what does GPU temperature #1, #2, #3 mean?

[moriz]

hi there,

i've recently reapplied thermal paste on my MSI HD 5850 (reference board), and i've noticed an increase in GPU temperature #1 (about 6C) and a sharp drop in temperatures #2 and #3. my question is, what are they exactly? as in, what's #1, #2, and #3 actually measuring?

 

[moriz]

so... nobody knows? great...

so does anyone know why the temperatures would change like that? as in, #1 rises, while 2 and 3 drops?

 

[trickson]

Post up A screen shot for us .

 

[moriz]

here you go. notice how GPU temperature #1 is much higher.

 

[trickson]

Look like GPU #1 is the GPU core temp .GPU #2 would be ( I am guessing here ) the shader temps and GPU #3 would be the RAM temps ( Again guessing on this one as well . ) .

 

[moriz]

sooo... why would reseating and reapplying thermal paste increase GPU core temp, but cause the shaders (which are part of the core) to drop? this doesn't make any sense.

 

[trickson]

sooo... why would reseating and reapplying thermal paste increase GPU core temp, but cause the shaders (which are part of the core) to drop? this doesn't make any sense.

Maybe you did not reset the HSF on right or maybe it will just take some time for the thermal past to set in . I would have to say it may take some time for the new thermal paste to set in .

 

[HammerON]

If you place your cursor over the description it will tell you the difference. Mine states one is "as reported by the graphics driver" and another one states "as reported by the ADT7473 sensor chip".

 

[moriz]

mine just says "exact location unknown" for all three.

anyway, if i didn't put the HSF on properly, wouldn't that just cause a global rise in temperature, instead of the one up and two downs? the card's been running for about a day now, so it SHOULD be set already.

 

[trickson]

mine just says "exact location unknown" for all three.

anyway, if i didn't put the HSF on properly, wouldn't that just cause a global rise in temperature, instead of the one up and two downs? the card's been running for about a day now, so it SHOULD be set already.

Yes . Maybe you did not put enough on or some thing. You can try doing it again and seeing if that helps or maybe some thing is just now showing up wrong is all . Hard to say really .

 

[Magikherbs]

What method did you use to spread the paste ? I always 'finger spread' it.

Even though the description isn't there, HammerON is right. The higher reading is usually the driver which has an offset built in.

 

[trickson]

What method did you use to spread the paste ? I always 'finger spread' it.

Even though the description isn't there, HammerON is right. The higher reading is the driver which has an offset built in.

http://gpuz.techpowerup.com/11/01/21/gzk.png

I apply a small amount onto the GPU ( or CPU ) then I place the HSF/Wb on the thing and press firmly and evenly . but e all have our own ways of doing things .

 

[Magikherbs]

I apply a small amount onto the GPU ( or CPU ) then I place the HSF/Wb on the thing and press firmly and evenly . but e all have our own ways of doing things .

Its your boat lol.. float it !

By pressing down on HSF, you maybe speading it too thin. I let it spread itself while I tighten all the screws.

EDIT
@moriz How did the original paste look when you first removed the HSF ? I know your temps will drop if you give the chip full coverage.

http://www.techpowerup.com/printarticle.php?id=134

 

[moriz]

the original was basically overflowing.

this is actually my second attempt. the first time, i used the credit card method and spread a very thin and even layer over the GPU, which resulted in VERY high temperatures. so i took it apart again and applied another drop, which decreased temperatures noticeably.

btw, what did you mean by "offset"? anyway, the fan is clearly linked to #1, since it's spinning up more than before as temperature rises.

also, is there a way to separate the copper heatsink portion and the plastic shroud? i can probably do a better job on contact if i can reattach them separately.

 

[Magikherbs]

the original was basically overflowing.

this is actually my second attempt. the first time, i used the credit card method and spread a very thin and even layer over the GPU, which resulted in VERY high temperatures. so i took it apart again and applied another drop, which decreased temperatures noticeably.

btw, what did you mean by "offset"? anyway, the fan is clearly linked to #1, since it's spinning up more than before as temperature rises.

also, is there a way to separate the copper heatsink portion and the plastic shroud? i can probably do a better job on contact if i can reattach them separately.

Temp offsets are there to help safeguard your system against faulty sensors and/or HSFs, from what I know.

Depends if the shroud is built that way. Doing that may cause the HSF to slide or break its seal and you risk damaging the bare/dry area on your gpu.

 

[moriz]

this is a reference board 5850, so yes, the shroud does come off, though i haven't figured out how just yet.

 

[Magikherbs]

this is a reference board 5850, so yes, the shroud does come off, though i haven't figured out how just yet.

The GTX 280 I recently bought and sold had part of the shroud, around the fan area, clamped on to the HSF. It was a bit of a pain but not hard to figure out.

Are you sure the ram and VRM thermal pads still line up ? What does your case/fan setup look like ?

 

[JousteR]

To take off the fan look on the underside and you'll see tiny crossheaded screws these remove the plastic shroud and let you have the heatpiped hsf for cleaning/hoovering jobs that may be needed.

 

[silkstone]

On a 4850 those temps are for #1- DISPIO #2-MEMIO 3#- SHADERCORE, i imagine it will be something similar for your card.. and if you applied the thermal paste and heatsink correctly they should be within 5deg of each other.
It took me multiple applications to get my heatsink seated evenly, the trick is to count the turns per screw and work your way slowly around the heatsink in order to get even pressure.

 

[W1zzard]

On a 4850 those temps are for #1- DISPIO #2-MEMIO 3#- SHADERCORE, i imagine it will be something similar for your card

that. but since i couldnt find anyone at amd to confirm or give further info i rather treat them as unknown locations. they are physically on the gpu die (not memory chip temperature)

 

[moriz]

so i just reseated the HSF again, and temperatures have dropped back to acceptable levels. apparently, the secret is to just use quite a bit more thermal paste than what i'm used to.

anyway, thanks everyone for your help.