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

Newegg Daily Deals

Reply
 
Thread Tools
Old 01-26-06, 04:10 AM   #1
d13f00l
Registered User
 
Join Date: Apr 2005
Posts: 197
Default Is there a memory leak in the 8xxx series drivers?

top
Code:
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 5779 root       0   0  327m 288m 2960 S  0.7 28.6 373:19.04 X
Ok, hm. It isn't my video card's mmaped.

Pmap of X after running 2 days.
5779: /usr/bin/X -br -nolisten tcp :0 vt7 -auth /var/run/xauth/A:0-Kkz3HS
Address Kbytes RSS Anon Locked Mode Mapping
00000000 640 - - - rwxs- [ shmid=0x0 ]
000a0000 128 - - - rwxs- mem
000c0000 192 - - - rwxs- mem
000f0000 64 - - - r-xs- mem
08048000 1752 - - - r-x-- Xorg
081fe000 200 - - - rwx-- Xorg
08230000 310196 - - - rwx-- [ anon ]
a4d9c000 256 - - - rwx-- [ anon ]
a6650000 112 - - - rwxs- [ shmid=0xd98007 ]
a676f000 384 - - - rwxs- [ shmid=0xe2000b ]
a69fb000 112 - - - rwxs- [ shmid=0xd90006 ]
a6a2f000 384 - - - rwxs- [ shmid=0xd48005 ]
a6b81000 164 - - - rwxs- nvidia0
a6baa000 112 - - - rwxs- [ shmid=0xeb000e ]
a6bcf000 384 - - - rwxs- [ shmid=0xd40004 ]
a6c46000 112 - - - rwxs- [ shmid=0xea800c ]
a6c78000 132 - - - rwxs- nvidia0
a6c99000 36 - - - rwxs- nvidia0
a6cb3000 68 - - - rwxs- nvidia0
a6cc4000 100 - - - rwxs- nvidia0
a6cdd000 260 - - - rwxs- nvidia0
a6d1e000 260 - - - rwxs- nvidia0
a6e00000 232 - - - rwx-- [ anon ]
a6e60000 36 - - - rwxs- nvidia0
a6e69000 68 - - - rwxs- nvidia0
a6e7a000 36 - - - rwxs- nvidia0
a6e83000 516 - - - rwxs- nvidia0
a6f04000 164 - - - rwxs- nvidia0
a6f2d000 8 - - - rwx-- zero
a6f2f000 4 - - - rwxs- nvidia0
a6f30000 4 - - - rwxs- nvidia0
a6f31000 64 - - - rwxs- nvidia0
a6f41000 92 - - - rwxs- [ shmid=0x98001 ]
a6f58000 4 - - - rwxs- nvidia0
a6f59000 1024 - - - rwxs- nvidia0
a7059000 3072 - - - rwxs- nvidia0
a7359000 64 - - - rwxs- nvidia0
a7369000 640 - - - rwxs- [ shmid=0x0 ]
a7409000 652 - - - rwx-- [ anon ]
a74ac000 480 - - - r-x-- libfreetype.so.6.3.8
a7524000 12 - - - rwx-- libfreetype.so.6.3.8
a752d000 16 - - - rwxs- nvidia0
a7531000 64 - - - rwxs- nvidia0
a7541000 4 - - - r-x-- libnvidia-tls.so.1.0.8178
a7542000 4 - - - rwx-- libnvidia-tls.so.1.0.8178
a7543000 7708 - - - r-x-- libGLcore.so.1.0.8178
a7cca000 204 - - - rwx-- libGLcore.so.1.0.8178
a7cfd000 16 - - - rwx-- [ anon ]
a7d01000 496 - - - r-x-- libglx.so
a7d7d000 140 - - - rwx-- libglx.so
a7da0000 168 - - - rwx-- [ anon ]
a7dca000 1092 - - - r-x-- libc-2.3.5.so
a7edb000 4 - - - r-x-- libc-2.3.5.so
a7edc000 12 - - - rwx-- libc-2.3.5.so
a7edf000 8 - - - rwx-- [ anon ]
a7ee1000 8 - - - r-x-- libXau.so.6.0
a7ee3000 4 - - - rwx-- libXau.so.6.0
a7ee4000 8 - - - r-x-- libpam_misc.so.0.78
a7ee6000 4 - - - rwx-- libpam_misc.so.0.78
a7ee7000 8 - - - r-x-- libdl.so.2
a7ee9000 8 - - - rwx-- libdl.so.2
a7eeb000 28 - - - r-x-- libpam.so.0.78
a7ef2000 4 - - - rwx-- libpam.so.0.78
a7ef3000 132 - - - r-x-- libm-2.3.5.so
a7f14000 8 - - - rwx-- libm-2.3.5.so
a7f16000 4 - - - rwx-- [ anon ]
a7f17000 60 - - - r-x-- libz.so.1.2.3
a7f26000 4 - - - rwx-- libz.so.1.2.3
a7f28000 4 - - - rwxs- nvidia0
a7f29000 4 - - - rwxs- nvidia0
a7f2a000 32 - - - rwxs- nvidia0
a7f32000 4 - - - rwxs- nvidia0
a7f33000 4 - - - rwxs- nvidia0
a7f34000 48 - - - r-x-- libfreetype.so
a7f40000 4 - - - rwx-- libfreetype.so
a7f41000 84 - - - r-x-- ld-2.3.5.so
a7f56000 4 - - - r-x-- ld-2.3.5.so
a7f57000 4 - - - rwx-- ld-2.3.5.so
afe02000 336 - - - rwx-- [ stack ]
ffffe000 4 - - - ----- [ anon ]
-------- ------- ------- ------- -------
total kB 333968 - - -


