CPU temp and fan speed

[ratamacue]

Hello, I just put together a new system for home use and I have a few questions. The components are as follows:

Gigabyte GA-7VRX
AMD Athlon XP 1800
Antec 660AMG (330W)
Antec Reference CPU fan/heatsink

1. The CPU fan is designed to be plugged into a standard 12V 4-pin connector. The fan moves a lot of air, but at 12V it's as loud as a small vacuum cleaner. So I bought one of those 12V to 7V converters that work by using the 5V wire as ground, resulting in a 7V current for the fan. As expected, the fan is quieter but moves less air. With the 7V current, my average CPU temp is around 115 deg F, and the average system temp is around 102 deg F. With the 12V current, the CPU temp averages 105 deg F, system temp 98 deg F. That's a big difference, but I consider it a good tradeoff for a much quieter box. Am I doing anything technically wrong by using the 7V current for the CPU fan? I have been monitoring the temps closely.

2. Oddly, with the CPU fan plugged into the 12V/7V converter, and the CPU fan sensor wire still plugged into the motherboard, I am now unable to monitor the CPU fan speed (it reads 0 RPM). Why is this? Using a straight 12V connector, the sensor works fine and the CPU fan runs at about 7000 RPM.

3. Because of the monitoring problem, I have opted to scrap the 12V/7V converter and run the CPU fan off one of the "fan only" connectors which are speed-controlled by the PSU for quiet operation. Now the CPU fan sensor works again, and it runs at about 3250 RPM. It seems to move less air than with the 7V adapter, and certainly less than with the straight 12V. My temps are still hovering around 115 deg F CPU / 102 deg F system. Is this technically OK? I figure as long as the temps stay relatively cool, it's not causing a problem. Would anyone happen to know what range of voltages these fan-only connectors use? I couldn't determine this from the available documentation.

Thanks for your help!

 

[OulOat]

Your CPU temp are 46.11 C at 7V, which is fine temp for it. Not the best for ocing, but certainly pretty good for normal operation. Thing is, if your system is stable at 70 C, then 70 C is fine too. Yeah, next time use metric, makes comparisions easier

 

[Zincq]

I've almost got the same setup:

Giga-byte 7VRXP
XP 2200+
Thermaltake Volcano 9
4 case fans - 3 80mm, 1 60mm.

My CPU runs ~53°C idle, and ~57°C full load. System temp is at a "nice" 45°C. My system is stable, but I'm still concerned that when the temp ever gets up into the 60's the CPU will overheat. However, I have read that Athlon XP with core revision A, meaning all those XP's up to 2200+, run quite hot. Also, my system does run rather loud - but if I turn down the fan speed on the Volcano 9, then the only humming noise coming out of the system is the case fans.

 

[ratamacue]

Thanks for the replies! Does anyone have any idea why the CPU fan sensor would be working fine with the normal 12V power connector OR the fan-only power connectors, but NOT the 7V converter?

 

[Smilin]

To get fan speed reported to bios you gotta break the yellow wire off and run it to the motherboard header.

I found this nifty product I wanted to tell you about...Thermaltake makes a High-med-low switch for fans that mounts in a PCI cover slot so the switch is on the outside. I'm using it for my CPU and it works great. I've got a dragon-orb 3 cooler that runs at about 7200rpm normally and is as loud as a jet. With the switch set to medium or low the sound is quite acceptable. It has a splitter that runs to your motherboard so your fan speed can still be read.

you can buy em for about six bucks I think.

 

[OulOat]

Only thing I can think of is the 5 V drop.

 

[jaybee]

Originally posted by: ratamacue
Thanks for the replies! Does anyone have any idea why the CPU fan sensor would be working fine with the normal 12V power connector OR the fan-only power connectors, but NOT the 7V converter?

Sure. You've tricked the fan by substituting 5V for ground. So the fan only receives 7V, which is great for underpowering it, but confuses the heck out of the fan monitoring. Since the monitoring and power circuits share a common ground, you've tricked both the sensor and the fan. Your motherboard doesn't understand the funky signal being returned by the sensor now.

