase this to any value up to 32.

Use - setenv AHIFragCount 8

Now for the caveats:

* Bigger/more buffers means data spends more time in the buffers.
It's no problem if you are just listening to MP3 music; But if
you are playing, say, a fast-paced action game, you want the sound
to match the action, and not be a fraction of a second behind.

* When recording, it is better to have many small(ish) buffers
than few larger ones --- because the data can be read as soon
as any one buffer is full. When playing back, few large buffers
are preferable to many small ones, because the delay is the same
either way, and it's more efficient to do things in larger blocks.

* Not all linux drivers support all sizes and/or counts. You can
only try, and if changing one parameter stops things from working
altogether, try changing the other one (or try different values).
You *might* get some useful information from "xcat /proc/kmsg",
but don't count on it.

The default buffer sizes, as set by the heuristic on modern machines,
gives a total of 1536 bytes; At 48000 samples/s 16 bit stereo playback,
the sound card will consume 192000 bytes per second, so the buffers are
about 8ms worth of sound. That's somewhat silly. Bump that up to roughly
40ms (i.e. setting AHIFragCount to 16, *or* AHIFragSize to 2048), and you
should still have a very nice game experience. Increase it to 166ms
(setting AHIFrageCount to 16 *and* AHIFragSize to 2048), and you should
be set for perfect MP3 playback and beautiful recording.

Hope this will help some people get the most out of their machines,


Bernie

P.S.: Oh, one more caveat --- using those AmigaOS environment variables is
completely and utterly untested.

[SnkBitten]

Just for fun I thought I'd take a look at booting Amithlon as AmigaOS 3.1.

That meant setting up a 3.1 boot partition and booting to it before the normal Amithlon system boots so that I don't get any of the 3.9 Setpatch features/modules loaded. Since I've modified my smallird.gz to have a patched kickstart it also meant using the stock smallird.gz and then catching the system booting to enter the 'Early Boot Menu' to choose the partition to boot.

I have tons of unallocated space on my 256 GB SSD drive Amithlon uses and I have a 3.1 boot CF-Card for my A4000T when I want to play with Toaster 3.x. I partitioned out a 1.5 GB partition, copied the CF-Card (USB Poseidon/ArakAttack) to the new partition and set it bootable. I then modified the startup-sequence, adding the Amithlon boot pieces at the beginning. I kept commenting out lines until it finally booted.

C:Mount DEVS:DOSDrivers/~(#?.info)
C:Mount DEVS:DOSDrivers/AMithlon/~(#?.info)

were the two that would stop with errorcode 20.

So booting using the stock smallird.gz (kickstart 3.1) and catching Amithlon before it fully boots using the two mouse buttons to enter the 'Early Boot Menu' and choosing the bootable partition with 3.1 installed.



I made sure to copy all the p96 stuff from my 3.9 SYS: volume to the 3.1 volume and all the commands referenced in the 3.9 startup-sequence as well as going through C:, S:, L:, LIBS: and DEVS: and copying what I felt was required like amithlon patches, pciinsmod, pci_modules, powerpci.library, etc...

[SnkBitten]

I was donated an MSI X58M motherboard with an i7-920 @ 2.67 GHz for testing Amithlon. It's a tricky board to work with, while it supports IDE and SATA it's chipset handles them in a strange way. Using FreeDOS to setup a boot partition it would boot off the CD and then fail to find the CD to copy data from it to the HD (CD connected IDE or SATA). Same, booting Amithlon off the CD (modified to use Kernel 4) it would boot and switch over to the amithlon_hd image and fail to see the CD. I finally got irritated enough that I just copied my Amithlon system partition to a formatted drive and used it.

USB works, networking works (with the added r1000.o file posted a few above), it has one PCI slot so my nVidia FX5500 went in that.

Speed wise I'm a little under impressed. Granted it's 2.67 GHz vs. my Core2Duo at 4.0 GHz but I was hoping for a little more. SysInfo reports it between 65-70 times faster than an A4000. My Core2Duo reports in at 237 times faster than an A4000. Switching to SysSpeed the results were similar. My Core2Duo is on the left, the i7-920 is on the right/middle.



Memory test showed significantly faster results on the Core2Duo (both are running dual channel setups).

I believe a lot has to do with the 3 MB L2 cache per core on the Core2Duo and only .25 MB L2 cache per core on the i7-920. The GHz difference plays a part as well but so far I'm extremely pleased with my current setup.

[SnkBitten]

A follow up with this i7 motherboard. A friend (who donated me this one) has the same board but with an i7-950 CPU and he's getting much closer to my numbers at 3.3 GHz.

I have an i7-965Xtreme on the way so hopefully I can do a better evaluation of the i7 CPU in Amithlon.

[ptyerman]

I have a Asus P6T-SE board with the X58 chipset and can verify odd behaviour, but it's usually to do with booting from USB and not the CD. Several distro's will begin to boot then fail with a can't find xxxx message. Very annoying at times and I have found no solution to it.
Also, the X58 chipset and boards are usually triple channel memory not dual. Unless yours is one of the later iterations after Intel shelved triple channel?

[SnkBitten]

Originally Posted by ptyerman

I have a Asus P6T-SE board with the X58 chipset and can verify odd behaviour, but it's usually to do with booting from USB and not the CD. Several distro's will begin to boot then fail with a can't find xxxx message. Very annoying at times and I have found no solution to it.
Also, the X58 chipset and boards are usually triple channel memory not dual. Unless yours is one of the later iterations after Intel shelved triple channel?

