nvidia GPU runs crazy Hot with new driver/distro
 

Ok so i have an HP dv2620us laptop. Its an amd machine with nvidia 7150m gpu.
It has always been a hot running machine.
I was running ubuntu 7.10 32bit and things were ok, i had decent and consistent 3D performance. 7.10 was using the 09x series of nvidia drivers, no PowerMiser or temp reporting. I upgraded to Hardy and my GPU was never less than 85C even in fluxbox and idle. If i did anything 3D it would quickly pass 105C and slow down massively.
I went back to 7.10. Now im running Deb Lenny, and having the exact same problem.
I rolled my own 2.6.25 kernel and am using 173 driver from nvidia.
Im trying to get a legacy driver running but it seems i need to switch to <=2.6.24.

has anyone else had this problem? Any ideas?
Thanks in advance

which tool are you using for the temperature?

You can install another kernel in the same distro.

You should try nvclock, you can change fanspeeds with it:
http://www.linuxhardware.org/nvclock/

for amani:
i'm using nvidia-settings and normal /proc interface to get temps.
And they are exactly the same values.
Yes i can use a diff precompiled kernel but that leaves no choice for the nvidia driver version, and Debian smp kernels include xen support that prevents nvidia installer from working.

for H_TeXMeX_H:
nvclock does not work with 7150m cards. As far as i can tell it doesn't work with laptops in general but some newer mobile GPU's might work.

i was mistaken in my original post, the nvidia version that runs fine in ubuntu7.10 is 1.0-14xxx. So it is a major generation past 09x drivers but not 169xxx or the like. The trouble now is that, i seem to be one of very few people with this problem and the older drivers wont build with newer kernels. sux, i wish nvidia would have the good sense to opensource their drivers, and godspeed to the "nouveau" project.

Thanks for all replies

where does it say that ? have you tried it ? I've checked google and there is no info on this particular card and whether nvclock will work with it.

Another data point (ok, it's an anecdote): My mother (an aeronautical engineer, not a cookie baker) had a laptop that would always get very hot (she used to run it on ice-packs to keep it cool). Anyway, it turned out that the problem was that the fan's mounting nuts had completely come off the motherboard due to the excessive heat melting the solder that held them in place. We had to take the thing apart and remount the nuts to get it working correctly.

HTH

Forrest

yeah i did try nvclock, it said something like "card not supported", i tried setting "Coolbits" to "1" but no dice. And on my maching the CPU and GPU share sink/fan and it is running full bore almost all the time (except in suspend). It's a strange problem that i should prolly take to nvidia forums, they just seem a bit oveloaded with vista isues at the moment :).

Some more info...
in gutsy with stock 2.6.22-15 kernel and nvidia-glx 100.14 i ran # glxgears and got ~1500fps for about 2min before it slowed down to ~850fps and that was as slow as it got.

in deb sid and hardy using >=2.6.24 kernels with 173.xx and 169.xx nvidia drivers glxgears runs ~850 frames for about a minute then slows down to ~200fps and stays there.

im trying to narrow it down. 169/173 drivers report temps and use "Powermiser" but work badly. The 100.xx drivers give no temps and dont seem to have powermiser but work ok.

At this point i'm not really expecting to solve the problem i just want to submit a good bug report and help the problem get solved.
i blame nvidia.

thanks again

Don't get me wrong, the fans in my mom's laptop would spin fine (and constantly). It's just that the heatsinks they were attached to were no longer attached to the chips they were supposed to cool.

Forrest

forrest, i just downloaded the maint. manual and will field strip this POS later. It's a good suggestion and easier to try than building a half dozen kernels on machine with heat problems :). thanks

If it performs differently with different nvidia drivers, then it's the nvidia drivers that are causing it. I've heard of them being more of a problem with the 8000 series, but from my experience with the rather sh*tty nvidia drivers, it can happen with any card. At least the drivers are better than the ati ones. Try not to run the latest kernel with the latest nvidia drivers, it just won't work.

Thanks Tex

Yeah im stuck with a bad mix of hardware, i need the newer b43 driver for wireless, or i have to stay with ndiswrapper, that means i cant run 64bit. im going to see if i can hack the 1.0-14 nvidia driver to build with 2.6.25
Not holding my breath, and now dualbooting.

hmm or maybe new b43 on older kernel......

This runs along the lines of what I've been seeing on my Acer Aspire 5520 with a GF7000M/nForce 610M IGP: it runs hot. Nvidia-settings says it's
stuck at the full 350 MHz clock and nothing I do will change it. Looking through the release notes in "Laptop Issues", Nvidia says that
Powermizer doesn't work on some systems, they don't know why, and they don't have a work-around. Sigh. I'm trying to upgrade the the newest
ACPI stuff manually, but I have a feeling it won't help: it seems I can control everything except the IGP. :(

Mike

ok i've done it!!
 

after much buggery, i've got everything i want working how i knew it could.
I'm now running Deb-Sid-amd64 with a hand rolled 2.5.25 kernel.
I got the .run for nvidia driver version 100.14.19 and applied the patch for 2.6.25 found here in post #10.

with this driver things run like they did in gutsy.
hope some people find this and it helps.
The best solution is not to buy this laptop in the first place, but live and learn i guess. I hope the coming ATI drivers from amd are up to snuff and we can all bail on nvidia.

have fun

I hope nouveau will deliver something useful within the next few years. I personally will never buy another ATI card, and I swore this even when I was using Window$, their cards suck.

hmmm, yeah i swore some 7-8 years ago that i was gonna stick with nvidia because of the swell linux drivers, but the last year or so has been pretty nasty. I think ATI cards and drivers will be getting much better in the coming years, but its pretty big gamble on a laptop with built in graphics. Anyway im back to productivity......











--support bacteria, the only culture most people have