Regarding running the cpu hsf from the ps-controlled headers, I wouldn't do it. Those headers are for case fans; when the case (power supply in this case) gets hot, they speed up. When it cools, they slow down (or even turn off possibly). That works for cooling the case, but doesn't carry over very well to the cpu. Your power supply might be very happy with the temps it's reading, while at the same time the cpu hsf is running at low speed and allowing the cpu to run too hot. If there's minimal air circulation over the cpu, that heat has little means to disperse -- meaning the cpu may get too hot, but also that most of the heat won't circulate to the psu sensor. In other words, if the hsf and case fans are running a very low speed, heat from the cpu will have a tough time getting to the psu sensor and the psu won't realize how hot the cpu is.

There are several solutions for your quiet or adjustable cpu cooling:

1. Buy a temperature-controlled fan. Install it on the cpu hs and test different locations for the sensor, finding a good balance between quiet and cooling. This would take some experimentation, but it's far preferable to a psu sensor controlling your cpu hsf. Also eliminates any manual control.
2. Buy a speed-adjustable fan. Crank it up when playing games, slow it down at idle. Kind of hit-or-miss -- how do you know what the correct amount of cooling is? I haven't found any way to get MBM to appear in the corner of my game screen (short of dual monitor setup)
3. Install a resistor on your cpu power line. This should allow the rpm sensor circuit to function normally. How do you know what resistance to use? Take a look at the fan rating; it will be 12V,x A. The circuit impedance will be 12/x. Let's say x is .25. Impedance=48. If you want the fan to run at 80% of its rated speed, you would want to cut your current (x) to .2A, meaning you would have to add 12 Ohms resistance to the circuit. You won't find that resistor, but settle for 10. Try to get a 1W (max power) resistor. P=I²R, where R is the resistance of your resistor (in this case) and I is the current (x above). This works on those noisy case fans too.
4. Make a quiet switch for your computer. This can be combined with the above if your fans are too noisy at full blast for even games. This is a lot easier than it sounds. I found a dual circuit switch at Radio Shack that I run my hsf and a case fan through. The "on" position is full power; "off" adds 20Ohms to the hsf circuit and 10Ohms to the case fan. When the machine is idle, I flip it off. Good for reducing noise when I'm not using the machine, but more importantly, lets me sleep while I download. The easiest way to do this is by running the grounds through the switch, with one output to ground and one to ground +xOhms. If you don't want to break the rpm sensors, run the power lines through. Little more work though.

That's enough for now. LMK if you have any ?'s.

jaybee

 

[ratamacue]

Thanks for the very informative reply jaybee. Looks like what I really need is one of those rheostat things or a variable resistor for the CPU fan.

 

[ratamacue]

Until I install the "clean" solution, the fan-only power connectors seem to keep the CPU cool enough. Last night I did a "ping -f localhost" for about 15 minutes straight (which generates 100% CPU load) and compiled the kernel a few times and then rebooted. After that, the BIOS was reporting 48 deg C for the CPU temp, 40 deg C for the ambient temp and that seems OK to me (not so far off my idle temps).

 

[jaybee]

Originally posted by: ratamacue
Until I install the "clean" solution, the fan-only power connectors seem to keep the CPU cool enough. Last night I did a "ping -f localhost" for about 15 minutes straight (which generates 100% CPU load) and compiled the kernel a few times and then rebooted. After that, the BIOS was reporting 48 deg C for the CPU temp, 40 deg C for the ambient temp and that seems OK to me (not so far off my idle temps).

My recommendation for cpu load testing is Prime95. It'll max out the cpu, giving highest load temps as well as indicating if you are overclocking too far (it'll return errors). Ping is going to be bottlenecked by your NIC, so I don't think you'll hit 100% load. Pull up Task Manager & check the cpu load if you like.

jaybee

PS If you want a cheap, quiet cpu hsf, try the SVC GC68. $6, very quiet, and it keeps my AXP (1.7GHz, 1.81v) under 60C load (measured at the socket). Not great, but good enough.

 

[ratamacue]

Hi jaybee, what I was doing was to flood ping the loopback interface (not the ethernet interface) which doesn't concern the NIC. According to top (I'm running Linux) it does indeed generate 100% CPU load. I'm not sure this is a viable method for testing because ping is a very simple operation. This is why I also compiled the kernel a few times on top of everything else.

Just did a quick search on freshmeat.net which yielded these 2 load testing programs:

Link 1
Link 2

When I get home from work today I'll give them a shot.

 

[jaybee]

So far your temps look fine. Try cpuburn for about 30 min. If you can stay under about 55-60C, you're in good shape. I wouldn't worry about the fan rpm sensor; since you have a temp monitor, you should be able to put an overheat alarm on it.

jaybee