 |
|
| My results and observations |
|
|
By trasnam - Posted on July 22, 2004 - 23:56:55 (#13512)
Current version when comment was posted: 0.01 |
|
|
Thus far, I have got 2 out of 3 video cards working using the fakedriver V2 and the AGP busmanager.
The first working device is a Sapphire Radeon 7000 with 64 MB DDR RAM. It has a PCI-ID of 5159. This is on a Shuttle SK41G (FX41 mainboard). The last three hex digits of AGP cmd readback are 304.
The second working device is an Asus V7100 Pro Geforce2 MX card with 32 MB RAM. It has a PCI-ID of 0110. The mainboard is an ECS K7S5A. The last three digits of AGP cmd readback are 302. In this case, I did not use the fakedriver, as it is an Nvidia card.
Now, my third card works sometimes, but not always, regardless of the settings I choose in agp.settings. It is an IGP 340 on a Compaq 2170US laptop. It has a PCI-ID of 4337.
Here are all observed scenarios:
1.) The solid background color would come up after the boot icons, then a freeze would occur before going any further.
2.) Bootup would appear to be complete, but the system went into a hard lockup right after boot. Also, the touchpad's on/off switch shut off and couldn't be turned back on (due to the hard lock, I assume).
3.) The system booted successfully and was usable.
Now, to try to gather further information, I used a serial debug cable and tried to determine what went wrong. The hangs I see appear to occur during USB initialization, as far as I can tell.
Here is the tail end of the syslog whenever a freeze occurs:
OHCI: std_ops():
OHCI: MODULE_INIT
OHCI: std_ops():
get_module(busses/usb/ohci/v2): wake failed
module -- wake_module(busses/usb/ohci/v2): wake_count is now 2
OHCI: ohci_init():
OHCI: new_hcd_info():
OHCI: hc = 0x01adfe00
OHCI: hcca_pa = 028f5e00
OHCI: hc_registers_base = 0x30a02000, intr_line = 0x0a
OHCI: new_roothub_info(0x01adfe00), rh = 0x01adfdd0
OHCI: RHDA 03000204, RHDB 00000000, RH status 00000000
OHCI: cb_anchor_ed.own_pa = 028f6300
OHCI: intr_anchor_ed.own_pa = 028f6330
OHCI: control_reg = 00000197
OHCI: SMM driver is active
So, this leaves me to believe that the problem is actually the fault of Zeta RC2's USB implementation. I will try to put on RC3 tonight and see if that fixes my problem or not. I may also try to install BeOS Max tonight, as I couldn't get R5 Pro to install at all (it couldn't find my laptop keyboard or mouse).
For reference, on the IGP card, I managed to get settings anywhere from 304 to 101 for the last three digits of AGP cmd readback, depending on how I edited agp.settings.
Oh yeah, a note for everyone with ANY kind of Radeon card. According to all of the research I've done, enabling Fast Writes on a Radeon card has ZERO effect at all. The only thing it does is add a possible instability to the system as a whole. As such, it should always be disabled in agp.settings for stability reasons (set block_fw true).
Alright, that's all I have for now. I'll write back if/when I learn anything else.
|
|
| serpentor |
|
|
By rudolfc - Posted on July 22, 2004 - 21:20:31 (#13511)
Current version when comment was posted: 0.01 |
|
|
Hi,
In B/W mode? That would mean Thomas his driver doesn't recognize your card I think. Have you mailed with him and told him your cardID from the Devices prefs app?
====
BTW People, it's funny I think that I got one single complaint about a nVidia card not working correctly, while with ATI there is definately more trouble...
Seems like FW is 'dangerous' indeed there: that is, some cards won't work with it. One user told me that enabling the FW option in his system BIOS was called something like 'AGP Fastwrite override' which sounds real nasty to me.
That's not how it's supposed to be called: it should just be an option that 'offers' the FW option to the busmanager and the busmanager then decides to use it or not depending on what both the hostbridge and graphicscard can do (and of course what the settings file says).
The BIOS here gives me the impression that it does something like: hey, whatever the programs want: we just ignore it and hardwire it to 'ON'...
Rudolf.
|
|
| IGP 320 still does not work here. |
|
|
By Serpentor - Posted on July 22, 2004 - 20:14:44 (#13510)
Current version when comment was posted: 0.01 |
|
|
Not sure what to think. I read all these comments and excitedly installed the AGP bus manager and AGP bus driver but it did not work. My machine still comes up in black and white mode. I built a win98 boot CD so I could boot into Win98 then boot to BeOS and that did not work either. Unless someone has some other ideas, I guess I am out of luck for now. FYI, it is an e-machines M2350 laptop that I am trying to run BeOS on.
