Sonnet 7200 PCI programming - Page 159

grelbfarlk · 23 April 2019, 23:25

Listen to Trixster, he speaks the truth.

As far as which PPC cards it goes:
Ragnarok
Sonnet 400/1M or Killer NIC K1
Ragnarok

I don't change anything between them except for the MMU-Configuration.
The ENV Variables should be mostly the same as on the wiki:
https://github.com/Sakura-IT/SonnetA...ki/Good-Config

I use Picasso 2.1e (unofficial)
OS 3.9 BB2+a bit of BB3 and BB4

On the Sonnet/KillerNIC system I use A4000 IDE.
The other systems I use the accelerator onboard SCSI.
Swap is open, that should be good.

I can't seem to find Dnetc on Aminet, I swore I got it there.

There was something Stedy was saying that the PCI spec expects either termination or a card in the last PCI slot.

trixster · 24 April 2019, 09:50

Also, Amigas with lots of cards - zorro and pci - are fickle and sometimes the pursuit of stability defies logic. My system had a 1 month period where it required maybe 10 minutes without the ppc card running in order to ‘warm up’. If i initialised the ppc immediately and then ran anything with warp3d I would get a crash or lockup. Pure WarpOS software ran fine. Wrangler and I spent a lot of time dismantling the system to try and trouble-shoot the problem. Then inexplicably the problem went away. It has been rock solid for the last 9 or so months.

Maybe it was a power issue? Maybe noise on a line? It was unlikely a problem with the Ragnarok or the sonnet.library as WarpOS ran fine. So maybe a problem with the voodoo3? However I tried a spare voodoo3 and that exhibited the same problem. Ultimately I made no real changes to the system and it just fixed itself. So who knows?

Stedy · 25 April 2019, 01:16

Quote:

Originally Posted by grelbfarlk

Listen to Trixster, he speaks the truth.

There was something Stedy was saying that the PCI spec expects either termination or a card in the last PCI slot.

It does indeed need a card in the last slot. If there is another PCI device after the last slot, in the old days, a southbridge, you can discount this rule.

PCI is the only reflective wave switched bus (google it) and to meet the timing requirements, you need to damp the signals down using diode clamping. You can add parts to do this or use a PCI card. Apparently Dell fixed a server PC that had issues by fitting a cheap USB card.

Quote:

Originally Posted by trixster

Also, Amigas with lots of cards - zorro and pci - are fickle and sometimes the pursuit of stability defies logic. My system had a 1 month period where it required maybe 10 minutes without the ppc card running in order to ‘warm up’. If i initialised the ppc immediately and then ran anything with warp3d I would get a crash or lockup. Pure WarpOS software ran fine. Wrangler and I spent a lot of time dismantling the system to try and trouble-shoot the problem. Then inexplicably the problem went away. It has been rock solid for the last 9 or so months.

Maybe it was a power issue? Maybe noise on a line? It was unlikely a problem with the Ragnarok or the sonnet.library as WarpOS ran fine. So maybe a problem with the voodoo3? However I tried a spare voodoo3 and that exhibited the same problem. Ultimately I made no real changes to the system and it just fixed itself. So who knows?

Older PowerPCs based around the TSI106/TSI107 (and MPC106/107) need all power supplies operational within 50ms of the first supply becoming active. If you violate this, the system will behave badly as the PLL system will not correctly initialise. I'd have to check the manuals for the E300/603 CPUs as well, it's been a few years.

If you have an oscilloscope or logic analyser, you could take some measurements. Basically the +5V, +3.3V, +1.8V (bridge supply) and the CPU core all need to be operational within a 50ms window and stable for 100us+ before reset de-asserts.

Hedeon · 25 April 2019, 10:52

@Stedy:

Not completely related, but does your knowledge also extend to video cards and can you say what is wrong if a video card starts up garbled in lets say a quarter of the cases? I was thinking of a memory PLL/timing initialization error. Can it here also be though that the driver initializes it too slow?

The code seems to work correctly on PPC but not on 68k.

drluking · 25 April 2019, 16:39

I'm trying to solve the problem, but it's not easy; in any case I'm installing amigaos 3.9 from scratch, step by step, but I want to ask you for important information right away, for the order in which to insert the PCI cards in the mediator 4000D 3v with 5 PCI slots.

