Reliable CPU fan speed controller?
in [Hardware]
|
Prev: USB SATA caddy size limit?
Next: Talk about vintage hardware, impressions from a visit to the Computer Museum in Mountain View, CA.
From: Paul on 27 Oct 2009 19:06 JD wrote: > > Many thanks again Paul. You are a huge help. > > Yesterday I spent an hour online looking for a good quality picture of > the motherboard and failed, even on the Intel web site. The images were > far too small to see any writing on the board. Of course I never even > dreamed > of visiting Japan to look for it ;-) You are a resourceful guy! Were you > thinking of a geisha girl when you visited? > > Then I opened up the computer and had a closer look. The case fan (only > one) > is connected to the power supply and not to the board. I looked around for > other connections and there were no more than appeared on the schematic we > looked at a few days ago. I'll change that case fan connection to the > motherboard when I get a plug for the connection and some time. > > I also looked in the area to the rear of the CPU between the AGP card > and the > power supply. There is the 12V board supply connector and two other small > connectors beside the AGP card. These are for ATAPI and the rear case fan > - not installed. > > I started up Speedfan and it does indeed show 4 fans. In the few minutes > it was > running it knocked the CPU-core voltage reading down to zero 3 times and > each time > the Intel monitor popped up a warning. That SpeedFan prog makes me worry. > > The CPU fan has been running and whining at about 4,300 rpm while all > the temperatures > are showing 25 C. Dumb design this system. > > I'm still trying to get through all your post above. Back soon, I hope. > Thanks again for > your efforts. If you have readings getting corrupted, they're probably coming from the LM85B, which sits on the SMBUS. It means there is a second piece of software accessing the chip and monitoring it. That could be the "Intel monitor". If one program runs at a time, and you exit the other program, you may find the readings more reliable. If the people writing the code for those programs could agree on an access method to the hardware, there might never have been any corruption in the first place. Since nobody in a leadership role, could define a semaphore for access to SMBUS, we're stuck with getting crappy readings. The only time the SMBUS is guaranteed reliable, is when the BIOS is doing its POST routine. In terms of connections then, you're saying the case cooling fan is connected to a 4 pin Molex power connector from the power supply ? Or is there a fan header cable coming from the supply, with the smaller (motherboard style) connector on it ? Some power supplies have a voltage controlled two or three pin fan connector, and the voltage is set according to how hot the power supply is getting. Since the power supply gets a stream of air from the computer case, the mechanism roughly causes the case fan to speed up, when the computer case air is getting hot. I presume the CPU cooler is plugged into the CPU header on the motherboard. If you're seeing evidence of the LM85B in Speedfan, then you should be able to make changes to its registers. If a BIOS or other software is also present and making changes, there is nothing to prevent the two programs or softwares from getting into a fight. I've had pretty good luck here, because my BIOSes tend to set the hardware just the once, and then they don't do any more writes. It allows the usage of Speedfan, without problems. You might wonder how the BIOS can run, while the operating system is present. There is something called SMM, which is a means by which the BIOS can take control from the OS, for short intervals. If implemented poorly, this can result in detectable issues at the OS level. If the outage caused by the BIOS is long enough, you can lose the counting of clock ticks (causing system time to drift while the OS is running). If the outage is a shorter interval, the system time may remain accurate, but the DPC latency (time until a deferred procedure call is serviced) may be lengthened. There is one utility for checking that, a DPC Latency utility, that can be used to check. For example, on my newest system, the average is perhaps 800 microseconds for DPCs to be serviced. I may see the occasional millisecond level response, if the system is heavily loaded. So if SMM steals control of the system for a long period of time, the only evidence might be collected indirectly, by measuring DPC latency. http://en.wikipedia.org/wiki/System_Management_Mode I've never seen any product documentation for motherboards, that references SMM or what the manufacturer may be doing or not doing. So whether there is any usage of SMM, is any one's guess. But if your SMBUS was getting corrupted, and you were absolutely sure that only Speedfan was running, it would suggest the BIOS is popping in occasionally, and doing something to the SMBUS. Paul