A fresh X
6241: X :1 -deferglyphs 16
Address Kbytes RSS Anon Locked Mode Mapping
00000000 640 - - - rwxs- [ shmid=0xda0009 ]
000a0000 128 - - - rwxs- mem
000c0000 192 - - - rwxs- mem
000f0000 64 - - - r-xs- mem
08048000 1752 - - - r-x-- Xorg
081fe000 200 - - - rwx-- Xorg
08230000 4384 - - - rwx-- [ anon ]
a6d12000 132 - - - rwxs- nvidia0
a6d33000 228 - - - rwx-- [ anon ]
a6d6c000 228 - - - rwx-- [ anon ]
a6db8000 196 - - - rwxs- nvidia0
a6de9000 8 - - - rwx-- zero
a6deb000 4 - - - rwxs- nvidia0
a6dec000 4 - - - rwxs- nvidia0
a6ded000 64 - - - rwxs- nvidia0
a6dfd000 1044 - - - rwx-- [ anon ]
a6f02000 92 - - - rwxs- [ shmid=0xda800a ]
a6f19000 4 - - - rwxs- nvidia0
a6f1a000 1024 - - - rwxs- nvidia0
a701a000 3072 - - - rwxs- nvidia0
a731a000 64 - - - rwxs- nvidia0
a732a000 640 - - - rwxs- [ shmid=0xda0009 ]
a73ca000 652 - - - rwx-- [ anon ]
a746d000 480 - - - r-x-- libfreetype.so.6.3.8
a74e5000 12 - - - rwx-- libfreetype.so.6.3.8
a74ee000 16 - - - rwxs- nvidia0
a74f2000 64 - - - rwxs- nvidia0
a7502000 4 - - - r-x-- libnvidia-tls.so.1.0.8178
a7503000 4 - - - rwx-- libnvidia-tls.so.1.0.8178
a7504000 7708 - - - r-x-- libGLcore.so.1.0.8178
a7c8b000 204 - - - rwx-- libGLcore.so.1.0.8178
a7cbe000 16 - - - rwx-- [ anon ]
a7cc2000 496 - - - r-x-- libglx.so
a7d3e000 140 - - - rwx-- libglx.so
a7d61000 160 - - - rwx-- [ anon ]
a7d89000 32 - - - r-x-- libnss_files-2.3.5.so
a7d91000 8 - - - rwx-- libnss_files-2.3.5.so
a7d93000 32 - - - r-x-- libnss_nis-2.3.5.so
a7d9b000 8 - - - rwx-- libnss_nis-2.3.5.so
a7d9d000 68 - - - r-x-- libnsl-2.3.5.so
a7dae000 8 - - - rwx-- libnsl-2.3.5.so
a7db0000 8 - - - rwx-- [ anon ]
a7db2000 24 - - - r-x-- libnss_compat-2.3.5.so
a7db8000 8 - - - rwx-- libnss_compat-2.3.5.so
a7dba000 8 - - - rwx-- [ anon ]
a7dbc000 1092 - - - r-x-- libc-2.3.5.so
a7ecd000 4 - - - r-x-- libc-2.3.5.so
a7ece000 12 - - - rwx-- libc-2.3.5.so
a7ed1000 8 - - - rwx-- [ anon ]
a7ed3000 8 - - - r-x-- libXau.so.6.0
a7ed5000 4 - - - rwx-- libXau.so.6.0
a7ed6000 8 - - - r-x-- libpam_misc.so.0.78
a7ed8000 4 - - - rwx-- libpam_misc.so.0.78
a7ed9000 8 - - - r-x-- libdl.so.2
a7edb000 8 - - - rwx-- libdl.so.2
a7edd000 28 - - - r-x-- libpam.so.0.78
a7ee4000 4 - - - rwx-- libpam.so.0.78
a7ee5000 132 - - - r-x-- libm-2.3.5.so
a7f06000 8 - - - rwx-- libm-2.3.5.so
a7f08000 4 - - - rwx-- [ anon ]
a7f09000 60 - - - r-x-- libz.so.1.2.3
a7f18000 4 - - - rwx-- libz.so.1.2.3
a7f1a000 4 - - - rwxs- nvidia0
a7f1b000 4 - - - rwxs- nvidia0
a7f1c000 32 - - - rwxs- nvidia0
a7f24000 4 - - - rwxs- nvidia0
a7f25000 4 - - - rwxs- nvidia0
a7f26000 48 - - - r-x-- libfreetype.so
a7f32000 4 - - - rwx-- libfreetype.so
a7f33000 84 - - - r-x-- ld-2.3.5.so
a7f48000 4 - - - r-x-- ld-2.3.5.so
a7f49000 4 - - - rwx-- ld-2.3.5.so
af933000 84 - - - rwx-- [ stack ]
ffffe000 4 - - - ----- [ anon ]
-------- ------- ------- ------- -------
total kB 26000 - - -