I have the following PCI cards available:
- two Voodoo3 PCI video cards (3000 and 2000)
- two PowerPC PCI cards (killer M1 @ 400 and K1 @ 333)
- some SB128 sound cards (CT4810)
- a PCI Fortemedia FM801-AS sound card
- two ethernet cards RTL 8100C PCI 10/100
- an RTL 8139C PCI 10/100 ethernet card
- some DFE-538TX PCI 10/100 ethernet cards
- an RTL 8029AS PCI 10 ethernet card
- a Nec USB card PCI (not Elbox flashed)
- an Adaptec AHA-2940UW SCSI card PCI
- several other PCI cards (USB, SCSI, firewire, parallel, ethernet, modem, gfx virge ..etc ..)

I should therefore terminate the PCI chain, filling all the 5 PCI slots and I would be interested to have:
- a Voodoo3 gfx, a PowerPC Killer, an ethernet card, a sound card, while the fifth card would not matter

So in what order do you recommend in which to insert the five PCI cards in the mediator 4000D 3v with 5 PCI slots ? and which cards do you recommend to insert ? (if the order of the cards and the number of cards is important to correctly terminate the PCI BUS)

thank you all for your help

trixster · 25 April 2019, 16:41

I would place the voodoo in the lowest (first) slot and the ppc card in the top (last) slot, and leave the remaining 3 slots empty for the moment.

drluking · 25 April 2019, 20:46

so Is it important to "terminate" the bus in this way (a card in the first PCI slot (PCI1) and another card in the fifth PCI slot (PCI5)) does it act as a terminator for the PCI BUS ?

Unfortunately it is not possible to use the Voodoo3 PCI or even the Killer M1 in the first PCI slot (PCI1) due to the heat sink that comes in contact with the Elbox PCI-Zorro3 card.
Do I leave the Killer M1 alone and put the Killer K1 in the first PCI slot (PCI1) and the Voodoo3 in the fifth PCI slot (PCI5) , or Do I put another card in the first PCI1 slot (exemple the USB card, or other ones ...) and the Killer M1 (or K1) in the fifth PCI slot (PCI5) ?

Should the PCI cards that terminate the PCI bus be special PCI cards ? That is, must they be bus mastering, DMA ... cards or have addressing in sharing ... ?

Stedy · 26 April 2019, 00:44

Quote:

Originally Posted by drluking

so Is it important to "terminate" the bus in this way (a card in the first PCI slot (PCI1) and another card in the fifth PCI slot (PCI5)) does it act as a terminator for the PCI BUS ?

Unfortunately it is not possible to use the Voodoo3 PCI or even the Killer M1 in the first PCI slot (PCI1) due to the heat sink that comes in contact with the Elbox PCI-Zorro3 card.
Do I leave the Killer M1 alone and put the Killer K1 in the first PCI slot (PCI1) and the Voodoo3 in the fifth PCI slot (PCI5) , or Do I put another card in the first PCI1 slot (exemple the USB card, or other ones ...) and the Killer M1 (or K1) in the fifth PCI slot (PCI5) ?

Should the PCI cards that terminate the PCI bus be special PCI cards ? That is, must they be bus mastering, DMA ... cards or have addressing in sharing ... ?

You don't need anything in the first slot, if that is adjacent to the Logic card as that provides the necessary termination. I'm looking at the install guide on the Elbox website as I write this. I guess slots 1 and 5 would share the same interrupt, so don't share a USB card with a CPU card. PIC has 4 interrupts, A,B,C, & D and each plug in card has one interrupt pin, the motherboard/backplanes deals with mapping each cards interrupt pin to the available PCI interrupts, A to slot 1, B to slot 2, C to slot 3, D to slot 4 and then A to slot 5.

Stedy · 26 April 2019, 01:05

Quote:

Originally Posted by Hedeon

@Stedy:

Not completely related, but does your knowledge also extend to video cards and can you say what is wrong if a video card starts up garbled in lets say a quarter of the cases? I was thinking of a memory PLL/timing initialization error. Can it here also be though that the driver initializes it too slow?

The code seems to work correctly on PPC but not on 68k.

Humm, interesting....

I have developed graphics cards for the embedded market. Don't know about the latest cards but Voodoo cards and Radeon 95xx are older tech, some of my knowledge is relevant.

The speed of initialisation should not matter too much. Normally you pre-load a number of registers then hit a 'go' bit or similar to initiate an init portion. This is especially true of the SDRAM on the graphics card. Normally the controller will help correctly initialise the RAM, there are minimum timing requirements for the RAM, the only maximum time would be RAM refresh related but you can be out by a factor of 2 and still work.

