Abit NF7-S Socket temp v. Diode

[craneage2001]

Well, I've been having temp issues with my RBX setup which I assumed were from a bad mounting...I pulled off the block and the AS5 was evenly compressed over the entire core and it was just a thin layer left as it should be. (a nice imprint was left on the block when removed). I placed it back on using the method of "just placing a dab in the center of the core" method and my temps are a hair lower at ~38°C load (CPU Burn). I was previously getting ~42°C after Prime-ing for a while, so I imagine CPUBurn would have hit 44-45°C or so. Do you think my temps are that high or is it just a discrepancy from the socket vs. diode temps? I've also heard Abit's NF7s run a bit high in the temps...any thoughts on that as well? (I'll ask that in the Abit section).

[craneage2001]

Here's the story (I did some testing...):

1) I left my PC off for a hour or so with my pump and rad fan running. My water was fully back at room temp.

2) I start up my PC from cold and immediately enter BIOS (~5 seconds on). I immediately go to the temps section and the CPU is at 32°C (This is within a maximum of 10 seconds of hitting power). I sit and watch the temps and it is a full 30sec before it goes to 33° and another 30sec to hit 34°. I exited BIOS and and boot into windows (~20secs) and it's still at 34°.

Is it even possible for the cpu to jump to 32° within the 5 seconds from powerup? I find that hard to believe---for it to jump that fast (especially since it's taking socket temps, not die) without having my load temps be insane. If I subtract 10°C from these temps, my bootup temps are just above ambient (which seems logical) and my load temps coincide with what the approximate values should be for my block (~.08 C/W, 100W heat from proc). Could I be right?

[ninety6]

Yes. I sometimes get the same thing on my NF7-S + 2500+ at 1.87V. Don't trust the temp on the ABIT boards. They're all wack :(

96

[SysCrusher]

Originally posted by craneage2001

Is it even possible for the cpu to jump to 32° within the 5 seconds from powerup? I find that hard to believe---for it to jump that fast (especially since it's taking socket temps, not die) without having my load temps be insane. If I subtract 10°C from these temps, my bootup temps are just above ambient (which seems logical) and my load temps coincide with what the approximate values should be for my block (~.08 C/W, 100W heat from proc). Could I be right?

I believe your right. If it's the socket temp, i bet the thermistor isn't even touching the bottom of your cpu. If it was taking it from the die, you would be able to better judge.(sp?) The socket thermistor is just reading heated air that is under the socket. I hover around 30C with an air temp of 25C. That's with three blocks in the loop also, although parallel to my cpu block.

How high do you get on load? Also, when I didn't use the socket hold down I had my springs fully compressed.

[craneage2001]

I hit 40-41°C load while running 2500mHz @ 1.7v (91 watts of heat).

Also, when I reseated my block, I compressed the springs to just about full, they're only probably about 1 turn away from full compress.

[craneage2001]

Come to think of it, the thermosister may actually be touching the bottom of the CPU...I recall that it was a 'tab' variety in the CPU cavity, but I don't recall just how high it stood...

[SysCrusher]

What is the air temp? 40-41C isn't bad on load. I'v noticed that once the AS sets in, the temps get better. It's hard to judge without air temp though. Where is your rad mounted? Does it have decent airflow?

My cpu diode tells me 10C above ambient with load being 2C more. Using a temp thermistor I get 5C above air with load being 2C more. Trying to read temps this way isn't best but atleast you can see a change. Motherboard temps are notorious for being wrong.

[I.M.O.G.]

Using an external thermistor is just as bad as mobo temps... The problem with each is that you have no idea how much error is in either reading. If you use an external thermistor, you are going to get a composite reading - that is, you will be reading the temps of the CPU, the air, the tape holding the sensor in place, etc. and your reading will reflect the average of all those.

You just can't get good temp readings. Even if you are reading an internal diode inside the CPU, the reading depends on where the manufacturer placed the diode - if they placed it close to the cache it could read 10C cooler than if it were in the hottest part of the chip.

You asked if it were even possible for the temp of the chip to jump 10C within 5 seconds of bootup? Let me ask you this, how long would you expect the chip to live if you had no heatsink on it?

Here's a good old story... a few years back before I knew anything about computers I used to hang out with this older kid who knows tons of shtuff about computers. He was working on some old systems and he had the heatsink off of one of the chips. I asked how quickly the chip gets hot without a heatsink and he said put your finger on it... Then he started it up. My finger wasn't there long at all. That was probably a 100-300 mhz CPU.

[SysCrusher]

Originally posted by IMOG
Using an external thermistor is just as bad as mobo temps... The problem with each is that you have no idea how much error is in either reading.

That's what I was trying to get at. I use both as I can see a change better with a thermistor. If the air temp is cool enough, I don't even see a change between idle and load using the diode. Then it can fluctuate greatly. It's almost annoying. The diode temp is worse than the thermistor on my board.

If anything their good to see a change and that's about it.