08230000 310196 - - - rwx-- [ anon ]

What on earth is that? They arn't pixmaps. According to an x resource monitor, only 30 megs or so or X's ram usage are pixmaps.

I'm having this issue in windows and Linux, in windows, my swap file hit 2.5G, in Linux, xorg grows over time.

I don't remember having this issue in either os with 7xxx drivers. Can anyone else confirm/deny this? I have a 6200A card.
d13f00l is offline   Reply With Quote
Old 01-26-06, 04:14 AM   #2
d13f00l
Registered User
 
Join Date: Apr 2005
Posts: 197
Default Re: Is there a memory leak in the 8xxx series drivers?

Code:
xrestop - Display: localhost:0
          Monitoring 33 clients. XErrors: 2
          Pixmaps:   65259K total, Other:     209K total, All:   65469K total

res-base Wins  GCs Fnts Pxms Misc   Pxm mem  Other   Total   PID Identifier
2600000  1244  125    5 1401  886    31927K     57K  31985K   ?   The buffer cache - Konqueror
3400000   147   31    1  246   51    12324K      6K  12330K 18522 Sharing Progress
1000000   479   40    1   81  143     7709K     16K   7726K   ?   #xorg
1800000   147   44    3  243  186     6499K     11K   6510K   ?   kicker
1600000    68   40    3  199  164     3409K      9K   3418K   ?   KDE Desktop
1400000   144  124    3  183  244      873K     15K    888K   ?   kwin
4000000   168   56    3  152   96      706K     10K    716K   ?   old.txt - KWrite
3200000   171   57    3  221  103      591K     10K    602K   ?   gmane.comp.emulators.wine.devel (moderated) - KNode
2c00000    52   53    3   68   55      327K      6K    334K   ?   mc - /home/diefool - Shell No. 3 - Konsole
2000000    50   47    3   64   51      311K      6K    318K   ?   xterm - Shell - Konsole
2400000    86   54    3   83   55      122K      7K    130K   ?   NVIDIA-README.txt - KWrite
3a00000    84   56    3   77   55      120K      7K    128K   ?   power_stone_2_c.txt - KWrite
3600000    84   55    3   83   59      118K      7K    126K   ?   power_stone_2_items.txt - KWrite
2200000   330   21    2   68   63       54K     11K     66K   ?   KMix
0a00000    20   20    1    4   49       42K      3K     45K 12499 kded
0400000    34   21    1   19   13       38K      2K     41K 12567 kio_uiserver
2e00000     2    6    0    6    4       35K    288B     35K 18468 kio_thumbnail
0c00000     2   17    0    1    8       20K    648B     20K 12513 ksmserver
1a00000    10    5    0   15    7       15K    528B     15K   ?   klipper
1c00000     6    7    0    9   18        9K    744B     10K   ?   Resize and Rotate
2800000     2    6    4    1   30        8B      4K      4K   ?   xterm
1e00000     2    6    4    1   30        8B      4K      4K   ?   xterm
2a00000    10    5    3    0   14        0B      3K      3K  5646 nspluginviewer
3c00000     0    1    1    0    3        0B      1K      1K   ?   <unknown>
0200000     0    1    1    0    0        0B      1K      1K   ?   <unknown>
1200000     3    1    0    0    4        0B    192B    192B 12510 kaccess
3000000     2    1    0    0    4        0B    168B    168B 12589 knotify
0600000     1    0    0    1    2        4B     72B     76B   ?   <unknown>
3e00000     1    1    0    0    0        0B     48B     48B   ?   <unknown>
3800000     1    1    0    0    0        0B     48B     48B   ?   xrestop
0800000     1    1    0    0    0        0B     48B     48B   ?   <unknown>
4400000     0    0    0    0    1        0B     24B     24B   ?   <unknown>
0e00000     0    1    0    0    0        0B     24B     24B   ?   <unknown>
I have some more stuff open, only 60 megs of images are loaded. What's the 300MB?
d13f00l is offline   Reply With Quote
Old 01-26-06, 10:14 AM   #3
ArmEagle
Registered User
 