What do you mean by a garbled display?
Do you see any recognisable text or graphics?
Are there any horizontal or vertical lines in the display?
Does the graphics driver report any errors?
Do you know how the graphics driver identifies the card type and initialises it?
Do cards with 128MB work fine but ones with 64 or 256MB fail or vice versa?

Ooh, is it specific to Radeon or Voodoo cards?

As the 68K and PowerPC are both big endian, we can disregard any big/little endian byte swaps to the GPU. That has caused some confusion in the past

A good test, is once you've initialised the card, to perform a memory test on the card. Look for the memory mapped Based Address Register (BAR) with >1MB allocated, then read/write chunks of data to it. This is typically the framebuffer used for the display.

Sorry for so many questions but there are a lot of variables.

drluking · 26 April 2019, 01:13

Quote:

Originally Posted by Stedy

You don't need anything in the first slot, if that is adjacent to the Logic card as that provides the necessary termination.

yes, It's adiacent, correct, OK

Quote:

I'm looking at the install guide on the Elbox website as I write this. I guess slots 1 and 5 would share the same interrupt, so don't share a USB card with a CPU card.

why ? because, the CPU PPC will it be activated manually via script ? or for other reasons?

for convenience and space I could put in the first PCI slot (PCI1) a card any USB, or ethernet, or audio (from what I saw, from pciinfo, only the USB PCI card shares the interrupts, is it possible?), while in the fifth slot PCI (PCI5) I could put any PCI card.
So what should be tried to insert in PCI1 and in PCI5?

Quote:

PIC has 4 interrupts, A,B,C, & D and each plug in card has one interrupt pin, the motherboard/backplanes deals with mapping each cards interrupt pin to the available PCI interrupts, A to slot 1, B to slot 2, C to slot 3, D to slot 4 and then A to slot 5.

where is the PCI microcontroller ? in the PCI-Zorro Elbox card ?

thanks

Hedeon · 26 April 2019, 02:51

What do you mean by a garbled display?

Looks like the video ram is corrupted. Things can be made out, sometimes it is 70% garbled sometimes almost not. It is not static, dots move around.

Do you see any recognisable text or graphics?

yes

Are there any horizontal or vertical lines in the display?

no, it looks like corrupted video ram.

Does the graphics driver report any errors?

no

Do you know how the graphics driver identifies the card type and initialises it?

it reads the bios and initializes the card with the values found there (not through x86 emulation).

Do cards with 128MB work fine but ones with 64 or 256MB fail or vice versa?

All cards exibit the same behaviour. With the Elbox driver, no such corruption is seen. With this driver it happens like in 1 out of 3 boots.

Ooh, is it specific to Radeon or Voodoo cards?

Radeon

As the 68K and PowerPC are both big endian, we can disregard any big/little endian byte swaps to the GPU. That has caused some confusion in the past

I don't think this is the problem here.

A good test, is once you've initialised the card, to perform a memory test on the card. Look for the memory mapped Based Address Register (BAR) with >1MB allocated, then read/write chunks of data to it. This is typically the framebuffer used for the display.

I could make a screenshot when the screen isn't garbled badly enough so I can still see things. Let's see what that does.

grelbfarlk · 26 April 2019, 03:21

Quote:

Originally Posted by drluking

yes, It's adiacent, correct, OK

why ? because, the CPU PPC will it be activated manually via script ? or for other reasons?

for convenience and space I could put in the first PCI slot (PCI1) a card any USB, or ethernet, or audio (from what I saw, from pciinfo, only the USB PCI card shares the interrupts, is it possible?), while in the fifth slot PCI (PCI5) I could put any PCI card.
So what should be tried to insert in PCI1 and in PCI5?

where is the PCI microcontroller ? in the PCI-Zorro Elbox card ?

thanks

When the system boots, it assigns interrupts via slot so when the card is initialized or not doesn't matter.

Yes you'll end up with the USB sharing interrupts no matter what unless you only have 3 PCI devices, but it seems to work. I have Spider in slot 1 and 10MB Ethernet in Slot 5(as only the 10MB card works in OS4 with CSPPC). And I don't think the 10MB card needs PCIDMA memory to work, which I think that's why it works on OS4.