Serpentor
|
|
| ATI IGP FIX fake_driver |
|
|
By rudolfc - Posted on July 22, 2004 - 15:51:16 (#13506)
Current version when comment was posted: 0.01 |
|
|
Hi,
I just removed the V1 fakedriver, and added a V2 fakedriver below.
ATI IGP users that did not have a working system with Thomas his ATI driver and the module/fake_driver before, are strongly encouraged to try the V2 fake_driver!
Please let me know if it helps!!
(The V1 driver aborted as soon as it saw that the first found AGP device was already setup for AGP: assuming that all other devices would be in the same state. As we now know ATI fails to do that, so the V2 driver checks ALL devices before aborting now, making sure ALL devices get enabled.)
Rudolf.
|
|
| IGP |
|
|
By rudolfc - Posted on July 22, 2004 - 14:18:31 (#13505)
Current version when comment was posted: 0.01 |
|
|
Hi,
Yes, is loaded nicely in VESA mode. No FW however.
Anyway, indeed wait for Thomas: it seems that IGP does not adhere to the official AGP standard. So while I program correctly 'officially' with this fake_driver (and busmanager), some ATI cards remain having trouble.
It's up to Thomas to filter out the exact troublemaking bits and then we should get it to work nicely.
This can be done from within a driver itself (as Thomas apparantly does (now)), and this can be done with help of this AGP busmanager. Thomas has to make sure that official legal, but non-working combination of settings is blocked somehow.
Personally I'm curious what exactly ATI decided to do wrong...
Rudolf.
|
|
| Re:Ilama |
|
|
By SD - Posted on July 22, 2004 - 14:00:02 (#13504)
Current version when comment was posted: 0.01 |
|
|
Thomas made recently new Radeon driver which starts IGP natively. Wait a bit, it will published very soon, i hope. (Currently in use in my laptop, wroks well without external AGP-initialization)
|
|
| Radeon IGP 340 |
|
|
By llama - Posted on July 22, 2004 - 13:57:00 (#13503)
Current version when comment was posted: 0.01 |
|
|
Hi there,
Some people have reported success with the IGP chipset using the radeon driver and the fake AGP driver with the AGP bus driver. This is not working for me however :-( Can anyone see anything wrong with the output of syslog?
KERN 'app_server'[18]: KERNEL: driver agp.driver does not have wake_driver and suspend_driver
KERN 'app_server'[18]: agp_drv: init_hardware called
KERN 'app_server'[18]: agp_man: bus module: init
KERN 'app_server'[18]: agp_man: found 2 AGP capable device(s)
KERN 'app_server'[18]: get_module(bus_managers/agp/v0): wake failed
KERN 'app_server'[18]: agp_drv: init_driver called
KERN 'app_server'[18]: agp_man: get_nth_agp_info called with index 0
KERN 'app_server'[18]: agp_drv: enabling AGP
KERN 'app_server'[18]: agp_man: enable_agp called
KERN 'app_server'[18]: agp_drv: AGP cmd readback: $0f000304.
KERN 'app_server'[18]: agp_drv: uninit_driver called
This is booting in VESA mode, if I boot with the radeon driver the computer crashes before it writes syslog.
Does it look like the AGP stuff is loading okay given the above? All help appreciated!
Cheers,
llama
|
|
| Radeon 0x1002/5144 - no crash, no "accelleration" |
|
|
By SD - Posted on July 22, 2004 - 12:38:51 (#13502)
Current version when comment was posted: 0.01 |
|
|
some ancient VIA M/B, old nonumber 32 MD DDR Radeon
No crash with all options enabled, no difference in GL-Teapot and Quake2 timedemo.
agp_man: found 2 AGP capable device(s)
agp_drv: init_driver called
agp_man: get_nth_agp_info called with index 0
agp_drv: enabling AGP
agp_man: enable_agp called
agp_drv: AGP cmd readback: $1f000302.
agp_drv: uninit_driver called
|
|
| Kraton.. |
|
|
By rudolfc - Posted on July 22, 2004 - 11:56:34 (#13500)
Current version when comment was posted: 0.01 |
|
|
Hi,
Thanks for the update. OK, so you have trouble getting AGP4x +FW to work. You have one option left to test, and this is (forcing max AGP speed 2x!):
max_speed 2
block_agp false
block_sba false
block_fw false
If that doesn't work, you can set:
block_sba true
and try once more.
If none of these two settings work, FW indeed refuses to let you use your system at all, and you should keep that disabled. All other options don't give any trouble on your system apparantly.