From: JD on 28 Oct 2009 13:29 Paul wrote: > JD wrote: > >> >> Many thanks again Paul. You are a huge help. >> >> Yesterday I spent an hour online looking for a good quality picture of >> the motherboard and failed, even on the Intel web site. The images were >> far too small to see any writing on the board. Of course I never even >> dreamed >> of visiting Japan to look for it ;-) You are a resourceful guy! >> Were you >> thinking of a geisha girl when you visited? >> >> Then I opened up the computer and had a closer look. The case fan >> (only one) >> is connected to the power supply and not to the board. I looked around >> for >> other connections and there were no more than appeared on the >> schematic we >> looked at a few days ago. I'll change that case fan connection to the >> motherboard when I get a plug for the connection and some time. >> >> I also looked in the area to the rear of the CPU between the AGP card >> and the >> power supply. There is the 12V board supply connector and two other small >> connectors beside the AGP card. These are for ATAPI and the rear case fan >> - not installed. >> >> I started up Speedfan and it does indeed show 4 fans. In the few >> minutes it was >> running it knocked the CPU-core voltage reading down to zero 3 times >> and each time >> the Intel monitor popped up a warning. That SpeedFan prog makes me worry. >> >> The CPU fan has been running and whining at about 4,300 rpm while all >> the temperatures >> are showing 25 C. Dumb design this system. >> >> I'm still trying to get through all your post above. Back soon, I >> hope. Thanks again for >> your efforts. > > If you have readings getting corrupted, they're probably coming from > the LM85B, which sits on the SMBUS. It means there is a second piece > of software accessing the chip and monitoring it. That could be the > "Intel monitor". If one program runs at a time, and you exit the other > program, you may find the readings more reliable. If the people > writing the code for those programs could agree on an access method > to the hardware, there might never have been any corruption in > the first place. Since nobody in a leadership role, could define > a semaphore for access to SMBUS, we're stuck with getting crappy > readings. The only time the SMBUS is guaranteed reliable, is > when the BIOS is doing its POST routine. If you are sure of the that SpeedFan won't damage the system, I'll use it and leave Intel's program off for the present. > > In terms of connections then, you're saying the case cooling fan is > connected to a 4 pin Molex power connector from the power supply ? This is correct - it is connected directly to the computer's power supply. Or is > there a fan header cable coming from the supply, with the smaller > (motherboard > style) connector on it ? Some power supplies have a voltage controlled two > or three pin fan connector, and the voltage is set according to how hot > the power supply is getting. Since the power supply gets a stream of > air from the computer case, the mechanism roughly causes the case fan > to speed up, when the computer case air is getting hot. > > I presume the CPU cooler is plugged into the CPU header on the motherboard. Yes, it is. > If you're seeing evidence of the LM85B in Speedfan, then you should be > able to make changes to its registers. If a BIOS or other software is > also present and making changes, there is nothing to prevent the two > programs or softwares from getting into a fight. I've had pretty good luck here, > because my BIOSes tend to set the hardware just the once, and then they > don't do any more writes. It allows the usage of Speedfan, without > problems. > > You might wonder how the BIOS can run, while the operating system is > present. There is something called SMM, which is a means by which the > BIOS can take control from the OS, for short intervals. If implemented > poorly, this can result in detectable issues at the OS level. If the > outage caused by the BIOS is long enough, you can lose the counting of > clock ticks (causing system time to drift while the OS is running). > If the outage is a shorter interval, the system time may remain accurate, > but the DPC latency (time until a deferred procedure call is serviced) > may be lengthened. There is one utility for checking that, a > DPC Latency utility, that can be used to check. For example, on > my newest system, the average is perhaps 800 microseconds for > DPCs to be serviced. I may see the occasional millisecond level > response, if the system is heavily loaded. So if SMM steals > control of the system for a long period of time, the only > evidence might be collected indirectly, by measuring DPC > latency. > > http://en.wikipedia.org/wiki/System_Management_Mode > > I've never seen any product documentation for motherboards, that > references SMM or what the manufacturer may be doing or not doing. > So whether there is any usage of SMM, is any one's guess. But if > your SMBUS was getting corrupted, and you were absolutely sure > that only Speedfan was running, it would suggest the BIOS > is popping in occasionally, and doing something to the SMBUS. OK, so I'll leave Intel's prog out for now and what do you suggest that I do with SpeedFan and, if necessary, the BIOS? > Paul Thanks again.