In the 4 slot Mediator the USB will end up sharing with something anyway, but keep in mind the USB is not really very useful for much, just mass storage devices like USB readers and input devices like mouse/keyboard which you can't use during early startup.

Also many non official Spider USB cards are flaky, yes they might appear to work, some will not work in in Hispeed mode, so will be limited to <1MB/s, they may crash, intermittently initialize, on and on. I have a couple with seem to work fine but I have several which do not.

I will regret saying this, as many people will probably tell me I'm wrong, but I think everyone is using the MMUlib, are you? Have you tried that?

grelbfarlk · 27 April 2019, 01:21

Here's some system details that might be of use to somebody. This is for the five slot Ragnarok system:
PCIINFO

Code:

Pci Device 1:
   VendorID: $1033
   VendorName: NEC Electronics
   DeviceID: $35
   DeviceName: udp9210 Dual OHCI controllers plus Single EHCI controller
   ClassCode: $c0310
   RevisionID: $41
   InterruptPin: 1
   SubsystemVendorID: $1033
   SubsystemID: $35
   MemSpace0: $60000000 - $60000fff

Pci Device 2:
   VendorID: $1033
   VendorName: NEC Electronics
   DeviceID: $35
   DeviceName: udp9210 Dual OHCI controllers plus Single EHCI controller
   ClassCode: $c0310
   RevisionID: $41
   InterruptPin: 2
   SubsystemVendorID: $1033
   SubsystemID: $35
   MemSpace0: $60100000 - $60100fff

Pci Device 3:
   VendorID: $1033
   VendorName: NEC Electronics
   DeviceID: $e0
   DeviceName: uPD720100A USB 2.0 Host Controller
   ClassCode: $c0320
   RevisionID: $2
   InterruptPin: 3
   SubsystemVendorID: $ee4
   SubsystemID: $89e
   MemSpace0: $60200000 - $602000ff

Pci Device 4:
   VendorID: $1319
   VendorName: Forte Media
   DeviceID: $801
   DeviceName: FM0801 PCI Card MediaForte made in singapore
   ClassCode: $40100
   RevisionID: $b2
   InterruptPin: 2
   SubsystemVendorID: $1319
   SubsystemID: $1319
   IOSpace0: $40c06000 - $40c0607f

Pci Device 5:
   VendorID: $1319
   VendorName: Forte Media
   DeviceID: $802
   DeviceName: FM801-as Xwave PCI Joystick
   ClassCode: $98000
   RevisionID: $b2
   InterruptPin: 0
   SubsystemVendorID: $1319
   SubsystemID: $1319
   IOSpace0: $40c06100 - $40c0610f

Pci Device 6:
   VendorID: $121a
   VendorName: 3dfx Interactive Inc
   DeviceID: $5
   DeviceName: Voodoo3 All Voodoo3 chips, 3000
   ClassCode: $30000
   RevisionID: $1
   InterruptPin: 3
   SubsystemVendorID: $121a
   SubsystemID: $57
   MemSpace0: $62000000 - $63ffffff
   MemSpace1: $64000000 - $65ffffff
   IOSpace2: $40c06200 - $40c062ff
   RomSpace: $60300000 - $6030ffff

Pci Device 7:
   VendorID: $1057
   VendorName: Motorola
   DeviceID: $480b
   DeviceName: Unknown
   ClassCode: $60000
   RevisionID: $2
   InterruptPin: 4
   SubsystemVendorID: $0
   SubsystemID: $1057
   MemSpace0: $60400000 - $60400fff
   MemSpace1: $60500000 - $60500fff
   MemSpace2: $70000000 - $7fffffff
   MemSpace3: $60600000 - $6063ffff
   MemSpace4: $68000000 - $6fffffff

Pci Device 8:
   VendorID: $10ec
   VendorName: Realtek Semiconductor
   DeviceID: $8029
   DeviceName: RTL8029(as)pci ethernet nic windot XPse
   ClassCode: $20000
   RevisionID: $0
   InterruptPin: 1
   SubsystemVendorID: $10ec
   SubsystemID: $8029
   IOSpace0: $40c06300 - $40c0631f

MMU-Configuration

Code:

ClearTTx  
SetCacheMode from 0x07000000 size 0x08F80000 Copyback
;SetCacheMode from 0x70480000 size 0x0FA7FFFF CacheInhibit
SetCacheMode from 0x60400000 size 0x00000FFF CacheInhibit
SetCacheMode from 0x60500000 size 0x00000FFF CacheInhibit
SetCacheMode from 0x60600000 size 0x0003FFFF CacheInhibit
;SetCacheMode from 0x68480000 size 0x03A7FFFF CacheInhibit
;SetCacheMode from 0x68480000 size 0x1797FFFF CacheInhibit
SetCacheMode from 0x00000000 size 0x00200000 CacheInhibit NonSerial Imprecise
SetCacheMode from 0x64D00000 size 0x00200000 CacheInhibit
P5Init

MuScan

Code:

MuScan 46.1 (02.07.2016) © THOR

68060 MMU detected.
MMU page size is 0x1000 bytes.

Memory map:
0x00000000 - 0x001FFFFF CacheInhibit Imprecise NonSerial 
0x00200000 - 0x00BBFFFF Blank 
0x00BC0000 - 0x00BFFFFF CacheInhibit I/O space 
0x00C00000 - 0x00D7FFFF Blank 
0x00D80000 - 0x00DFFFFF CacheInhibit I/O space 
0x00E00000 - 0x00EFFFFF Blank 
0x00F00000 - 0x00F00FFF CacheInhibit 
0x00F01000 - 0x00F21FFF CacheInhibit I/O space 
0x00F22000 - 0x00F7FFFF CacheInhibit 
0x00F80000 - 0x00FFFFFF ROM CopyBack Remapped to 0x0FEE4000 
0x01000000 - 0x06FFFFFF Blank 
0x07000000 - 0x0801DFFF CopyBack 
0x0801E000 - 0x08020FFF CacheInhibit 
0x08021000 - 0x0FEE3FFF CopyBack 
0x0FEE4000 - 0x0FF63FFF ROM CopyBack 
0x0FF64000 - 0x0FF7FFFF CopyBack 
0x0FF80000 - 0x0FFFFFFF CacheInhibit 
0x10000000 - 0x3FFFFFFF Blank 
0x40000000 - 0x40FFFFFF CacheInhibit I/O space 
0x41000000 - 0x5FFFFFFF Blank 
0x60000000 - 0x7FFFFFFF CacheInhibit I/O space 
0x80000000 - 0xFFEFFFFF Blank 
0xFFF00000 - 0xFFF7FFFF CacheInhibit 
0xFFF80000 - 0xFFFFFFFF Blank

P96 Vars

Code:

AlertSupport=Yes
AmigaVideo=31kHz
AreaEndSupport=Yes
BlackSwitching=No
Debug=No
DirectColorMask=Yes
DisableAmigaBlitter=No
DoNotSetMMU=No
EnableBootVGA=No
Experimental=No
PlanarOnlyDIPFDBUF=No
PlanesToFast=No
ShowModes=All

Stedy · 28 April 2019, 14:02

@Hedeon

Can you get a screenshot or photo of the screen when the graphics driver misbehaves?

A RAM error is a strong possibility here. A photo of the screen can help identify some common RAM errors. The earlier Radeon cards don't have microcode/shaders running from the GPU framebuffer so the failure mode is a bit more subtle.

drluking · 28 April 2019, 16:31

Quote:

Originally Posted by grelbfarlk

I will regret saying this, as many people will probably tell me I'm wrong, but I think everyone is using the MMUlib, are you? Have you tried that?

I can't install MMUlib because I get this error:

"The version of SetPatch you are using is unknow to the installation script and cannot be updated, sorry. Please try to download the 43.6 edition, or buy OS 3.5 to get 44.xx editions"

I tried everything ... my setpatch 44.38 (not..), 44.16 (os3.9 not..), 44.13 (os3.9 not), 44.2 (os3.9 not and os3.5 not), 43.6 (os3.5 not), 43.6beta (aminet not) ...

I have found only this about this problem of MMULib install
https://forum.amiga.org/index.php?topic=64734.0

I'd like to try MMULib with WarpEngine040 but no way at the moment ...

Other question ... I reinstalled everything from scratch, OS3.9 + BB1 + BB2 base, Picasso96 2.1b and driver / libraries of Elbox's mediator, but I always have the same problem ... can I use the latest Sonnet powerpc.library version also in WB3.1 or is it necessary to use OS3.9 ?