-----
If AGP2x plus FW works on your system, GLteapot should indeed speedup over there. Aprox. values would probably be:
125fps or so in AGP2x + FW,
140fps or so in AGP4x + FW.
Without FW enabled you won't get any speed increase at the time, and you can disable AGP alltogether for now.
Hope this helps!
Rudolf.
|
|
| AGP false and FW false wont work here ATI8500(64MBSDR/DDR) |
|
|
By kraton. - Posted on July 22, 2004 - 10:05:20 (#13499)
Current version when comment was posted: 0.01 |
|
|
Yes only this combinations are working here:
false,true,true
and
false,false,true
and
true,false,false
and
true,true,false
and this one dont work:
false,true,false
and
false,false,false
That means FW only works in PCI Fallbackmode or?
I cannot use 4-AGP and FW together.(Blank Monitor).
I just try to test this feature. Teapot demo runns here
at 100fps with 1200x800 and 32Bit. So if FW works it must be much more isnt it?
|
|
| Voodoo3 2000 |
|
|
By rudolfc - Posted on July 22, 2004 - 09:37:53 (#13498)
Current version when comment was posted: 0.01 |
|
|
Hi,
Thanks for reporting! Cool that it's OK also there :-)
Your card supports spec AGP1.0,
It's set to it's max capability, being
-AGP2x speed,
-SBA enabled,
-8 outstanding AGP requests can be handled.
Unfortunately no FW support, so your system won't have a graphics speedup currently.
It's nice to now have this info on your card, I am 'collecting' this stuff you could say ;-)
Rudolf.
|
|
| VIA KTE133 + VT82C686B + Voodoo 3 2000 |
|
|
By SD - Posted on July 22, 2004 - 00:33:10 (#13494)
Current version when comment was posted: 0.01 |
|
|
Vendor: 3dfx 0x121a CardID:5
M/B chipset VIA KTE133 + VT82C686B
no crash with all options enabled.
3dfx_v4: init_hardware - Enter - PCI 3dfx-driver: Oct 24 2000 17:42:54KERN 'app_server'[19]: kprof_load_image called way too earlyKERN 'app_server'[19]: KERNEL: driver agp.driver does not have wake_driver and suspend_driverKERN 'app_server'[19]: agp_drv: init_hardware calledKERN 'app_server'[19]: kprof_load_image called way too earlyKERN 'app_server'[19]: agp_man: bus module: initKERN 'app_server'[19]: agp_man: found 2 AGP capable device(s)KERN 'app_server'[19]: get_module(bus_managers/agp/v0): wake failedKERN 'app_server'[19]: agp_drv: init_driver calledKERN 'app_server'[19]: agp_man: get_nth_agp_info called with index 0KERN 'app_server'[19]: agp_drv: enabling AGPKERN 'app_server'[19]: agp_man: enable_agp calledKERN 'app_server'[19]: agp_drv: AGP cmd readback: $07000302.KERN 'app_server'[19]: agp_drv: uninit_driver calledKERN 'app_server'[19]
|
|
| to Das Jott |
|
|
By rudolfc - Posted on July 21, 2004 - 19:31:26 (#13488)
Current version when comment was posted: 0.01 |
|
|
Hi,
You don't have to be dumb not to understand this stuff right away! It's quite confusing I think too :)
Anyway, I tested G100-G550 over here and I'm sorry to report that FW is not supported on them. This means that at this time Matrox users will _not_ profit from the busmanager. (unless their system didn't boot at all before, so the kind of trouble that this busmanager fixed for some Radeon graphics users (IGP).
Still, I am interested in feedback on cards that don't supportthe FW feature: as I plan to make better use of AGP in the future, by updating the driver to let it's engine use 'real' AGP transfers. Can take a while though before I get the hang of it ;-)
Rudolf.
|
|
| kraton |
|
|
By rudolfc - Posted on July 21, 2004 - 19:26:51 (#13487)
Current version when comment was posted: 0.01 |
|
|
Hi,
Sorry, but I still not quite understand what your problem is? Do you mean to tell me it only works with the setting you did, and not without?
Anyway, did you try to remove the block_fw thing and see if it works then? (so set block_fw false and don't touch the rest..)
The FW option is the _only_ thing that at the moment gives a speed boost. There's actually no real need to install the AGP busmanager if your hardware doesn't support FW (both hostbridge and card). If you don't set 'block_fw true' and the busmanager/fake driver still don't report FW active, one or both of your devices don't support that feature.
You can recognize it by seeing a '1' instead of a '0' in the second digit from the right in the AGP cmd readback line.
