180.29 on XPS-1730 Report

[zBeeble42]

I've been away for awhile. My last look at the driver was 180.03.

On the XPS-1730 (which has dual 8700M SLI video), the nvidia binary driver still soaks up 100% of one CPU while running.

I havn't had a chance to update wine (to test EVE) or to update mplayer (to test video support --- only -vo gl and -vo gl2 are working with my system --- but mplayer is a bit old)... so this 100% CPU thing is my largest, longest outstanding issue.

Xander: you said that you thought you'd nipped it awhile back... any thoughts?

[zander]

Nipped, what, exactly? I reproduced and fixed the mode validation problem you reported on an XPS-1730 notebook (said problem turned out to be specific to that notebook), but I didn't see 100% CPU utilization, interrupt storms or any other problem of this sort. The only thing I can suggest is that you identify what needs to happen in order for the problem to reproduce - i.e. does it happen if you just start X with e.g. just `xlogo` to keep it alive, does it happen with different versions of FreeBSD, does it happen with both UP and SMP kernels, etc.?

[zBeeble42]

Well... I don't have unlimited resources. It has happened with every version of the driver that has worked with this card (I believe we were at 169 or somesuch?). I don't have the ability to run other versions of FreeBSD easily.

If I run systat with no arguments, the "int16" process (nvidia) takes 90% to 100% of the CPU when X is running. This is true even if X is running with no clients. If I run "systat -vmstat" the number of interrupts per second attributed to nvidia0 is 11644 (+- a bit) and 274 to nvida1 (plus some other drivers).

Obviously, I can't tell you if this is specific to the 8700M SLI or if it also affects the 8800M SLI --- I don't have the other hardware. I also don't know what hardware you're testing.

I will test the SMP kernel tomorrow. It definitely happens with the SMP kernel.

[zander]

I've tried to reproduce the interrupt storm problem on a variety of systems and FreeBSD versions - most recently on a XPS-1730 notebook. Like I said, however, I have not yet seen this kind of problem.

[zBeeble42]

Did your 1730 have the 8800M-SLI or the 8700M-SLI?

[zBeeble42]

BTW... I can offer my machine as a test system if you want to poke around.

[zBeeble42]

OK. UP kernel: disaster. Still mostly interrupts. Interestingly only about 95% or 96%... which means that I could still switch to the console to get enough CPU to revert the kernel and reboot.

So, using FreeBSD-7.1, both SMP and UP on a Core-2-Extreme X7900 with the 8700M-SLI in the Dell 1730 laptop gives most of one CPU full of interrupts... somewhere over 11000 interrupts per second.

I havn't tried UP kernels before, but I've had this problem from 7.0 forward.

The interrupts start when I start X (works with just starting X from the console command line). It doesn't require any X clients to be running. The interrupts stop when I stop X (CTRL-ALT-Backspace or CTRL-C if I'm on the console).

[zBeeble42]

I was actually giving some consideration to buying an 8800M GTX SLI. They seem to be in the $1000 range --- and would give me a boost on some of the newer more demanding games. If I do this --- I'll have another perspective on this --- but since I don't like taking my laptop apart a bunch of times, it would be convenient if we could address any tests on the 8700M GT SLI beforehand.