How can a cpu temperature jump 10 degrees in 1 second?

[Ymmy]

I used to have an epox 8rda+ with an xp2100 o/c to 2ghz with an slk800 and 80mm case fan for hsf. well, that mobo died (*weeps*) and now i just bought an asus a7n8x-x as a replacement.

before, under load, the temperature difference was from 5-9 degrees, while the ambient was usually a 1-2 degrees diference. now, the ambient is more or less the same, but the temp under load is wacky. from 29 degrees, it would go to 30 and then jump to 38 in under a second. then from 38, it'll gradually climb up to 41 or something, and then spike to 50. then a few seconds later, it would jump back to 43 or whatever. these numbers are all arbitrary, but the spikes are truly that sporatic.

right now, there is a 13 degree difference under load with the cpu at 43 and the case at 31 (it's cold in my room). i'm measuring all of this with the latest mbm and using artic silver 5.

i have tried reseating the hsf on the cpu. the as5 is thinly spreaded across the cpu. Any clue to what's wrong?

 

[Yanagi]

seems like your mobo does bad readings or you havent set mbm up right. check in bios if the temps fluctuate as much.

 

[Ymmy]

Originally posted by: Yanagi
seems like your mobo does bad readings or you havent set mbm up right. check in bios if the temps fluctuate as much.

mbm has the auto motherboard setting thingy where i just chose a7n8x-x and it does its own thing. i can't check the mobo's temp cause the temp only fluctuates rapidly when it's under load.

 

[mechBgon]

The mass of the CPU core is very low, maybe 5 grams. When its heat production abruptly swings from ~30W to 60W+ under load, why do you expect it would take a long time for its core temperature to change on a 5-gram piece of material? It is not logical, Captain.

(MBM5 can extract actual core temperatures on the A7N8X's, btw, as opposed to slow-reacting socket-thermistor temps like on most boards)

 

[Macro2]

Mech is right. Doesn't take long to heat yee of little mass.

 

[Ymmy]

Originally posted by: mechBgon
The mass of the CPU core is very low, maybe 5 grams. When its heat production abruptly swings from ~30W to 60W+ under load, why do you expect it would take a long time for its core temperature to change on a 5-gram piece of material? It is not logical, Captain.

(MBM5 can extract actual core temperatures on the A7N8X's, btw, as opposed to slow-reacting socket-thermistor temps like on most boards)

so i take it you have an a7n8x-x. this is the "x" version that i have, not the deluxe. does your temperature jump like that also?

 

[mechBgon]

I've had an A7N8X-X briefly and I have an A7N8X Deluxe right now in my home system.

I don't do temperature monitoring on an ongoing basis, but I've tried MBM5 on the A7N8X Deluxe for the heck of it, and it does indeed get real-time temperature jumps and drops if you pick the proper sensor. By contrast, a board that uses a socket thermistor is making an indirect reading, and you can expect that indirect reading to ramp up & down over about a two-minute interval (based on the testing I did with my A7V333-RAID, a socket-thermistor board).

 

[Operandi]

Originally posted by: mechBgon
The mass of the CPU core is very low, maybe 5 grams. When its heat production abruptly swings from ~30W to 60W+ under load, why do you expect it would take a long time for its core temperature to change on a 5-gram piece of material? It is not logical, Captain.

(MBM5 can extract actual core temperatures on the A7N8X's, btw, as opposed to slow-reacting socket-thermistor temps like on most boards)

I'm not claiming to be an expert but I don't think it works like that. While the CPU die is small all that heat is transferred immediately to the much larger mass of the heatsink, large temp jumps shouldn't be possible. That?s why athlons burn up within 1-2 seconds with out a heatsink.

I think most likely it?s just a sensor error/fluctuation.

 

[mechBgon]

Originally posted by: Operandi

Originally posted by: mechBgon
The mass of the CPU core is very low, maybe 5 grams. When its heat production abruptly swings from ~30W to 60W+ under load, why do you expect it would take a long time for its core temperature to change on a 5-gram piece of material? It is not logical, Captain.

(MBM5 can extract actual core temperatures on the A7N8X's, btw, as opposed to slow-reacting socket-thermistor temps like on most boards)

I'm not claiming to be an expert but I don't think it works like that. While the CPU die is small all that heat is transferred immediately to the much larger mass of the heatsink, large temp jumps shouldn't be possible. That?s why athlons burn up within 1-2 seconds with out a heatsink.

I think most likely it?s just a sensor error/fluctuation.

I had a nice reply to this, and my dial-up connection died and ate it Basically, because the thermal interface between the core and the heatsink is relatively poor, there is going to be a significant temperature delta required to drive the CPU's heat across that interface to the heatsink. If the CPU's heat output doubles, then yes, the CPU core can jump in temperature significantly and quickly on its side of the relatively-high-resistance interface. When the delta is big enough to drive the additional "load wattage" across the interface, the CPU's temperature stabilizes. As the heatsink itself begins to rise in temperature a bit from the higher power input, you see the CPU move up a little more to maintain the delta that drives the wattage across the interface to the heatsink.

That make any sense?

 

[Operandi]

Yeah, that dose make sense. Just that 10c seems like alot to cover in a second, but I have noticed similar rises wiht my Athlon XP/ Aopen board more like 6-7f though.

dial- up? i'm sorry.

 

[VirtualLarry]

Can MBM even be set up to update every second? I thought that the minimum interval was 5s or 10s or something.
But either way, yes, a tiny little chip like that, that's drawing so much power, can easily have internal core temp slew rates of 10C/s.

 

[Sonic587]

Originally posted by: VirtualLarry
Can MBM even be set up to update every second? I thought that the minimum interval was 5s or 10s or something.
But either way, yes, a tiny little chip like that, that's drawing so much power, can easily have internal core temp slew rates of 10C/s.

Goto General Options and look for "MBM 5 should work at an interval of". It can go as low as 1 second updates.