Thanks for your trouble ;-)
Best regards,
Rudolf.
|
|
| What about Matrox? |
|
|
By Das Jott - Posted on July 21, 2004 - 18:12:58 (#13485)
Current version when comment was posted: 0.01 |
|
|
I have a Matrox G550, that uses your mga-driver V0.16.
What do I have to do, to get the advantages of that agp-thing? (Yes, I do not know much about that...)
How do I recognize, that it is working?
Please explain it for a dumb user ;-)
Btw: Thanks for your great work!
|
|
| syslog ATI 8500(64MB) |
|
|
By kraton. - Posted on July 21, 2004 - 14:30:49 (#13483)
Current version when comment was posted: 0.01 |
|
|
KERN 'app_server'[17]: KERNEL: driver agp.driver does not have wake_driver and suspend_driver
KERN 'app_server'[17]: agp_drv: init_hardware called
KERN 'app_server'[17]: kprof_load_image called way too early
KERN 'app_server'[17]: agp_man: bus module: init
KERN 'app_server'[17]: agp_man: found 2 AGP capable device(s)
KERN 'app_server'[17]: get_module(bus_managers/agp/v0): wake failed
KERN 'app_server'[17]: agp_drv: init_driver called
KERN 'app_server'[17]: agp_man: get_nth_agp_info called with index 0
KERN 'app_server'[17]: agp_drv: enabling AGP
KERN 'app_server'[17]: agp_man: enable_agp called
KERN 'app_server'[17]: agp_man: max AGP 4x speed allowed (agp.settings)
KERN 'app_server'[17]: agp_man: blocking FW (agp.settings)
KERN 'app_server'[17]: agp_drv: AGP cmd readback: $1f000304.
KERN 'app_server'[17]: agp_drv: uninit_driver called
KERN 'app_server'[17]: kprof_load_image called way too early
KERN 'app_server'[17]: KERNEL: driver gx00.driver does not have wake_driver and suspend_driver
KERN 'app_server'[17]: kprof_load_image called way too early
KERN 'app_server'[17]: KERNEL: driver mga.driver does not have wake_driver and suspend_driver
KERN 'app_server'[17]: kprof_load_image called way too early
KERN 'app_server'[17]: KERNEL: driver mobility.driver does not have wake_driver and suspend_driver
KERN 'app_server'[17]: AKD init_hardware(): checking pci index 0, device 0x8086/0x1a30
|
|
| ATI8500 with agp and fake driver |
|
|
By kraton. - Posted on July 21, 2004 - 11:21:50 (#13481)
Current version when comment was posted: 0.01 |
|
|
Sorry,
It is working here with:
-4 times AGP
-false
-true
-true
|
|
| quake2 benchmarking |
|
|
By rudolfc - Posted on July 21, 2004 - 10:15:18 (#13480)
Current version when comment was posted: 0.01 |
|
|
Hi,
Try this:
->start quake2
->hit the '~' button so you get the games terminal 'window'
->type: timedemo 1
->type: map demo1.dm2
Now wait until you are returned to the games terminal window: it will display the FPS reached.
If you want back in normal mode, type:
->timedemo 0
->hit '~' once again.
Rudolf.
|
|
| syslog output |
|
|
By rudolfc - Posted on July 21, 2004 - 10:12:39 (#13479)
Current version when comment was posted: 0.01 |
|
|
Hi,
In order to get the stuff logging to syslog you need to have the kernel settings file 'installed' (copied from example up to settings folder).
Remove the # character on the line that says syslog output is true (or something like that).
If you reboot now, you should be able to find the agp messages in there. If not, the fake driver is not correctly installed so both this fakedriver _and_ the busmanager won't load and log ;-)
Rudolf
|
|
| to Lt_Henry |
|
|
By BeAR - Posted on July 21, 2004 - 09:43:33 (#13478)
Current version when comment was posted: 0.01 |
|
|
Try to disable FW in agp.settings I have the same situation with my Radeon 9000 Pro. I think this card does not support FW, even in Windows I can't enable it.
|
|
| Kt-400 not working |
|
|
By Lt_Henry - Posted on July 21, 2004 - 07:43:33 (#13476)
Current version when comment was posted: 0.01 |
|
|
I cant find agp_drv or agp_man on var/syslog, and the fake-driver hangs at booting (last icon). My board is a QDI Kudoz 7a, VIA KT400 and the graphics card a radeon 9000. if you need more information just tell me.
|
|
| ATI Radeon 8500(QL) working! |
|
|
By kraton. - Posted on July 20, 2004 - 23:32:32 (#13472)
Current version when comment was posted: 0.01 |
|
|
If i change the settingsfile all to true and the AGP speed to 4 my card is working now. See down:::
Anyway i dont know to test it now. I have downloaded Quake II and Demofiles but i dont know how to use the console in there to show me the framerates...
|
|
| FW |
|
|
By rudolfc - Posted on July 20, 2004 - 23:04:41 (#13471)
Current version when comment was posted: 0.01 |
|
|
Hi,
No, that's not possible. The busmanager asks every device what it supports and by default enables the highest common mode. You can force the busmanager to do less, but not more: that would probably crash your system.