Other question ...
setting VoodooMemOS=Yes variable there are problems (crash, reboot ..etc ..) trying to use video modes 16 BE (example opening a screenmode monitor amiga in 640x480 16bit or 800x600 16bit), while there is not this problem using PC LE resolutions (example opening screenmode monitor amiga in 640x480 16bit PC or 800x600 16bit PC)
(I think "640x480 16bit" is BE = big endian, while "640x480 16bit PC" is LE = little endian, is It correct ?
And is the variable "VoodooMemOS=Yes" strictly necessary to use PPC Killer M1/K1 correctly ? or not and I can use correctly PPC Killer M1/K1 with "VoodooMemOS=No" ?

Wrangler · 28 April 2019, 16:54

If you have followed the set up instructions on the wiki step by step and you're still getting crashes, you're problem is poor power distribution to the cpu IMHO. You may need to try a different ATX adapter or a 5v injector into the Zorro slot (be very careful!!!)

Stick with LE video modes as they're more compatible.

Duke Nukem should play nicely on your setup. Warcraft 2 via Stratagus might be a bit of a challenge.

Hedeon · 28 April 2019, 23:21

ANY switch from what you started in (being it LE or BE screenmode) will corrupt the memory list when VoodooMemOS=yes. This is not related to powerpc.library. This is because the voodo gfx mem holds a memory header and when switched from LE to BE or vice versa, the memory header also gets switched (corrupted). Any request for CHIP memory will give a guru as the CHIP memory header is mostly the last node in the memory list. The guru should be memory list corrupt (81000005).

Just set it to 'no'.

drluking · 29 April 2019, 01:13

Quote:

Originally Posted by Wrangler

If you have followed the set up instructions on the wiki step by step and you're still getting crashes, you're problem is poor power distribution to the cpu IMHO. You may need to try a different ATX adapter

tested with ATX 550W psu and an AT 200W psu; with AT 200W I have about +4.98V on P9 AT connector of the mediator 4000D 3v and about +4.95V on 5V pin of the core logic oscillator of mediator pci/zorro card (while about +4.92V using the ATX 550 psu), when the ppc card is working (K1 or M1) ... but I have reset/guru testing voxelspace demo (after for exemple 15 minutes)

Quote:

or a 5v injector into the Zorro slot (be very careful!!!)

how to do this?

Will be possibile to use two PSU, or Is It dangerous for the A4000 ? the AT 200W psu connected only with P9 AT to the mediator 4000D 3v and the ATX 550W psu connected to the A4000 mainboard and hd/cdrom/floppy.
If It's possibile, I should first turn on the AT 200W (to power on the mediator) and then the ATX 550W to power on the A4000,
or they should be turned on at the same time ?
I'm not sure if it is possible, also because if I remember correctly, if only the P9 AT turns on, it also partially powers the Amiga 4000 ... (+ 5V, but not + 12V)

drluking · 29 April 2019, 01:29

Quote:

Originally Posted by Hedeon

ANY switch from what you started in (being it LE or BE screenmode) will corrupt the memory list when VoodooMemOS=yes. This is not related to powerpc.library.

yes, I have seen this problem without ppc card

Quote:

This is because the voodo gfx mem holds a memory header and when switched from LE to BE or vice versa, the memory header also gets switched (corrupted). Any request for CHIP memory will give a guru as the CHIP memory header is mostly the last node in the memory list. The guru should be memory list corrupt (81000005).

Just set it to 'no'.

ok
when I run "voxelspace G" and select 320x240 8 bit, Is It LE or BE ?

About "VoodooInt=Yes" is it necessary for the correct functioning of the ppc M1 / K1 cards, or I can disable it ?

About "VoodooMem=13" is it OK, or better "Voodoo3=12" ?
considering also the variable "VoodooMemOS=No"

I have also read an interesting thing here https://github.com/Sakura-IT/SonnetA...ki/Good-Config
I have not the varabiles "Terminator" and "Force" for the powerpc.library in my ENV:Mediator/ directory ... Do I need them ?
If I need them, I will have to set Terminator=2 and Force=1 (using only sonnet powerpc.library and not ppc.library emulation) OR
I will have to set Terminator=0 and Force=1 (using sonnet powerpc.library and also ppc.library emulation) ?

grelbfarlk · 29 April 2019, 01:34

Ignore Terminator=2 and Force=1 those are from the CSPPC config. You don't need these variables.

VoodooInt=Yes YES you want this, otherwise Warp3D will be slow and crashy.

