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

Newegg Daily Deals

Reply
 
Thread Tools
Old 04-03-03, 02:39 AM   #1
Haesslich
Registered User
 
Join Date: Apr 2003
Location: Alton, IL
Posts: 2
Default NVidia drivers causing audio problems

Okay, this has been driving me nuts for a while....

I'm running a triple head setup using an MSI GF4 TI4400 128MB to drive the primary and secondary displays (21" and 17" CRTs respectively) and a TNT2 M64 PCI to drive the tertiary display (also a 17" CRT).

My problem is that with the accellerated NVidia drivers enabled, my audio sounds scratchy, like I'm listening to a very weak radio station.

The problem seems to be caused by the PCI card, because if I disable it, the sound is fine. If I disable the NVidia drivers and use the plain nv driver that comes with XFree86, the sound is also fine. There's nothing wrong with the PCI card, I tried replacing it with a PCI GF4 MX440 with the same result. The sound card is an SB Live and I've tried both OSS and ALSA with no discernible difference in the sound.

I'm currently using the 4191 drivers, I tried the 4349 but they refuse to give me any output on the secondary head of the TI4400 - and the sound is still scratchy.

I'm currently using Mandrake 9.1, the problem was the same with 9.0. I've tried different kernels, compiling my own and modifying Mandrake's enterprise and multimedia kernels. I've also trying changing PCI latency settings with setpci, all to no avail. I've also tried moving the PCI boards to different slots in case there was an IRQ conflict causing the problem.

Hardwarewise, I'm using a Tyan Thunder K7 motherboard with dual Athlon MP2000+ CPUs and 1 gb RAM.

Any ideas?
Haesslich is offline   Reply With Quote
Old 07-26-03, 03:09 AM   #2
xiphmont
Registered User
 
Join Date: Jul 2003
Location: Somerville, MA
Posts: 17
Send a message via AIM to xiphmont
Default *exact same problem*, tried alot of hardware

I have the same problem. As this is supposed to be an *audio editing machine*, this problem is a killer.

I have a three head setup with one dual-head AGP card (FX5200) and one single-head PCI card (Geforce2). Any display to the PCI card causes the sound to fall apart.

I've tried four (!) different Geforce PCI cards (Geforce 2 and Geforce 4) and six (!) different sound cards (SB Live, SB Live 5.1, ESS Maestro3 , Ensoniq AudioPCI 1371 and two nonames). I've tried every imaginable combination in the PCI slots. I've tried different latency settings. The only solution is to yank the PCI video card from the machine or disable it in the server config.

The mobo is a MSI K7D Master-L with dual Athlon MP2400+, AMD 760 MPX rev 2.

Monty
xiphmont is offline   Reply With Quote
Old 07-26-03, 03:18 AM   #3
Thunderbird
 
Join Date: Jul 2002
Location: Netherlands, Europe
Posts: 2,105
Default

Problems like this have existed in the past. It useally happens when the videocard and soundcard for some reason are sharing an irq. (especially with creative sblive cards ..)

Unfortunately not much drivers contain options to change the irq, so the only way is in various case to move cards to different slots. Perhaps it helps but first check if the cards are sharing an irq.

In both your cases it can also be the second pci videocard. The bandwith of the pci bus is limited and videocards consume lots of bandwith. (likely more when using the nvidia drivers instead of the nv because of more features and optimizations)
Thunderbird is offline   Reply With Quote
Old 07-26-03, 04:04 AM   #4
xiphmont
Registered User
 
Join Date: Jul 2003
Location: Somerville, MA
Posts: 17
Send a message via AIM to xiphmont
Default

Quote:
Originally posted by Thunderbird
Problems like this have existed in the past. It useally happens when the videocard and soundcard for some reason are sharing an irq. (especially with creative sblive cards ..)

Unfortunately not much drivers contain options to change the irq, so the only way is in various case to move cards to different slots. Perhaps it helps but first check if the cards are sharing an irq.

In both your cases it can also be the second pci videocard. The bandwith of the pci bus is limited and videocards consume lots of bandwith. (likely more when using the nvidia drivers instead of the nv because of more features and optimizations)
I already addressed every suggestion you made in what I originally wrote.

a) they are not sharing IRQs. detailed dmesg information upon request. Also, given that I've tried many ordering combinations, that also takes care of the possibility.

b) You can't change the IRQ of a PCI card without physically moving it. IRQ allocation on a PCI bus is hardwired.

c) The bandwidth of the PCI bus is much lower, yes, and that's exactly what the latency timer mechanism is for; to prevent one device from hogging the bus and driving latency-sensitive devices into starvation.

d) I tried four cards that were not SBLive variants, in fact, that were utterly unrelated to any chipset Creative ever used.

I appreciate the desire to help, but we're past the point of vague rules-of-thumb.

