Go Back   nV News Forums > Linux Support Forums > NVIDIA Linux

Newegg Daily Deals

Reply
 
Thread Tools
Old 08-17-05, 08:59 PM   #1
JaXXoN
Registered User
 
Join Date: Jul 2005
Location: Munich
Posts: 910
Default Reproducable freeze at 2D/3D clock switching with driver version 7676

Hi all!

I have a Gainward Ultra/3500PCX (7800GTX) in an Asus A8N-SLI
Premium and i discovered a procedure that can easily freeze the
computer and which might be related to the "screen frozen, but
mouse pointer moves" bug as discussed at:
http://www.nvnews.net/vbulletin/showthread.php?t=31858


1. run fvwm95 with the following parameters in ~/.fvwm95rc:

DeskTopSize 3x2
EdgeScroll 100 100
EdgeResistance 0 0

This basically means that you will have six virtual desktops
(three columns, two rows) and you can switch from one desktop
to the next one by simple moving the mouse beyond the edge
of the current visible desktop.

2. run glxgears and place it "between" the upper left and
the lower left virtual desktops, so that the upper half of
the glxgears window is rendered on the upper left virtual
desktop and the lower half of the glxgears window is rendered
in the lower left virtual desktop.

3. move the mouse downward, so that you will leave the upper
left virtual desktop and enter the lower left virtual desktop.


Now two things can happen, depending on the 2D/3D clock settings:

When 2D/3D clock settings are the same (450+40MHz in this case,
+40 is the "Geometric Delta Clock" of the 7800GTX), then you will
see that the mouse jumps from the lower edge of the screen to the
upper edge as expected when entering the lower left virtual desktop
from that upper left one, but you will still see the content of the
upper left virtual desktop (with the upper half of the glxgears window)
for a second or two (the time varies). There is no activity going on
in that time (i.e. no rotating glxgears), but the mouse pointer can
be moved. After that one or two seconds, the content of the lower
left virtual dektop appears (with the lower half of the glxgears window)
and the system works normal again (i.e. rotating glxgears).
Let's call it a "hickup".

With different 2D/3D clock settings (150+0MHz and 450+40MHz),
the system freezes after that one to two seconds (after moving
the mouse downward).

GPU voltage has been set to 1.4V volt for both 2D and 3D mode.

The problem does not show up when moving the window from
one to the next virtual desktop. It also doesn't show up when
glxgears is started on both virtual desktops in parallel (one
glxgears window on each virtual desktop).


AFAIK, when switching virtual desktops, fvwm95 makes all windows
of the previous virtual desktop "invisible" and the other windows
related to the next virtual desktop visible. I could imagine that
this will lead to a very short 3D-2D-3D switch causing the hickup
or freeze. Maybe the PLL doesn't ramp up/down fast enough?

So the interim solution to at least avoid a freeze is to make
2D/3D clocks (and voltages) identical:

http://www.nvnews.net/vbulletin/show...hlight=nibitor

regards

Bernhard
JaXXoN is offline   Reply With Quote
Old 08-17-05, 09:02 PM   #2
JaXXoN
Registered User
 
Join Date: Jul 2005
Location: Munich
Posts: 910
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Please find attached the .fvwm95rc file i'm using. BTW.: the window manager
is fvwm95-2.0.43f compiled from source on a Fedora Core 4.

regards

Bernhard
Attached Files
File Type: zip fvwm95rc.zip (13.1 KB, 290 views)
JaXXoN is offline   Reply With Quote
Old 08-19-05, 11:38 AM   #3
chunkey
#!/?*
 
Join Date: Oct 2004
Posts: 662
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Code:
NVRM: Xid: 25,  L1 -> L0
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fc28fb80 00000002
NVRM: Xid: 25,  L0 -> L0
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fc28fb00 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fc000080 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 0000fb80 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 0028fb00 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fc000080 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 0000fb80 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 0028fb00 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fc280080 00000002
NVRM: Xid: 25,  L1 -> L0
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fe08fb80 00000002
NVRM: Xid: 25,  L0 -> L0
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fe08fb00 00000002
NVRM: Xid: 13, 0002 beff000c 0000009f 00000300 fe080080 00000002
Ahh, a relative to the 0xbeef bug! Beef
Seems like that atleast all > NV43 Cards have this problem. and noone knows why!
chunkey is offline   Reply With Quote
Old 08-19-05, 09:41 PM   #4
JaXXoN
Registered User
 
Join Date: Jul 2005
Location: Munich
Posts: 910
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Quote:
Originally Posted by chunkey
Ahh, a relative to the 0xbeef bug!
Silly me, didn't checked /var/log/message. I get pretty similar output:

NVRM: Xid: 25, L1 -> L0
NVRM: Xid: 13, 0001 beff000c 0000009f 00000300 fbc0f300 00000002
NVRM: Xid: 25, L0 -> L0
NVRM: Xid: 13, 0001 beff000c 0000009f 00000300 fb50f300 00000002
NVRM: Xid: 13, 0001 beff000c 0000009f 00000300 0070f300 00000002
NVRM: Xid: 13, 0001 beff000c 0000009f 00000300 fbc0f300 00000002
NVRM: Xid: 13, 0001 beff000c 0000009f 00000300 fb50f300 00000002
NVRM: Xid: 13, 0001 beff000c 0000009f 00000300 0070f300 00000002

Thanks for the feedback!

regards

Bernhard
JaXXoN is offline   Reply With Quote
Old 08-23-05, 12:29 PM   #5
chunkey
#!/?*
 
Join Date: Oct 2004
Posts: 662
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Jaxxon, since you've a 7800 Card... can you get:


Get NvCrash-Cone
(~13 Mb)

and look if it also _works_ for you?
(Be ready to reboot!)
Thanks!