23 April 2019, 23:25	#3161
grelbfarlk Registered User Join Date: Dec 2015 Location: USA Posts: 2,958	Listen to Trixster, he speaks the truth. As far as which PPC cards it goes: Ragnarok Sonnet 400/1M or Killer NIC K1 Ragnarok I don't change anything between them except for the MMU-Configuration. The ENV Variables should be mostly the same as on the wiki: https://github.com/Sakura-IT/SonnetA...ki/Good-Config I use Picasso 2.1e (unofficial) OS 3.9 BB2+a bit of BB3 and BB4 On the Sonnet/KillerNIC system I use A4000 IDE. The other systems I use the accelerator onboard SCSI. Swap is open, that should be good. I can't seem to find Dnetc on Aminet, I swore I got it there. There was something Stedy was saying that the PCI spec expects either termination or a card in the last PCI slot. Last edited by grelbfarlk; 23 April 2019 at 23:36.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
WANTED: PCI Sound Card VORTEX2 SuperQuard / SQ2500 or Turtlebeach SantaCruz pci card	keropi	MarketPlace	12	15 March 2017 02:48
Wanted:Sonnet Technologies QuadDoubler 50 (Doubler 4000)	kronuz	MarketPlace	0	16 June 2013 00:29
eBay : Sonnet Quaddoubler	alexh	MarketPlace	3	07 May 2013 00:50
Overlay problem on GeForce 7200	Leandro Jardim	support.WinUAE	1	10 August 2010 15:50
Prometheus PCI & Voodoo 3 PCI GFX Card	Slayer	support.Hardware	21	05 September 2006 10:57

24 April 2019, 09:50	#3162
trixster Guru Meditating Join Date: Jun 2014 Location: England Posts: 2,354	Also, Amigas with lots of cards - zorro and pci - are fickle and sometimes the pursuit of stability defies logic. My system had a 1 month period where it required maybe 10 minutes without the ppc card running in order to ‘warm up’. If i initialised the ppc immediately and then ran anything with warp3d I would get a crash or lockup. Pure WarpOS software ran fine. Wrangler and I spent a lot of time dismantling the system to try and trouble-shoot the problem. Then inexplicably the problem went away. It has been rock solid for the last 9 or so months. Maybe it was a power issue? Maybe noise on a line? It was unlikely a problem with the Ragnarok or the sonnet.library as WarpOS ran fine. So maybe a problem with the voodoo3? However I tried a spare voodoo3 and that exhibited the same problem. Ultimately I made no real changes to the system and it just fixed itself. So who knows?

25 April 2019, 10:52	#3164
Hedeon Semi-Retired Join Date: Mar 2012 Location: Leiden / The Netherlands Posts: 2,039	@Stedy: Not completely related, but does your knowledge also extend to video cards and can you say what is wrong if a video card starts up garbled in lets say a quarter of the cases? I was thinking of a memory PLL/timing initialization error. Can it here also be though that the driver initializes it too slow? The code seems to work correctly on PPC but not on 68k.

25 April 2019, 16:39	#3165
drluking Registered User Join Date: Sep 2016 Location: Italy Posts: 225	I'm trying to solve the problem, but it's not easy; in any case I'm installing amigaos 3.9 from scratch, step by step, but I want to ask you for important information right away, for the order in which to insert the PCI cards in the mediator 4000D 3v with 5 PCI slots. I have the following PCI cards available: - two Voodoo3 PCI video cards (3000 and 2000) - two PowerPC PCI cards (killer M1 @ 400 and K1 @ 333) - some SB128 sound cards (CT4810) - a PCI Fortemedia FM801-AS sound card - two ethernet cards RTL 8100C PCI 10/100 - an RTL 8139C PCI 10/100 ethernet card - some DFE-538TX PCI 10/100 ethernet cards - an RTL 8029AS PCI 10 ethernet card - a Nec USB card PCI (not Elbox flashed) - an Adaptec AHA-2940UW SCSI card PCI - several other PCI cards (USB, SCSI, firewire, parallel, ethernet, modem, gfx virge ..etc ..) I should therefore terminate the PCI chain, filling all the 5 PCI slots and I would be interested to have: - a Voodoo3 gfx, a PowerPC Killer, an ethernet card, a sound card, while the fifth card would not matter So in what order do you recommend in which to insert the five PCI cards in the mediator 4000D 3v with 5 PCI slots ? and which cards do you recommend to insert ? (if the order of the cards and the number of cards is important to correctly terminate the PCI BUS) thank you all for your help

25 April 2019, 16:41	#3166
trixster Guru Meditating Join Date: Jun 2014 Location: England Posts: 2,354	I would place the voodoo in the lowest (first) slot and the ppc card in the top (last) slot, and leave the remaining 3 slots empty for the moment.