Yes this one is triple channel but supports dual channel ram if only two ram modules are installed in particular slots.

[ptyerman]

You would gain a speed advantage by using triple channel, that is possibly where your slowness is coming from. Removing a full channel of memory bandwidth restricts CPU access somewhat!

[SnkBitten]

Originally Posted by ptyerman

You would gain a speed advantage by using triple channel, that is possibly where your slowness is coming from. Removing a full channel of memory bandwidth restricts CPU access somewhat!

I agree and I have three 512 MB modules (only two installed). I'd need to find some 256 MB modules which I'm not sure exists. It was hard enough finding the 512 MB ones.

EDIT*
Just popped in the third 512 MB module and Amithlon still booted. No additional ram in Amithlon but at least it should allow the system to run in triple channel memory mode.

Memory test were better, not a huge difference but scores were higher.

[ptyerman]

Never even crossed my mind about the memory limitation! But yeah, should be OK as long as Amithlon ignores it.
I've never run Amithlon on anything considered modern so have no idea what the consequences are. It's usually Pentium II and III's and AthlonXP CPU's I run it and X-Amiga on. It's just a viable way to use older hardware.

[SnkBitten]

A little update on the MSI X58M motherboard.

The third ram module created an issue with being able to pciinsmod the r1000.o module for the network driver. This required me to use the mem= setting in the kernel init string.

So mem=832 works, allowing me to have 1.5 GB on the motherboard in Triple Channel mode and a healthy amount in AmigaOS.

I'm definitely having a thermal issue, I'll visit that when I replace the CPU (once it arrives). Initial boot I can get a 100+ in SysInfo but as time goes on the scores continually drop down to 80, 70, 60, 40 and 30 in speed compared to an A4000. I don't use the system so not too worried though if it takes too long for the new CPU to arrive I'll pull the heatsink and reapply thermal paste.

IDE/SATA is extremely weird on this board. Amithlon will not see IDE devices and Amithlon doesn't support SATA optical drives (kernel 2.4 doesn't support SATA optical drives). It may require running IDE only and disabling the SATA function. I'll have to test though currently my AmigaOS is on a SATA SSD and I'm booting from an old IDE HD.

[ptyerman]

Yeah, support for the X58 chipset is really sketchy at times, specially with early Linux kernels and drivers!
You will probably find it's down to lack of drivers in the kernel, it's usually the USB that fails for me in some situations.
I do tend to run my SATA settings as IDE in the BIOS though and don't normally have problems with anything. I think I came across some problems some time ago with a Linux distro when I had the BIOS set to AHCI, can't remember fully but it did disable the drives if I remember right.

[SnkBitten]

Some small changes. mem=850M works, anything above that fails. So 1.5 GB on the motherboard, mem=850M and all is well. Using three 512MB modules to allow the motherboard to use triple channel memory access.

The chipset driving the IDE/SATA bites. No SATA devices, IDE only, booting from a modified Amithlon CD (uses kernel4) does not see any devices on amithlon.device, so no HD to install AmigaOS. IDE is really borked with this system. This means no CDRom available to Amithlon, which makes installation a much more difficult task.

My i7-965xtreme CPU should be arriving Monday so hopefully that'll happen and I'll have the better CPU installed.

[SnkBitten]

The i7-965xtreme arrived and is running much better. It's a 3.2 GHz CPU and my SysInfo scores are in the low 180s (typically 181).

Memory speeds have jumped tremendously as well, in fact every performance test under SysSpeed has increased. I swapped out the video card from the PCI 5500 256MB to a PCI 5200 128MB card, so I may try to see if I can bump up the mem= statement in the kernel init string.

Overall it's running much more like I expected it to run. I could have had a bad i7-920 or the i7-965 is just that much better.

Posting this from the i7-965xtreme Amithlon system.

EDIT**
Was able to up the memory a little with mem=896M, up from mem=850M previously. This is because all the memory of the video card is mapped into the Amiga address space (even though it'll only use 16MB), so the lesser memory on the 5200 128MB allowed more available fast ram for AmigaOS.

[malko]

[towo2099]

Sadly, Amithlon runs fine in Virtualbox nowdays, it's bad i can't get networking to work.
Doesn't matter if NAT or Bridged network. With bridged network i can Ping tha Amithlon VM from outsite, my router sees the System toom, but fom inside Amithlon i can't do any network task.

[jdog320]

Originally Posted by towo2099

Sadly, Amithlon runs fine in Virtualbox nowdays, it's bad i can't get networking to work.
Doesn't matter if NAT or Bridged network. With bridged network i can Ping tha Amithlon VM from outsite, my router sees the System toom, but fom inside Amithlon i can't do any network task.

What networking stack are you using?

[towo2099]

I'm using genesis. Yesterday i found out, if i write the nameserver to ../db/resolv.conf ping can resolve adresses, but again, 100 % packet loss. I don't know, why genesis-prefs isn't writing anything to resolv.conf.
on a bare metal dell notebook, genesis is working like it should. I have no clue, what's the problem with virtualbox. I can ping any IP but not use any browser, ftp or irc app in vbox.

[sigma63]

Originally Posted by SnkBitten

You need 1.9a of ROMSplit. Doobrey hasn't released it yet but he did give me permission to share the link he provided me. I had to ask him to include the Amithlon 'shape.rom' in ROMSplit which he gladly did.

http://www.doobreynet.co.uk/beta/ROMsplit_1.9a.lha

Hello SnkBitten,
unfortunately your link is dead. Do you have a newer or correct one?

Thanks
Sigma