> -----Original Message----- > From: Jean Delvare [mailto:khali@xxxxxxxxxxxx] > Sent: Tuesday, December 11, 2012 3:56 AM > To: Leslie Rhorer > Cc: lm-sensors@xxxxxxxxxxxxxx > Subject: Re: Fried CPU? > > Hi Leslie, > > On Mon, 10 Dec 2012 12:40:07 -0600, Leslie Rhorer wrote: > > I fear I know the answer to this question before I ever ask it, but I'm > > going to ask anyway, just in case there is a better answer out there. > > > > I purchased a used AMD Phenom II x 6 CPU and installed it along with > > a water cooling system in one of my servers. It was working perfectly, > > except that the sensors binary evidently does not support alarms for the > > atk0110-acpi-0 interface, so I could not trivially implement a shutdown > in > > the case of cooling failure, and like an idiot I didn't take the time to > > write an interface at the time. I put it on the back burner. > Unfortunately > > that burner was the CPU itself and it was lit three weeks later when the > > cooling pump failed while I was at work. I came home to a CPU whose > > temperature was 101°C, and had failed processing several times during > the > > day. Most of the fluid was boiled away inside the coolant feed tubes, > and > > had I been gone longer, the CPU temp would have gone through the roof. > In > > any case, I pulled the water cooler and put the old aluminum fin cooler > back > > on. The system came up, and seemed to work fine for a couple of more > weeks, > > except that the atk0110-acpi-0 sensors were not quite stable. I did not > > have a monitor on them previously, so had no way to know if this were an > > artifact of the CPU's having been excessively hot for a time, or not. > When > > the replacement pump finally came in, I once again installed the water > > cooler and brought the system up. It only worked for a few minutes, and > > then shut itself down, because the monitor was saying the CPU was > > ridiculously hot. It was difficult to even get it to bring up the BIOS, > but > > after letting the system sit without power overnight, I was able to get > it > > back up and disable the monitor before it shut down the system again. > > CPU temperature issues showing up right after changing the cooling > system usually indicates bad thermal contact between the CPU and the > heatsink (or whatever the equivalent part is for liquid-cooled > systems.) For example missing or improperly applied thermal paste, or > loose mounting. That might be true if it showed a reasonably high temperature, like 80°C, or 200°C, not temperatures higher than those found in the core of the Sun. I know how to mount a cooling system, and this one is particularly simple. > > Processing seems stable, but the chip is reporting absurdly high > > temperatures to the sensors command. At first boot, it reported a > > reasonable 36°C in the BIOS, but after coming up fully, the sensors > command > > is reporting temperatures of 429496652.5°C. It is interesting the chip > > You will notice that 429496652.5 is 0xfffffcfd/10 i.e. this looks like a > negative value being improperly displayed as a positive value. Yeah, that is why I was thinking perhaps there was a software error somewhere. Well, not really "thinking", but grasping at straws. > This > could either be an error code being improperly processed as actual > data, or a dead thermal sensor. It's not dead, at least not entirely. First of all, the 400 million dollar number does vary with the actual temperature. What's more, the chip has been reporting good numbers since yesterday. Right now it is reporting 41°C. > > reported a correct temperature to the BIOS, but that `sensors` is > reporting > > garbage. > > Even more so when using the asus_atk0110 driver, which gets its > readings straight from the BIOS. I don?t understand what you mean. What is even more so? > > I did run the `sensors-detect` command prior to shutting the > > system down in order to check something. > > When did you do that exactly? sensors-detect is known to have caused > serious trouble on a small number of systems, but given the history of > your system this doesn't seem like the prime suspect for your specific > problem. It was run just prior to swapping out the solid state cooling for the liquid cooing when the replacement arrived. > What motherboard is this? Asus Crosshair II Formula > Which version of lm-sensors or sensors-detect > did you use? 1:3.1.2-6 > > > My forlorn hope is that may have > > caused some issue or that there is some other software fix for this. I > > could run it without CPU temperature monitoring, but with an active > cooling > > system, that does not seem wise. (Hind sight is 20/20!) I suppose I > could > > rig an external temperature probe. The motherboard does have external > > thermal sensor inputs, but the case of the waterblock is plastic, making > the > > measurement of the CPU temperature with a thermistor problematical. > > > > Is it at all possible this CPU is not fried, or that there is some > > software work-around? > > I can't think of any workaround. If the weird temperature values > started being reported right after your first pump died, I'd say the > CPU is fried. If it started right after mounting the new pump, I'd say > it was improperly installed. Neither is the case. _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