25 April 2019, 20:46	#3167
drluking Registered User Join Date: Sep 2016 Location: Italy Posts: 225	so Is it important to "terminate" the bus in this way (a card in the first PCI slot (PCI1) and another card in the fifth PCI slot (PCI5)) does it act as a terminator for the PCI BUS ? Unfortunately it is not possible to use the Voodoo3 PCI or even the Killer M1 in the first PCI slot (PCI1) due to the heat sink that comes in contact with the Elbox PCI-Zorro3 card. Do I leave the Killer M1 alone and put the Killer K1 in the first PCI slot (PCI1) and the Voodoo3 in the fifth PCI slot (PCI5) , or Do I put another card in the first PCI1 slot (exemple the USB card, or other ones ...) and the Killer M1 (or K1) in the fifth PCI slot (PCI5) ? Should the PCI cards that terminate the PCI bus be special PCI cards ? That is, must they be bus mastering, DMA ... cards or have addressing in sharing ... ?

26 April 2019, 02:51	#3171
Hedeon Semi-Retired Join Date: Mar 2012 Location: Leiden / The Netherlands Posts: 2,039	What do you mean by a garbled display? Looks like the video ram is corrupted. Things can be made out, sometimes it is 70% garbled sometimes almost not. It is not static, dots move around. Do you see any recognisable text or graphics? yes Are there any horizontal or vertical lines in the display? no, it looks like corrupted video ram. Does the graphics driver report any errors? no Do you know how the graphics driver identifies the card type and initialises it? it reads the bios and initializes the card with the values found there (not through x86 emulation). Do cards with 128MB work fine but ones with 64 or 256MB fail or vice versa? All cards exibit the same behaviour. With the Elbox driver, no such corruption is seen. With this driver it happens like in 1 out of 3 boots. Ooh, is it specific to Radeon or Voodoo cards? Radeon As the 68K and PowerPC are both big endian, we can disregard any big/little endian byte swaps to the GPU. That has caused some confusion in the past I don't think this is the problem here. A good test, is once you've initialised the card, to perform a memory test on the card. Look for the memory mapped Based Address Register (BAR) with >1MB allocated, then read/write chunks of data to it. This is typically the framebuffer used for the display. I could make a screenshot when the screen isn't garbled badly enough so I can still see things. Let's see what that does.

28 April 2019, 14:02	#3174
Stedy Registered User Join Date: Jan 2008 Location: United Kingdom Age: 46 Posts: 733	@Hedeon Can you get a screenshot or photo of the screen when the graphics driver misbehaves? A RAM error is a strong possibility here. A photo of the screen can help identify some common RAM errors. The earlier Radeon cards don't have microcode/shaders running from the GPU framebuffer so the failure mode is a bit more subtle.

28 April 2019, 16:54	#3176
Wrangler Registered User Join Date: Sep 2015 Location: London, UK Posts: 419	If you have followed the set up instructions on the wiki step by step and you're still getting crashes, you're problem is poor power distribution to the cpu IMHO. You may need to try a different ATX adapter or a 5v injector into the Zorro slot (be very careful!!!) Stick with LE video modes as they're more compatible. Duke Nukem should play nicely on your setup. Warcraft 2 via Stratagus might be a bit of a challenge.

28 April 2019, 23:21	#3177
Hedeon Semi-Retired Join Date: Mar 2012 Location: Leiden / The Netherlands Posts: 2,039	ANY switch from what you started in (being it LE or BE screenmode) will corrupt the memory list when VoodooMemOS=yes. This is not related to powerpc.library. This is because the voodo gfx mem holds a memory header and when switched from LE to BE or vice versa, the memory header also gets switched (corrupted). Any request for CHIP memory will give a guru as the CHIP memory header is mostly the last node in the memory list. The guru should be memory list corrupt (81000005). Just set it to 'no'.

29 April 2019, 01:34	#3180
grelbfarlk Registered User Join Date: Dec 2015 Location: USA Posts: 2,958	Ignore Terminator=2 and Force=1 those are from the CSPPC config. You don't need these variables. VoodooInt=Yes YES you want this, otherwise Warp3D will be slow and crashy.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)