Last edited by chunkey; 08-23-05 at 09:10 PM. Reason: aber, jetzt passt's!
chunkey is offline   Reply With Quote
Old 08-23-05, 09:43 PM   #6
JaXXoN
Registered User
 
Join Date: Jul 2005
Location: Munich
Posts: 910
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Quote:
Originally Posted by chunkey
Get NvCrash-Cone and look if it also _works_ for you?
I have downloaded Cone3-nvcrash-x86 and very annoyingly, it freezes
the machine (ping or ssh doesn't work). However, the computer doesn't
seem to be fully dead, because i still get messages on the serial console
(kernel boot parameter "console=ttyS0,115200n8").

After an initial "beef" message right after the crash, i get "Head"/"Channel"
messages every few seconds (with "Count" increasing):

Code:
NVRM: Xid: 25,  L1 -> L0
NVRM: Xid: 6, PE0002 0184 beef0201 001d0568 00000000 01a10113
NVRM: Xid: 25,  L0 -> L0
NVRM: Xid: 6, PE0002 0184 beef0201 00000000 00000000 00000000
NVRM: Xid: 6, PE0002 0184 beef0201 00000000 00000000 00000000
NVRM: Xid: 25,  L1 -> L0
NVRM: Xid: 12, COCOD 00000002 beef7201 00000072 00000214 00000000
NVRM: Xid: 6, PE0002 0804 00003300 00000030 00000000 011f012b
NVRM: Xid: 16, Head 00000000 Count 00000db4
NVRM: Xid: 25,  L1 -> L0
NVRM: Xid: 8, Channel 00000002
NVRM: Xid: 16, Head 00000000 Count 00000db5
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 16, Head 00000000 Count 00000db6
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 16, Head 00000000 Count 00000db7
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 16, Head 00000000 Count 00000db8
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 16, Head 00000000 Count 00000db9
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 16, Head 00000000 Count 00000dba
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 8, Channel 00000020
NVRM: Xid: 16, Head 00000000 Count 00000dbb
NVRM: Xid: 16, Head 00000000 Count 00000dbc
NVRM: Xid: 16, Head 00000000 Count 00000dbd
NVRM: Xid: 16, Head 00000000 Count 00000dbe
NVRM: Xid: 16, Head 00000000 Count 00000dbf
NVRM: Xid: 16, Head 00000000 Count 00000dc0
The machine actually was pretty stable after clocking down the
Golden Sample to 450 MHz (470 MHz factory default) and setting
2D/3D clocks and voltages to the same values. UT2004 runs like
a charm for many hours and there are no problems with the desktop
(except the hickups described above). But now, i know that this
doesn't mean anything :-(

Could it be that the driver loops in the interrupt service routine?
This would explain why there are still messages on the console, but
no more other activities: the console output is done in polling mode.

I once had a Geforce 5950 card that always sucked up 70% of the
CPU time in the interrupt service routine (checked that by hacking
the "open source" part of the nvidia driver). So the system worked,
but was slow like hell - i certainly returned the card.

Thanks for the hint. I guess i will fill in a bug report to nvidia.
Seems like quite a number of people are having that problem - hopefully
nvidia doesn't leave us alone.

regards

Bernhard
JaXXoN is offline   Reply With Quote
Old 08-23-05, 09:47 PM   #7
JaXXoN
Registered User
 
Join Date: Jul 2005
Location: Munich
Posts: 910
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

BTW.: just checked with my backup system (Athlon64 3500+, nforce4, 6800GT)
and it crashes pretty much the same way.

regards

Bernhard
JaXXoN is offline   Reply With Quote
Old 08-24-05, 05:38 AM   #8
chunkey
#!/?*
 
Join Date: Oct 2004
Posts: 662
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Thanks for your feedback...

Nvidia has already confirmed that _this_ is a (well) _known_ bug, (aka beef-bug)
Code:
NVRM: Xid: 25,  L1 -> L0
NVRM: Xid: 6, PE0002 0184 beef0201 001d0568 00000000 01a10113
chunkey is offline   Reply With Quote

Old 08-24-05, 08:16 AM   #9
JaXXoN
Registered User
 
Join Date: Jul 2005
Location: Munich
Posts: 910
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Quote:
Originally Posted by chunkey
Nvidia has already confirmed that _this_ is a (well) _known_ bug
Thanks for the info! Is there anything i (we) could do to assist nvidia fixing the bug?

regards

Bernhard
JaXXoN is offline   Reply With Quote
Old 08-24-05, 10:10 AM   #10
chunkey
#!/?*
 
Join Date: Oct 2004
Posts: 662
Default Re: Reproducable freeze at 2D/3D clock switching with driver version 7676

Quote:
Originally Posted by JaXXoN
Thanks for the info! Is there anything i (we) could do to assist nvidia fixing the bug?
Nvidia never replied any of my mails about _this_ issue...
chunkey is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


Similar Threads
Thread Thread Starter Forum Replies Last Post
nvidia + 2.4.19 gentoo = messed up ulukay NVIDIA Linux 18 11-01-02 11:58 PM
X Failed to load NVdriver c00lr4c3r NVIDIA Linux 13 10-22-02 01:44 PM
Error messages... HELP!!!! Imperito NVIDIA Linux 3 09-24-02 10:46 PM
GForce drivers installed...but dont work tomfullery NVIDIA Linux 6 09-22-02 08:23 AM
Need help to get the X to work on my Acer TravelMate 630 knchee NVIDIA Linux 16 09-19-02 10:16 PM

All times are GMT -5. The time now is 05:45 AM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.