If the BIOS supports an AGP option to be enabled or disabled, it in fact 'modifies' the hostbridge hardware to adhere to what the BIOS says, making that option available (or not) to the busmanager to be used.
The same applies for graphicscards sometimes. For instance on nVidia NV11 you can make the card support FW. In fact, the nVidia driver supports letting you do that, which will allow the busmanager to activate FW: but ONLY if also the hostbridge support that feature.
Look at it this way:
If I speak French, and you (listener) only understands English, our 'communication' would fail ;-)
Rudolf.
|
|
| Fast write in BIOS? |
|
|
By Tek_No - Posted on July 20, 2004 - 22:59:23 (#13470)
Current version when comment was posted: 0.01 |
|
|
Hi Rudolf,
sorry but I can't find this option in my Thinkpad's bios... Is there a way to try to force this in the driver???
Thx
Tek_No
|
|
| FW support? |
|
|
By rudolfc - Posted on July 20, 2004 - 19:15:40 (#13462)
Current version when comment was posted: 0.01 |
|
|
Hi,
Glad it's working OK.
Your system is set to AGP V2 4x, with SBA enabled, but no FW (fastwrites). Did you check if your system BIOS has an option to enable that? If so, maybe you could enable it and see if FW gets enabled then ;-)
Rudolf.
|
|
| Works fine on my Thinkpad A30 |
|
|
By Tek_No - Posted on July 20, 2004 - 15:44:36 (#13450)
Current version when comment was posted: 0.01 |
|
|
Works fine with fake agp driver:
KERN 'app_server'[18]: agp_drv: enabling AGP
KERN 'app_server'[18]: agp_man: enable_agp called
KERN 'app_server'[18]: agp_drv: AGP cmd readback: $1f000304.
KERN 'app_server'[18]: agp_drv: uninit_driver called
KERN 'app_server'[18]: kprof_load_image called way too early
KERN 'app_server'[18]: KERNEL: driver radeon2 does not have wake_driver and suspend_driver
KERN 'app_server'[18]: Radeon - init_hardware: Version: 4.1.0.1
KERN 'app_server'[18]: Radeon - Radeon_FindRom: Signature: P6
KERN Last message repeated 1 times
KERN 'app_server'[18]: Radeon - Radeon_FindRom: Signature: RADEON
KERN 'app_server'[18]: Radeon - Radeon_FindRom: found ROM @0xc0000
KERN 'app_server'[18]: Radeon - Radeon_CardDetect: found supported device pci index 2, device 0x1002/0x4c59
KERN 'app_server'[18]: Radeon - init_driver:
KERN 'app_server'[18]: Radeon - Radeon_FindRom: Signature: P6
KERN Last message repeated 1 times
KERN 'app_server'[18]: Radeon - Radeon_FindRom: Signature: RADEON
KERN 'app_server'[18]: Radeon - Radeon_FindRom: found ROM @0xc0000
KERN 'app_server'[18]: Radeon - Radeon_MapDevice: device: 010000
KERN 'app_server'[18]: Radeon - Radeon_MapDevice: old PCI command state: 0x00000387
KERN 'app_server'[18]: Radeon - Radeon_MapDevice: physical address of memory-mapped I/O: 0xc0100000-0xc010ffff
KERN 'app_server'[18]: Radeon - Radeon_GetPLLInfo: ref_clk=2700, ref_div=60, xclk=16600, min_freq=12000, max_freq=35000 from BIOS
KERN 'app_server'[18]: Radeon - Radeon_GetBIOSDFPInfo: Panel ID string: Samsung LTN150P1-L02
KERN 'app_server'[18]: Radeon - Radeon_GetBIOSDFPInfo: Panel Size from BIOS: 1400x1050
KERN 'app_server'[18]: Radeon - Radeon_GetFPData: h_disp=1400, h_blank=288, h_over_plus=40, h_sync_width=112
KERN 'app_server'[18]: Radeon - Radeon_GetFPData: v_disp=1050, v_blank=13, v_over_plus=0, v_sync_width=3
KERN 'app_server'[18]: Radeon - Radeon_GetFPData: pixel_clock=108000
KERN 'app_server'[18]: Radeon - Radeon_DetectRAM: 16 MB DDR SGRAM found
KERN 'app_server'[18]: Radeon - Radeon_UnmapDevice:
KERN 'app_server'[18]: Radeon - probeDevice: found Radeon Mobility; ASIC: m6
KERN 'app_server'[18]: Radeon - probeDevice: making /dev/graphics/1002_4c59_010000
KERN 'app_server'[18]: Radeon - Radeon_ProbeDevices: 1 supported devices
Thx!
