Hi, Am Freitag 25 August 2006 19:50 schrieb Jean Delvare: > Hi Prakash, > > > I set them to: > > powernow-k8: 0 : fid 0xc (2000 MHz), vid 0x11 > > powernow-k8: 1 : fid 0xa (1800 MHz), vid 0x13 > > powernow-k8: 2 : fid 0x2 (1000 MHz), vid 0x1b > > The K8 family uses the VID conversion table described in AMD document > 26094 (Table 74). Inside the kernel it is known as VRM version "2.4. > This table leads me to the following values: Ah nice to know, thx! > (2000 MHz), vid 0x11 -> 1.125 V > (1800 MHz), vid 0x13 -> 1.075 V > (1000 MHz), vid 0x1b -> 0.875 V > > These values are significantly lower than the ones I am used to (1.40 > V, 1.35 V and 1.10 V, respectively). I am surprised by your words: "I > set them to"... Do you mean that you somehow _decided_ to use these > values instead of the standard ones? Yes, I used a hack with which I can override the default vids. Of course I did stability tests and ALU and FPU (->prime95) seems stable so far. (Using gentoo I would have nonoticed if the compiler creates junk. ;-) So, by your values given above, I have got a EE SFF version for the price of a regular one. :-) > Please check that the value of /sys/class/hwmon/hwmon0/device/vrm is > 8.2 as expected. It gives 24. > > converting the values properly. This could be caused by a device > misconfiguration, or maybe by the IT8716F chip itself being > underpowered? I don't think the chip is responsible for it, if I do it on purpose, or do you think it also causes some deviation? > The driver should print some messages in the log when you load it, can > you please copy them here? it87: Found IT8716F chip at 0x290, revision 0 it87: in3 is VCC (+5V) it87: in7 is VCCH (+5V) it87: fan4 is enabled it87: fan5 is enabled Not too interesting, it seems, or is it? > If temp2 is taken on the northbridge, and if you have no northbridge > fan, and if your CPU is underpowered and has a good fan, it may make > sense. I used to tell users that CPU temperature must always be above > motherboard temperature, but I seem to be proven wrong more frequently. > The best way to differenciate between CPU temperature and motherboard > temperature is to put some load on the CPU, and see which temperature > raises faster. A simple "md5sum /dev/zero" makes the trick. Yes, I also concluded it this way, as the chipset is cooled passively and if I stress the CPU temp1 goes up fast and when going idle again, it goes down fast, as well. :-) > Yeah, there are interesting similarities between your report and this > one. Same motherboard manufacturer... I have had other reports for > IT8716F chips where the voltages were in line with what the BIOS said. > I wonder if these motherboards could be voluntarily underpowering all > the devices, instead of only the CPU, to same power? That would be > weird. > > One way to test this is as follows: > Reboot your system, and make sure that the it87 driver will not be > loaded at boot time. > Dump the contents of the chip: > isadump -y 0x295 0x296 > /tmp/before.dump > Load the it87 driver, run sensors once. > Dump the contents of the chip again: > isadump -y 0x295 0x296 > /tmp/after.dump > Send both files here. Do you still want me to get the dumps, if you know now that I underpower on purpose? > If the voltage registers in both dumps are the same, it means the it87 > driver has nothing to do with it. > > > > The decoding depends on the CPU. In theory the kernel detects it and > > > picks the right decoding formula but if your CPU is really new, it > > > might use a table we don't know yet. What is your exact CPU model? > > > > cpu family : 15 > > model : 75 > > model name : AMD Athlon(tm) 64 X2 Dual Core Processor 3800+ > > stepping : 2 > > > > It is the new AM2 "regular" version. > > Do you know where I can get technical documentation for your > processor? My reference document for K8 CPUs was AMD document 30430 so > far, but I can't find your CPU revision there. > > > > > CPU: 780 > > > > POWER: 777 > > > > > > > > wheres sensors reports too high values > > > > fan1: 1323 RPM (min = 6490 RPM) ALARM > > > > fan2: 0 RPM (min = 6490 RPM) ALARM > > > > fan3: 1564 RPM (min = 0 RPM) > > > > > > So you are affected by the problem too... > > > > > > What do you think are the correct values for your fans, the BIOS' ones, > > > or sensors' ones? > > > > I think the bios ones are right, as if I disable fan management, it shows > > CPU fan at 2335 which fits well enough to Artcic Cooling Freezer 64 specs > > at 2200rpm. > > Almost all reporters seem to agree on that fact, and one ITE tech guy > just confirmed it to me... I'll fix the driver. > > > (Yes my machine is "ultra silent", so it is possible that fans run > > at ~800 rpm when idle. It also just consumes 50Watt when idle...) > > Nice. I tried something similar for my home server, except that it's > using old hardware (Pentium III 800 MHz...) It is somewhere around 50 W > too, and pretty silent as well (mostly thanks to the Fortron PSU and the > Seagate hard disk drive.) > > > On the other hand, when I disable fan management, the fan makes quite > > some noise which might be too high for a 2200rpm fan, but maybe I am just > > used to silence. ;-) > > Depends on the size of the fan. A 120 mm fan at 2200 RPM should be > almost silent. A 60 mm fan at 2200 RPM can already make some noise, > although it should be acceptable. > > > > > fan1: 46551 RPM (min = 6490 RPM) ALARM > > > > fan2: 0 RPM (min = 6490 RPM) ALARM > > > > fan3: 1551 RPM (min = 0 RPM) > > > > > > Very strange. Are you using fan speed control? This could induce some > > > noise in the speed sensing. > > > > Yes Asus' so called "q-fan2". > > I guess you won't observe the phenomenon with it disabled. > > > > Do you have similar effects when looking at > > > the values in the BIOS? > > > > Well, I haven't observed the bios long enough, as such values usually > > just appear for short times. But I noticed at times the bios also shows > > wrong values, as it shows "0" instead of ~800. Perhaps here the same > > thing hapens, but the bios clips the value correctly? > > I'm not sure about the "correctly" but yet the BIOS is most certainly > trimming the values it considers too high to be relistic. But I am > reluctant to do this in our drivers. It's up to the hardware to improve > to make it possible to control fans and still monitor them. I think > that's where we are going with 4-wire fans. -- (?= =?) //\ Prakash Punnoor /\\ V_/ \_V -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060825/a1c7eef6/attachment.bin
