Last version is faster!

timedemo increases from 9.8 to 13.2 FPS
on X1000 + HD 7970, full details, 1920x1080

Thx HunoPPC!

@LiveForIt

yes, I know. I receive this XE like DMA-fixed. Thanks to walkero I received Eyetech original DMA-fix manual.
Them my friend makes DMA-unfix.

I tested the PCI 66 MHz, graphics cards works here ootb with no problems, speed is the same like AGP card in PCImode or AGPx1 mode.
Interrupts works. So everything OK with slots.

But not with drivers.
Other cards ( tried SATA, USB2.0, 1GbE ) act strangely. I see them in Ranger, they receive correct interrups, interrupts correctly mapped in IntHandler - I will post screenshots.

But partially works only 1GbE - like I wrote above.
1GbE also have no errors in debuglog, USB and SATA yes, something like "cannot read"...

SATA drivers normally initialized, in early Amiga bootmemu ( pressed L+R mouse) SATA disk in PCI 66 normally visible.
But after boot the disk is not mounted, HDtoolbox see that there is the device, but cannot go inside to partitions. Sometimes even cannot read correct diskname.

I tested also change interrupt level in PCI66 from "level" to "edge" ( PCI66 have same interrupt line, pin nr like PCI33 slot 2 ), but no change. I also think, that if the level was the reason, all system will be stucked ( because on unsolved interrupt ). But it in no case. System works, only cards not.

It seems that my problem is not connected directly with interrupts.

When I place PCI card ( for example 1GbE NIC ) in 66 MHz slot it receive correct interrupt line, pin and nr. It emerges correctly in IntHandler.

But cards in 66 MHz slot ( except gfx one ) works like not works.
With NIC I can ping ( receive DNS translation, ping works ). But nothing more complex. If I try Odyssey - it starts first sockets but no advance. It is not stuck, simply there no transfer.

Please, do somebody utilize PCI card in 66 MHz slot? ( except gfx, of course )

@Evillord68

thank you for info. I am thinking about registration there - in short time it is second case which is interested me. But my german is good for ordering the wine and beer only. And google translates from Czech language is very far from ideal.

About 7448. Don't know.
If somebody used 7447 in AmigaOne XE, there is big chance it will works. It is nearly identical CPU. Other very similar microarchitectures are MPC7445, MPC7441 all developed from 7450.
And AmigaOne has 7451 which also derives from 7450 but other way. So I wonder...

@LiveForIt

Sharing irq nr is not optimal, but is possible with PCI bus.
One from my working configurations:

AGP card
PCI66 -
slot 1 SATA
slot 2 USB 2.0
slot 3 ethernet 1GbE

with this config AGP card and ethernet have the same:
Line=0B(= 11dec), PIN=A, int nr.=27

they differs in priority on interrupt handler, where gfx have 100 and ethernet 10 or 5.

As I know, older ISA cannot share interrups and also MSI (newer PCI interrupt system) not share interrupts. But I thing that was for speed oprimalization, not for need.

yes it is a puzzle.
As I know, Interrupt controller is legacy PCI ( pre - MSI ), have only four IRQ hw pins and have APIC ( two data registers ). I don't understand APIC.

But mapping according Ranger is this:

PCI bus 0 - i.e. 3x 33MHz:
PCI slot 1 = LINE 9
PCI slot 2 = LINE 10
PCI slot 3 = LINE 11

PciBus 1 - i.e. AGP + PCI 66MHz
PCI 66 slot = LINE 10
AGP slot = LINE 11

This concers if PCI card have inside one device. If card have more devices inside ( my USB2.0 card have three PCI ids), the numbers is shifted:
PCI slot 1 = second device LINE 10, third device LINE 11
PCI slot 2 = second device LINE 11, third device LINE 7
PCI slot 3 = second device LINE 7, third device LINE 9

It looks similar like U-boot menu.

I ( theoretically ) know, how to modify interrupt LINE for single PCI id - with pci write U-boot command, but how to with menu? It is a question...


I need this becouse I made experiments with PCIe cards on XE. Unfortunatelly PCIe bridge must be on 33 MHz bus, so I want to utilize PCI66 MHZ slot for other cards. And there is interrupt problem.

Please, do somebody know, what exactly do settings in U-Boot, section PCI/AGP, Menu PCI Interrupt?

There are four PCI interrupts A, B, C, D - it probably means PCI slots interrupts pins.
I checked the southbridge specs ( there is interrupt controller ) and he have only four hw interrupt lines, so it is also can be interrupt controller input.

But if I changed for example PCI interrupt A level from 10 to 13 (#A to #D), whole interrupts mappings ( line, pin, number ) in Ranger stays the same. Also mappings in InterruptHandler are not changed.

@Evillord68

you are right. It is small car with powerfull engine. But:

HDD will be faster, all network tasks will be faster ( with 1GbE card ), games will be faster - especially those that heavily use CPU - like Tower57.

And especially upgrading and optimizing old computers is really great fun!!

P.S. and if you will have such great AmigaOne XE, I recommend also change graphics card to Radeon 9000PRO. It is fastest and additionally sometimes capable of overclocking.