Tek_No
|
|
| hidden/2 :) |
|
|
By rudolfc - Posted on July 20, 2004 - 13:12:26 (#13441)
Current version when comment was posted: 0.01 |
|
|
Hope should hav read: Nope :-/
Sorry about that... typing too fast I guess ;-)
Bye!
Rudolf.
|
|
| hidden :) |
|
|
By rudolfc - Posted on July 20, 2004 - 13:11:38 (#13440)
Current version when comment was posted: 0.01 |
|
|
Hope SD,
They are hidden for users. The busmanager is a kernel module, and only accessible to kerneldrivers and such (so I'm told), and the fakedriver is a graphics kerneldriver that does not export any names. It's in /dev/graphics though ;-)
Rudolf.
|
|
| /dev ? |
|
|
By SD - Posted on July 20, 2004 - 13:01:01 (#13437)
Current version when comment was posted: 0.01 |
|
|
I tried to look if you see those two components in /dev/ tree but didn't find entries neither under /bus nor under /graphics.
Should those be exposed?
|
|
| more general info.. |
|
|
By rudolfc - Posted on July 20, 2004 - 12:52:45 (#13434)
Current version when comment was posted: 0.01 |
|
|
Hi,
First off, thank you all for testing and reporting! I must say I am very surprised to see that even hanging situations are fixed with ATI cards using Thomas' driver!
Cool :-)
OK, I have some 'updates' on the info I just crossposted:
===============
->Everyone BUT nVidia graphicscard users NEED to install the fakedriver. Untill their official drivers directly program the bus via the busmanager that is.. Keep an eye out for possible updates for that on your graphicsdriver's webpage.
->If you enable syslog output, the fakedriver and busmanager log a bit of stuff in there. You can at least see if both are working: agp-man is the busmanager talking, agp-drv is the fake driver.
Here's a snippet from my syslog over here:
KERN 'app_server'[25]: agp_drv: enabling AGP
KERN 'app_server'[25]: agp_man: enable_agp called
KERN 'app_server'[25]: agp_drv: AGP cmd readback: $1f000302.
This value tells me what mode has been set BTW:
if bit set:
b3 clear:
b0 - agp pre V3 1x mode
b1 - agp pre V3 2x mode
b2 - agp pre V3 4x mode
b3 set:
b0 - agp V3 4x mode
b1 - agp V3 8x mode
general:
b4 - FW enabled(!)
b8 - AGP transfers enabled
b9 - SBA enabled.
The rest is less interesting ;-)
=======
If you have trouble: tweak the agp.settings file to get it working!
You should AT LEAST be able to fallback to PCI mode (block_agp true).
If that doesn't work your graphicsdriver is likely messing up stuff... (it may NEVER touch AGP command register directly, only via the busmanager (doing both the host bridge AND graphicscard in THIS specific order.
----------
Your preferred fallbacks if automatic (default) does not work for you:
->block SBA; still trouble:
->try max speed AGP 4x, trouble:
->try max speed AGP 2x, trouble:
->try AGP1x, trouble:
->use PCI mode.
Another technical note:
->AGP FW ALSO works if AGP transfers are NOT enabled: as long as the speed scheme is setup correctly. Otherwise enabled FW WILL hang your system!
As this 'option' (FW only) is not crystal clear from the AGP specs I block this in the busmanager code. This info is just interesting for driver writers.
=======
OK, even more info:
->Your system BIOS might control if your motherboard tells the AGP busmanager it supports certain features.
Make SURE you check it out for AGP FW (fastwrites) set to enabled, and the highest possible max. AGP speed!
(SBA might be in there as well.)
Note: this is CONFIRMED info.
---
Well, happy tweaking! Keep me posted please, I need the feedback to ascertain what options are reasonalbe failsafe to have enabled by default on a next version of the busmanager... ;-)
Best regards,
Rudolf.
|
|
| nv crosspost: AGP, speed, card, hostbridges.. |
|
|
By rudolfc - Posted on July 20, 2004 - 12:37:26 (#13433)
Current version when comment was posted: 0.01 |
|
|
Hi there,
I've got a few pointers maybe that clears stuff up a bit.
