|
|
#1 | |
|
Kris Moore
Join Date: May 2006
Posts: 19
|
[ Solved below]
I'm trying out the latest beta 195.22 driver, and am experiencing an almost immediate hard-lock after bringing up X. This is on a 9800GT, but another of my systems with an older 6800 works fine. This hard-lock occurs anywhere from 1-15 seconds after x starts. I've tried running the nvidia-bug-report.sh script in that time, but something is that script causes it to lock up immediately, and I can't get a saved log file from it. Attached is a log from nvidia-bug-report.sh while using the "nv" driver, hopefully it'll still have enough info for you ![]() BTW Running FreeBSD 8.0-Release, tried this on amd64 and i386 versions, doesn't make a difference. On this system the older drivers worked great on FreeBSD 7.x. Last edited by kmoore134; 12-11-09 at 09:31 AM. Reason: [Solved] |
|
|
|
|
|
|
#2 | |
|
Kris Moore
Join Date: May 2006
Posts: 19
|
Solved it here!
Turns out it the crash was caused by having powerd enabled in rc.conf: powerd_enable="YES" powerd_flags="-a adaptive -b adaptive" # set CPU frequency As soon as I commented out those lines, the driver works like a champ again. Dual-head and everything ![]() |
|
|
|
|
|
|
#3 |
|
Lobo
Join Date: Dec 2009
Location: Brasil
Posts: 10
|
I had an instability problem too.
AMD phenon 955 / FBSD 8-STABLE amd64 Nvidia 9800 GT Even if I had powerd disabled, the system was still randomly unstable but I wanted to keep powerd running to keep the CPU cool so I started to dig. I noticed I had those 3 lines from sysctl: -dev.cpu.0.freq_levels: 3200/28215 2800/24688 2500/18625 2187/16296 2100/12650 1837/11068 1575/9487 1312/7906 -dev.acpi_throttle.0.freq_settings: 10000/-1 8750/-1 7500/-1 6250/-1 5000/-1 3750/-1 2500/-1 1250/-1 -dev.hwpstate.0.freq_settings: 3200/28215 2500/18625 2100/12650 800/5070 with debug.cpufreq.lowest=1312 in sysctl.conf The acpi_throttle list seemed weird because it only had weird frequencies. I wanted to know which frequencies powerd uses. I got a hint from http://wiki.freebsd.org/TuningPowerConsumption on how to disable acpi_throttle (and overall power savers) and get it out of the way of powerd, so I added to loader.conf: hw.pci.do_power_nodriver="3" hint.acpi_throttle.0.disabled="1" hint.pcm.0.buffersize="65536" hw.snd.latency="7" Then this showed up: dev.cpu.0.freq_levels: 3200/28215 2500/18625 2100/12650 dev.hwpstate.0.freq_settings: 3200/28215 2500/18625 2100/12650 800/5070 So I found where powerd (dev.cpu.0) gets its frequency list from! a much more uniform list, so I set debug.cpufreq.lowest=800. Now it shows dev.cpu.0.freq_levels: 3200/28215 2500/18625 2100/12650 800/5070 dev.hwpstate.0.freq_settings: 3200/28215 2500/18625 2100/12650 800/5070 800 is really low, so, so far I haven't seen the system drop to that but right now, the machine has been on for an hour (KDE4 w/OpenGL composite, e-mail, downloads, 2 VMs) and... dev.cpu.0.temperature: 45.0C dev.cpu.1.temperature: 45.0C dev.cpu.2.temperature: 45.0C dev.cpu.3.temperature: 45.0C dev.amdtemp.0.%desc: AMD K8 Thermal Sensors dev.amdtemp.0.%driver: amdtemp dev.amdtemp.0.%parent: hostb4 Which is pretty cool for this CPU type. Plus, the lock-ups (openGL+VBox) are gone ! I still have to do the final test, which is to run a VM with 3D acceleration enabled. This one always froze the machine, no matter what! |
|
|
|
|
|
#4 | |
|
Registered User
Join Date: Dec 2009
Location: Germany
Posts: 7
|
Quote:
I have exactly the same cpu with a ASUS M4a78T-E Board, and run in the same problem, you need to be make a Bios update that solve the powerd problem. See also here - Martin |
|
|
|
|
![]() |
| Thread Tools | |
|
|