That tests was for graphics experience. If I compared systems I use this:

RAM: stream
VRAM + 2D: gfxbench2D
CPU: Dhrystone1 ( no FPU, no VMX ), MFlops ( FPU ), lame ( all with VMX )
HDD: ScsiSpeed, DiskSpeed ( only average and max from Read, long alligned buffers)
GFX drivers:Cow3D, MPlayer, SDLBench
NET+HTML: TCPSpeed, wirple.com/bmark
Games: GLQuake, ioQuake3, Tower57, 1941, WingsBattlefield - all this use diferrent gfx drivers and libraries. Than also RTCW.
Emu: E-UAE ADoom, DOSBox Quake


for this I have results from my AmigaOS machines

Hi,

very nice idea! I am interested in CPU modding. It will be fine, if you also post some photos and experience.

And regarding the gfx tests, I made this with Sam440ep-flex graphics cards:
https://www.amigaportal.cz/node/149415
it is in Czech language, so google helps.

Btw I have bad experience with ragemem Videobus benchmark. It gives weird results, gfxbench2D is much better here.

@Spectre660 @LiveForIt

I agree, that X1000 cannot boot from USB completely.
But can partially, with this workaround:

1. stage execution of amigaboot.of from CFE:
CFE see USB, so amigaboot.of can be booted from USB normally - tested by me. Only has to be on CFE readable partition (FFS, ext2, fat, iso? )

2. stage amigaboot.of looking for kernel modules:
this is the weak point, amigaboot.of not see USB devices, only SATA devices from 1st port, then IDE and maybe PCI SATAs.
The workaround is to have for example CF-IDE adapter with CF card in IDE controller. With small FFS partition, and on this partition is /Kickstart directory - i.e. loader and kernel modules.
Of course, it can be also on separate SATA partition instead of CF card.
amigaboot.of find kernel modules location, execute loader from there and loader loads all modules
This stage is not tested, but it works for me similar on Pegasos machines, where firmware cannot read SFS2 partitions.

3. stage selection of booting AmigaOS device
after loading of all modules ( incl. USB and other controller drivers ) is again scanned all devices for booting AmigaOS System disk with the highest priority. This disk should be anywhere ( drivers loaded ), not only on 1st SATA.
So here we can boot from booting USB prepared according to Spectre660 manual.
This is also not tested, but it should work. It is firmware independent, and it works for Spectre660 on other machine.

All what we can do is to synchronize /Kickstart on CFcard with /Kickstart on System: disk after every kickstart change.
Note: amigaboot.cf can be on the same partition like /Kickstart, if this partition is visible by both OF and amigaboot.of (- i.e. probably on 1st SATA only)

So it look like mess, but it is easy. I use it often on old machines to boot from firmware unsupported SATA cards.
It all depends on what Kas1e wants to use it for. If he wants easy-modifiyng portable System: disk with no frequent /Kickstart changes it will works fine. If he need USB System: bootable from every X1000 machine, it will not works easy way.

https://wiki.amigaos.net/w/index.php?t ... 15#AmigaOS_boot_procedure

There is AmigaOS booting procedure and manual from Spectre660 how to make bootable USB stick. It can be tried with X1000.

@Mlehto

Peg2 can boot .iso image, but MorphOS only, with parameter bootimage. AmigaOS not as I know.

Bootable USB stisk works on Sam440/460, I am not sure on X1000. It have to be tested.
And if it not works, we can make a workaround with small partitition (with /Kickstart only) on SATA disk. It depends, if amigaboot.of see USB devices or not.

@kas1e
Quote:

No, as I know. Steps are:
- put amigaboot.of to USB, connected to first port
- enter OpenFirmware
- type: boot fs=ext2 usbdisk0:amigaboot.of

My USB has ext2, normally you have fs=fatfs here. I have sometimes read problems when I accessed USB from different systems, so I change it to ext2.

This also assumes that System: disk is on device which is visible by amigaboot.of. From https://wiki.amigaos.net/wiki/AmigaOne_X1000

"Specifically, the SATA CD-ROM and SATA drive connected to Port1, SATA2 and IDE devices are scanned for partitions which are set as AmigaOS bootable, using the FFS2 or SFS file systems and containing an AmigaOS system."

I have one more manual:

Setting up dual Graphics Cards on the AmigaONE X1000:
Version 1.2 June 2012 Hans de Ruiter/Trevor Dickinson
AmigaONE X1000 Dual Graphic Card Installation Guide v1.2 ©2012

Booting of AmigaOS from USB is also possible, tested.

sorry, not yet.
garden, vacations, DMA-unfix AmigaOne XE, RadeonHD v5 X1000... ))

and X1000 + Polaris, Efika and iMac G5 bus overclocking still waits...

But I will for sure.

I think string.gadget is OK, problem with newlib.library

Just boot your Amiga with DVD ( or with other system, if you have installed two systems on disk ), find backuped AmiUpdate files ( Sys:System/Amiupdate/RallBack/2022-last-date ), read newlib_lib.restore and copy backuped files ( newlib.library.kmod, libc.so )to their right location.
Reboot from your system disk.

see my story:
https://www.amigans.net/modules/newbb/ ... id=133867#forumpost133867

I have problems on X1000 with new newlib from Amiupdate:

- libc.so v 53.80 + newlib.library.kmod v 53.80 crashed - i.e. breaks after loading cca 90% of kernel modules ( info from botton red stripe, have not serial debug )

- if I revert to older libc.so v.53.61 + newlib.libraty.kmod v 53.62 all works OK.

do you have any ideas?

Which 1080p video do you use? And do you have Prefs/Power set to High power?