About the fake driver: you MAY install it, but it's NOT needed. The nVidia driver takes care of it for you, but it doesn't mind a bit if the fake driver does it first.
The nVidia driver just 'adds' a few 'features':
->it dumps extensive info to logfile if logging is enabled;
->it enables NV11 users to get FW support which the fake driver CANNOT do! (still it doesn't matter if you have the fake driver installed, the nVidia driver just reprograms the AGP bus after the fake driver is done ;-)
About AGP:
->AGP has different 'features', but you can divide it in two major parts I think: FW (fastwrites), and AGP cycles.
==========
1. Fastwrites:
This option gives you a nice speed increase for unaccelerated graphics. It just speeds up existing comms between the system and graphicscard by using the AGP-FW style comms instead of standard PCI comms. For this to work, you only need to have a motherboard that has a AGP2.0 or higher slot WITH FW support, and a AGP2.0 or higher graphics card WITH FW support. The speedup happens 'automatically' as soon as you installed just the AGP busmanager and the latest nVidia driver.
Well, you can all just lookup the docs of your hardware and see which AGP spec your stuff supports, but the FW feature is mostly not mentioned in there. Some pointers I do have:
nVidia graphicscards:
->NV15 (GeForce2Ti) and higher probably support AGP2.0 and FW,
->NV11 (GeForce2MX) even support it unofficially: you need to 'tweak' the nv.settings file for 'unhide_fw' to get the card to support this feature (over here it works!)
->TNT1 and TNT2(M-64) over here don't support FW, although they are AGP2.0 compliant (I think). I don't know if different versions of those cards exist (die revisions) that might support it after all, but I doubt it.
Motherboard hostbridges:
->The Intel BX chipset is AGP1.0, so does max. AGP2x speed and does NOT support FW: no gain here.
->I have a Asus motherboard with Intel chipset for P4/533FSB which is AGP2.0 and does support FW: definate speedup there!
->Someone tested a VIA hostbridge for me (vendorID $1106, deviceID $0691) which does support FW. Unfortunately VIA chipsets are notorious for messing up with AGP with certain settings. At this users system the screen got corrupted up and the system hanged with the default settings (using AGP4x, SBA and FW automatically selected). After tuning the busmanager's settings file to do max. AGP2x speed it
worked OK (FW still enabled!), speed increase to about 125% (AGP4x does about 140%).
------------
2. AGP cycles:
->While this 'feature' (style of comms) is enabled now, it will only be used if the graphicsdriver is updated to issue AGP style commands to the system from the acceleration engine. This is targeted (only used for) hardware 3D acceleration! This has not yet been setup, so this isn't used yet now.
HINT:
Now you know why AGP trouble on windows mostly occurs when playing 3D games, and not when using other 'uses' of the system... apart from FW that is, which is also used for those other uses: just as it is now in BeOS.
(AFAIK)
SBA BTW is an extra feature only used for AGP cycles, and so isn't used yet neither, though the 'feature' is enabled.
==============
Tuning the busmanager:
If your system hangs with the default settings, you should try to tune the busmanager. If you want speed increase, do not block FW use. You should try to disable SBA first, and if that does not help, decrease max. AGP speed. Try 4x, then 2x. 1x won't speedup graphics even with FW enabled BTW.
PLEASE report to me what setting worked and what not! I will modify the default settings for the busmanager later on to be stable on all (most) systems. But I can only do that if you report back!
-----
Getting info about what happens:
If you want to know exactly what happens, just turn on logging for the nVidia driver. It will dump all found devices, their power-up settings, their newly programmed settings, and their ID's to file.
It's needles to say, that IF you report to me about what works and what not (if you encounter trouble you need to fix via settings), I would LOVE to see those logfiles!!!
--------
The AGP busmanager AND the fake driver also log a bit of info: for you to see this you need to enable syslog output debugging, and then you can find the messages in the syslog file (search for 'agp': agp-drv is the fake driver, agp-man is the manager logging).
==========
visible speed increase effects:
->everything that's not accelerated by the 2D engine will speedup.
So openGL (if the CPU load is not the limiting factor that is!), video playback (bitmap bigtime speedup, overlay also speeds up because overlay bitmaps STILL need to be transferred over the AGP bus, although this is less info to 'transfer' than bitmap output), every app that draws stuff onscreen will be faster, so paint apps, net browsers, hey: you name it (quake2 software rendering mode speeds up, but not openGL as the scenes are soo complicated your CPU is the bottleneck (software openGL, MESA), not the buswidth to the graphicscard (that happens only with HW OGL ;-).
OK,
Hope this helps a bit.
Good luck!!