Join Date: Jan 2006
Posts: 12
Default Re: Is there a memory leak in the 8xxx series drivers?

I've been changing drivers this week quite a bit, hoping to get better performance playing Battlefield2 using Cedega.

I had 81xx drivers installed for some time, and noticed that memory usage would go from 700MB to all 1G ram, plus 700MB swap.
Then someone told me that the 81xx drivers had issues with AGP memory, so i reverted back to 7676. This gave me the best performance, and memory usage is stable for multiple maps around 700MB (and try that on winXP!

I really hope the next 8xxx drivers will fix the issues at hand; memleak, AGP problem (related perhaps?), and a bugged implementation of 'EXT_packed_depth_stencil'.
ArmEagle is offline   Reply With Quote
Old 01-26-06, 12:51 PM   #4
Bratag
Registered User
 
Join Date: Sep 2005
Posts: 62
Default Re: Is there a memory leak in the 8xxx series drivers?

I have had similar issues and posted about them - nvidia claim no memory leak but all I know is the 7676 drivers didnt have the issue. I am reluctant to go back to 7676 though as 81xx is more stable for other things.

A.
Bratag is offline   Reply With Quote
Old 01-27-06, 03:35 AM   #5
d13f00l
Registered User
 
Join Date: Apr 2005
Posts: 197
Default Re: Is there a memory leak in the 8xxx series drivers?

I had talked to kon colivas about this, the developer of the CK kernel.

He said that there probably is a leak in the nvidia driver, as he has experienced this as well.

Coupled with some of the VM problems in 2.6.15, and a busted kernel patch I was running, the problem became apparent as my PC was swapping like hell.

Now, I'm on a fixed kernel. My PC isn't swapping, but X is still using way more ram than it should, quite possibly due to a Nvidia problem, but who knows.

Windows on my machine runs terrible, it swaps like hell with the latest drivers and there is nothing I can do about it.


Assuming this does affect windows users as well, maybe we will see a fix soon.



Also, the 8xxx drivers will lock up instantly if I touch my GPU clock, even 1mhz up or down, the ram clock I can still toy with. This is a known bug on the 6200A.

I'm in limbo here, 7xxx is much more unstable, but allows overclocking and doesn't suffer from this ram problem. 8xxx is faster, stable, but has a clocking bug, and chews ram(maybe...)

It's all speculation, I can't get any hard proof, as nothing seems to know where the ram is going, and profiling X for ram usage would be very complicated and confusing.
d13f00l is offline   Reply With Quote
Old 01-31-06, 04:25 PM   #6
Sentertainment
Registered User
 
Join Date: Jan 2006
Posts: 6
Default Re: Is there a memory leak in the 8xxx series drivers?

wierd, even so...I still notice way better speeds in the 8XXX series than in 7676, even after running it for a few hours.
Sentertainment is offline   Reply With Quote
Old 02-03-06, 07:14 AM   #7
d13f00l
Registered User
 
Join Date: Apr 2005
Posts: 197
Default Re: Is there a memory leak in the 8xxx series drivers?

Quote:
Originally Posted by Sentertainment
wierd, even so...I still notice way better speeds in the 8XXX series than in 7676, even after running it for a few hours.
Yeah, it definatly is faster in terms of 3d performance, but after a few days, my machine starts swapping like heck and I have to restart X. Something is up.
d13f00l is offline   Reply With Quote
Old 02-06-06, 03:20 PM   #8
Christian_Frank
Registered User
 
Join Date: Dec 2005
Posts: 85
Default Re: Is there a memory leak in the 8xxx series drivers?

Hi @all,

Same strange memory problems here:

I have a big problem with batllefield 2 under cedega.
The wine process is growing and growing until my hole system memory and swap is used.