From: JD on 28 Oct 2009 14:08 JD wrote: > Paul wrote: >> JD wrote: >> >>> >>> Many thanks again Paul. You are a huge help. >>> >>> Yesterday I spent an hour online looking for a good quality picture of >>> the motherboard and failed, even on the Intel web site. The images were >>> far too small to see any writing on the board. Of course I never even >>> dreamed >>> of visiting Japan to look for it ;-) You are a resourceful guy! >>> Were you >>> thinking of a geisha girl when you visited? >>> >>> Then I opened up the computer and had a closer look. The case fan >>> (only one) >>> is connected to the power supply and not to the board. I looked >>> around for >>> other connections and there were no more than appeared on the >>> schematic we >>> looked at a few days ago. I'll change that case fan connection to the >>> motherboard when I get a plug for the connection and some time. >>> >>> I also looked in the area to the rear of the CPU between the AGP card >>> and the >>> power supply. There is the 12V board supply connector and two other >>> small >>> connectors beside the AGP card. These are for ATAPI and the rear case >>> fan >>> - not installed. >>> >>> I started up Speedfan and it does indeed show 4 fans. In the few >>> minutes it was >>> running it knocked the CPU-core voltage reading down to zero 3 times >>> and each time >>> the Intel monitor popped up a warning. That SpeedFan prog makes me >>> worry. >>> >>> The CPU fan has been running and whining at about 4,300 rpm while all >>> the temperatures >>> are showing 25 C. Dumb design this system. >>> >>> I'm still trying to get through all your post above. Back soon, I >>> hope. Thanks again for >>> your efforts. >> >> If you have readings getting corrupted, they're probably coming from >> the LM85B, which sits on the SMBUS. It means there is a second piece >> of software accessing the chip and monitoring it. That could be the >> "Intel monitor". If one program runs at a time, and you exit the other >> program, you may find the readings more reliable. If the people >> writing the code for those programs could agree on an access method >> to the hardware, there might never have been any corruption in >> the first place. Since nobody in a leadership role, could define >> a semaphore for access to SMBUS, we're stuck with getting crappy >> readings. The only time the SMBUS is guaranteed reliable, is >> when the BIOS is doing its POST routine. > > If you are sure of the that SpeedFan won't damage the system, I'll use > it and leave Intel's program off for the present. >> >> In terms of connections then, you're saying the case cooling fan is >> connected to a 4 pin Molex power connector from the power supply ? > > This is correct - it is connected directly to the computer's power supply. > > Or is >> there a fan header cable coming from the supply, with the smaller >> (motherboard >> style) connector on it ? Some power supplies have a voltage controlled >> two >> or three pin fan connector, and the voltage is set according to how hot >> the power supply is getting. Since the power supply gets a stream of >> air from the computer case, the mechanism roughly causes the case fan >> to speed up, when the computer case air is getting hot. >> >> I presume the CPU cooler is plugged into the CPU header on the >> motherboard. > > Yes, it is. > >> If you're seeing evidence of the LM85B in Speedfan, then you should be >> able to make changes to its registers. If a BIOS or other software is >> also present and making changes, there is nothing to prevent the two >> programs or softwares from getting into a fight. I've had pretty good >> luck here, >> because my BIOSes tend to set the hardware just the once, and then they >> don't do any more writes. It allows the usage of Speedfan, without >> problems. >> >> You might wonder how the BIOS can run, while the operating system is >> present. There is something called SMM, which is a means by which the >> BIOS can take control from the OS, for short intervals. If implemented >> poorly, this can result in detectable issues at the OS level. If the >> outage caused by the BIOS is long enough, you can lose the counting of >> clock ticks (causing system time to drift while the OS is running). >> If the outage is a shorter interval, the system time may remain accurate, >> but the DPC latency (time until a deferred procedure call is serviced) >> may be lengthened. There is one utility for checking that, a >> DPC Latency