Rudolf.
|
|
| About fake driver... |
|
|
By cvincent - Posted on July 20, 2004 - 11:17:56 (#13429)
Current version when comment was posted: 0.01 |
|
|
Hi,
We need it or no ?
i have installed AGP module + nvidia 0.22 drivers, fake driver is for drivers not support the "fake" ? yes ? no ?
Thanks,
CKJ - Vincent Cedric
|
|
| About Motherboard etc... |
|
|
By cvincent - Posted on July 20, 2004 - 11:15:17 (#13428)
Current version when comment was posted: 0.01 |
|
|
Hi,
I have try this module, but i don't know if that work,
i have ASUS P2B-Dual with AGP but 1 or 2 i don't know
and i have GForce2 MX 400 chipset on grapics card.
do you mean that's work with this configuration ?
can you list on your website a ready hardware list ?
CKJ - Vincent Cedric
|
|
| How to check? TNT2 64MB |
|
|
By kraton. - Posted on July 20, 2004 - 09:12:34 (#13426)
Current version when comment was posted: 0.01 |
|
|
With my TNT2 64MB and AGPmanager+fake driver i see no difference to before. At least it seems working. But how to check it?.
BeOs Max +P2 400MHz TNT2 64MB AGP +440BX Board
|
|
| AGP manager Radeon 8500 fake driver Radeon driver 4.01 dont work |
|
|
By kraton. - Posted on July 20, 2004 - 09:08:55 (#13425)
Current version when comment was posted: 0.01 |
|
|
With an ATI Radeon 8500(Ql) and the AGP Bus manager + fake driver... the system seems to boot but my monitor wont awake.(black screen and blinking standby mode-lights only.)
Dano
|
|
| Nice! |
|
|
By tuishimi - Posted on July 20, 2004 - 04:30:21 (#13421)
Current version when comment was posted: 0.01 |
|
|
Maybe it is all in my head, but I really seem to see a difference with Quake 2 in software mode, especially full screen! I have an old ATI Radeon card.
Mike
|
|
| not crashed yet |
|
|
By jonas.kirilla - Posted on July 20, 2004 - 04:23:02 (#13420)
Current version when comment was posted: 0.01 |
|
|
Installed Haiku AGP 0.01, NVidia 0.22 + fake AGP device, rebooted to make sure it got loaded alright, and so far it hasn't crashed my R5.03/BONE. Been a few hours now.
|
|
| Another effect - browsers |
|
|
By SD - Posted on July 20, 2004 - 03:55:08 (#13419)
Current version when comment was posted: 0.01 |
|
|
Now animated gifs also take much less of CPU, which improves Mozilla speed
|
|
| Is it possible??? |
|
|
By SD - Posted on July 20, 2004 - 03:47:09 (#13418)
Current version when comment was posted: 0.01 |
|
|
Tried at old Radeon 32 DDR AIW. Inspite VLC works in Overlay (not bitmap!), CPU load decreased 1.5-2 times!!!.
Same result reported for ATI Rage 128 AIW AGP working with Be's driver - CPU load is lower with VLC and DivX almost 2 times
|
|
| Works great! |
|
|
By Kevin_Field - Posted on July 20, 2004 - 02:04:34 (#13411)
Current version when comment was posted: 0.01 |
|
|
...I think. :) How can I tell whether it's loaded or if it's kicked into PCI mode? I forgot to take note of my GLTeapot frame rate before I rebooted. :)
Kev
|
|
|
|
|
|
|
|
|
 |
|
|
|
|
|
|
|
|
 |
| |
Recent Downloads - # 661
Total Downloads - # 351
Total Views - # 69
User Ratings - # 5
|
|
|
|
 |
| |
Video Drivers
|
|
|
|
 |
|
1. BePodder - 9.80
2. QEMU - 9.68
3. ScummVM - 9.57
4. cpu_fix - 9.42
5. Jukebox - 9.40
6. libdl.so - 9.40
7. Haiku AGP busm... - 9.35
8. vim6 - 9.31
9. Beezer - 9.26
10. HandBrake - 9.25
|
|
|
|
 |
|
1. BeOS 5 Perso... - 13,601
2. Realtek RTL8... - 13,114
3. Ati Radeon G... - 12,583
4. Ensoniq Audio... - 7,566
5. ATI Rage 128... - 7,442
6. USB Joystick... - 5,663
7. Broadcom 440x... - 5,411
8. S3 Trio 64 v2... - 4,727
9. USB Serial dr... - 4,707
10. Intel Extreme... - 4,469
|
|
|
|
 |
| You are not logged in.
Login
|
|
|
|
 |
|