Monty
xiphmont is offline   Reply With Quote
Old 07-27-03, 10:43 AM   #5
xiphmont
Registered User
 
Join Date: Jul 2003
Location: Somerville, MA
Posts: 17
Send a message via AIM to xiphmont
Default Found it the problem! Two weeks *not* wasted!

...granted, the current form of the solution is a bit heavy handed.

It's MTRR write-combines on the PCI bus that are causing the huge latency spikes and crippling audio. Ripping MTRR out of the kernel entirely solves the problem (simply disabling the write combines via /proc is temporary. The driver will just reenable them if it can).

Eliminating MTRR support also kills the nvidia driver performance.

In my case, I can live with it if need be; I meant it when I said flawless audio was more important to me than fast video. OTOH, I can't help but think there's a happy medium here that doesn't involve either abusing the PCI bus's latency handling or backing off so far as to make video unbearably sluggish. I mean, we're talking back to the days of unaccellerated Oak cards here ;-)

Hopefully now that there's a glimmer of hope the nVidia folks will comment...

Monty
xiphmont is offline   Reply With Quote
Old 09-10-03, 10:07 AM   #6
vvarder
Registered User
 
Join Date: Sep 2003
Posts: 1
Default

Hi all,

I had the same problem with the sound stutter (I have a Athlon XP 1700 with a nVidia GeForce 2 PCI and a Soundblaster PCI128). I took the mtrr out -- not wanting to recompile the kernel again, I just issued the command (from the mtrr linux kernel documentation):
echo "disable=#" >| /proc/mtrr
For each of the registers and I no longer recieved any stutter. However, I agree that this seems a band-aid way to fix it, and it does kill performance (although my few test opengl apps still had high frame-rates). For me, the sound was like listening to a radio station that was out of range with mtrr. Is there a "compromise" way to have at least partial mtrr support without killing sound performance? Maybe using another register to burst the sound across the pci bus? Or maybe just reducing the size of the register entries? Any thoughts would be appreciated.

-Josh

Last edited by vvarder; 09-10-03 at 10:11 AM.
vvarder is offline   Reply With Quote
Old 09-10-03, 11:13 AM   #7
xiphmont
Registered User
 
Join Date: Jul 2003
Location: Somerville, MA
Posts: 17
Send a message via AIM to xiphmont
Default Long since given up

I eventually just decided I could have my cake and eat it too, and bought a triple-head Parhelia, put it on AGP and sold off the nVidia cards.

The PCI bus problem isn't really nVidia's fault (it looks like most PCI video drivers have this problem, and I suspect Athlon chipsets make it worse), but I too wish there had been some way to more intelligently handle PCI video writes than 'screw the rest of the system!'

What really eventually killed nVidia in the deal was the XFree86 driver crashing three times a day. I've had the Parhelia running for a month without a glitch now, and I have accel across all three heads.

Monty
xiphmont is offline   Reply With Quote
Old 09-10-03, 11:18 AM   #8
xiphmont
Registered User
 
Join Date: Jul 2003
Location: Somerville, MA
Posts: 17
Send a message via AIM to xiphmont
Default

Quote:
Originally posted by vvarder
Hi all,

I had the same problem with the sound stutter (I have a Athlon XP 1700 with a nVidia GeForce 2 PCI and a Soundblaster PCI128). I took the mtrr out -- not wanting to recompile the kernel again, I just issued the command (from the mtrr linux kernel documentation):
echo "disable=#" >| /proc/mtrr
For each of the registers and I no longer recieved any stutter. However, I agree that this seems a band-aid way to fix it, and it does kill performance (although my few test opengl apps still had high frame-rates). For me, the sound was like listening to a radio station that was out of range with mtrr. Is there a "compromise" way to have at least partial mtrr support without killing sound performance? Maybe using another register to burst the sound across the pci bus? Or maybe just reducing the size of the register entries? Any thoughts would be appreciated.

-Josh
Oh, BTW! You don't need to kill *all* mtrr; just the one range used by the video card. Look up the memory range via lspci -v, then disable the one register associated with that range.

That avoids disabling write-combine for things that need it, like the bridges and system controller.

Monty
xiphmont 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 Drivers Receive Windows 8 Certification News Archived News Items 0 06-01-12 06:30 AM
Rumor regarding lack of 680 availability ViN86 Rumor Mill 6 05-09-12 05:48 PM
Radeon 9700 not all that? sancheuz Other Desktop Graphics Cards 200 10-12-02 10:31 PM
Official detonator 40.71 feedback thread Crake NVIDIA Windows Graphics Drivers 88 10-01-02 12:38 PM
Problems with Mandrake and NVidia drivers..... xMonoxide187x NVIDIA Linux 2 08-22-02 02:55 PM

All times are GMT -5. The time now is 10:58 AM.


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