utility, that can be used to check. For example, on >> my newest system, the average is perhaps 800 microseconds for >> DPCs to be serviced. I may see the occasional millisecond level >> response, if the system is heavily loaded. So if SMM steals >> control of the system for a long period of time, the only >> evidence might be collected indirectly, by measuring DPC >> latency. >> >> http://en.wikipedia.org/wiki/System_Management_Mode >> >> I've never seen any product documentation for motherboards, that >> references SMM or what the manufacturer may be doing or not doing. >> So whether there is any usage of SMM, is any one's guess. But if >> your SMBUS was getting corrupted, and you were absolutely sure >> that only Speedfan was running, it would suggest the BIOS >> is popping in occasionally, and doing something to the SMBUS. > > OK, so I'll leave Intel's prog out for now and what do you suggest that I > do with SpeedFan and, if necessary, the BIOS? > > > Paul > > Thanks again. I forgot to mention that the case fan is connected to the power supply and has only 2 wires and is probably not compatible with the board connection. I guess I should look for a 3 wire fan of the same size and connect it to the motherboard..
From: Paul on 28 Oct 2009 13:52 JD wrote: > > I forgot to mention that the case fan is connected to the power supply > and has only 2 wires and is probably not compatible with the board > connection. > I guess I should look for a 3 wire fan of the same size and connect it > to the > motherboard.. You can connect the two wire case fan to the fan header. It would still get power. The only issue with a connection to the motherboard header, is if the fan draws a lot of current. There are a few pre-built computers, where a fan drawing one ampere is used, and that is too much for a regular fan header. The motherboard manual sometimes tells us the current flow limitations of a header, and so that is something to keep in mind. If the case cooling fan draws somewhere between 12V @ 100mA and 12V @ 350mA, that might be safe to connect to the header. Check what is printed on the hub of the fan. The label sometimes includes power in watts or current in milliamps, from which you can decide whether the fan really should stay connected directly to the power supply. I have one large fan here, on another computer, and it draws close to 1000mA (1 amp). I have it connected directly to the power supply. I don't think the motherboard on that computer, could handle it. A 2 wire fan cannot be monitored for RPMs (at least the normal way via an RPM signal). But you could still vary the voltage to the fan, if the motherboard has control capabilities on that particular fan header. Paul
From: JD on 30 Oct 2009 01:41
Paul wrote: > JD wrote: >> >> I forgot to mention that the case fan is connected to the power supply >> and has only 2 wires and is probably not compatible with the board >> connection. >> I guess I should look for a 3 wire fan of the same size and connect it >> to the >> motherboard.. > > You can connect the two wire case fan to the fan header. It would still > get power. The only issue with a connection to the motherboard header, > is if the fan draws a lot of current. There are a few pre-built > computers, where a fan drawing one ampere is used, and that is > too much for a regular fan header. The motherboard manual sometimes > tells us the current flow limitations of a header, and so that is > something to keep in mind. If the case cooling fan draws somewhere > between 12V @ 100mA and 12V @ 350mA, that might be safe to connect > to the header. Is this oddball name "fan header" the fan connection point to the motherboard? > Check what is printed on the hub of the fan. The label sometimes > includes power in watts or current in milliamps, from which you > can decide whether the fan really should stay connected > directly to the power supply. Yes, I think I'm in luck. The rating is 12v, 0.15A. > I have one large fan here, on another computer, and it draws close > to 1000mA (1 amp). I have it connected directly to the power supply. > I don't think the motherboard on that computer, could handle it. Phew!That's quite a fan. > A 2 wire fan cannot be monitored for RPMs (at least the normal way > via an RPM signal). But you could still vary the voltage to the > fan, if the motherboard has control capabilities on that particular > fan header. I'll look for a connector for that fan. > Paul Thanks again :-) |