This problems were starting at my side after i changed from:
Suse 10.0 32 bit, 1.076xx driver, 6600GT 768 MB Ram
to:
Suse 10.0 64bit 1.08xxx driver, 7800GT 1024 MB Ram.

I tried with kernel 2.6.13 and now with 2.6.15-arck3

Here is my top output:

op - 21:07:13 up 38 min, 3 users, load average: 2.06, 1.60, 1.04
Tasks: 83 total, 1 running, 82 sleeping, 0 stopped, 0 zombie
Cpu(s): 42.9% us, 3.3% sy, 0.4% ni, 42.0% id, 10.9% wa, 0.1% hi, 0.3% si
Mem: 1026308k total, 891776k used, 134532k free, 7968k buffers
Swap: 1574328k total, 1033976k used, 540352k free, 198716k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
6792 christia 6 0 1821m 630m 343m S 98.1 62.9 13:17.07 wine
5120 christia 0 0 104m 2956 1944 S 2.0 0.3 0:01.38 kded

So maybe this is a 64 bit problem or just a memory leak in 1.08xxx. ?

I would like to test 1.076xx but i don't know if my card 7800GT is already supported in 1.076xx ?

Regards, Christian
Christian_Frank is offline   Reply With Quote

Old 02-07-06, 02:11 AM   #9
Linuxhippy
Registered User
 
Join Date: Feb 2004
Posts: 585
Default Re: Is there a memory leak in the 8xxx series drivers?

I've exactly the same problems, I am running Xorg-6.8.2 (+composite) using the 8178 drivers on my FX5200 and sometimes after runnig X for some hours X uses huge amount of memory.
The 7xxx series of drivers never had problems like this so I guess its really the nvidia driver causing the problems.

lg Clemens
Linuxhippy is offline   Reply With Quote
Old 02-08-06, 03:34 AM   #10
d13f00l
Registered User
 
Join Date: Apr 2005
Posts: 197
Default Re: Is there a memory leak in the 8xxx series drivers?

Bump, anyone from NVidia around with any comments?
d13f00l is offline   Reply With Quote
Old 02-10-06, 03:30 AM   #11
Christian_Frank
Registered User
 
Join Date: Dec 2005
Posts: 85
Default Re: Is there a memory leak in the 8xxx series drivers?

Hi,

Now i found same mor time to test with cedega and battlefiled 2.

I switched back to 1.0.7676 on my Suse 10.0 64 Bit System.
battlefiled 2 runs much mor stabel now ( i played about 30 Min and the mem was about 1 GB) .
The only problem is, that my card is not supported by this driver version.
It shows only G70 Chip Card and not 7800GT.
And the top ouput of the memory usage looks horrible ( about 300 MB for X, 2 GB for Wine and abpout 200 MB for each K Apllication), but this can not be correct, cause i only have 2 GB ram and the machine did not swap.
So i think i get this wrong outpout, cause my card is not fully supported


Then i installed Suse 10.0 32 bit (Same cedega and battlefield2 version) and 1.0.8178.
After 10 Minutes of playing Battlefiled 2 with cedega, the memory jumps over 1GB and was still growing !!

So there musst be definitivly a memory problem with 1.0.81xx !!

So please, NVIDIA, get cedega and try Battlefield2 and you will see what i mean.

Please support us !!

Thanks,
Christian
Christian_Frank is offline   Reply With Quote
Old 02-12-06, 10:51 PM   #12
shifted
Registered User
 
Join Date: Dec 2004
Posts: 2
Default Re: Is there a memory leak in the 8xxx series drivers?

I am also experiencing the same issue. I'm using Xorg 6.8.2. I've had Xorg consume the entirety of my 2 GB of swap. pmap output is similar.

I'm currently using using the nvidia drivers 7667. I suspect it has something to do with Xorg, as this rapid memory leaking happened with a recent upgrade from 6.8.0 (along with other kubuntu updates).
shifted 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
First Sign of AMD's Radeon HD 8000 Series Found in Catalyst Drivers News Archived News Items 0 06-08-12 06:40 AM
Nvidia GeForce 301.42 WHQL drivers DSC NVIDIA Windows Graphics Drivers 5 05-29-12 11:12 PM
Measuring card memory usage peter_ga NVIDIA Linux 0 05-24-12 09:07 PM
Creative drivers ? SparrowHawk General Hardware 39 11-18-02 09:03 AM
Radeon 9700 not all that? sancheuz Other Desktop Graphics Cards 200 10-12-02 10:31 PM

All times are GMT -5. The time now is 05:10 PM.


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