GPU Temperature #1, #2 and #3, what do they mean for the 4800 series?

[EastCoasthandle]

Does anyone know what:
GPU Temperature #1
GPU Temperature #2
GPU Temperature #3
actually means for a 4870? Did I overlook/miss a thread/post that already discussed this?

[dark2099]

I think they are the GPU, MEM, and VRM temps. I think erocker knows, obviously w1izzard.

[EastCoasthandle]

Originally Posted by dark2099

I think they are the GPU, MEM, and VRM temps. I think erocker knows, obviously w1izzard.

Thanks for the input

[erocker]

They are all off of the GPU.

[EastCoasthandle]

EDIT:
SO
it's GPU #1
Mem #2
VRM #3

[W1zzard]

no ... there are three sensors inside the gpu die at different strategic locations

[tomatomanz]

Originally Posted by W1zzard

no ... there are three sensors inside the gpu die at different strategic locations

So.. there are not any of this as below?..
GPU #1
VRM #2
Mem #3

I gotta say.. the GPU Temp. #2 seems to be the highest temps...

Cheers :P

[W1zzard]

the locations are display io, memory io and shader core. all those locations are inside the gpu in different areas

[ShadowFold]

I thought 3 was the vrm's cause if I dont have a HS on them they get really hot.

[Wshlist]

In another thread a guy installed his VRM heatsink of his new coolersetup wrong and the GPU#2 temps went up to 117C or thereabouts then he did it right after people advised him and they went down to half that, so in short: these sensors are NOT all in the GPU, at least #2 is the VRM area.
I think it was on a 4850 in the thread though but obviously it does show those sensors aren't all in the GPU, but since the 4870 has a different layout and VRM setup I'm not sure which temp is monitoring what parts of the card.

Perhaps a can of compressed air blown on various part of the card while it's dong some work and while monitored by GPU-Z will help show what's monitored by which sensors.
(I'd say cooling spray but sudden freezing of areas might not be healthy).

[blaher]

Originally Posted by ShadowFold

I thought 3 was the vrm's cause if I dont have a HS on them they get really hot.

I'd have to agree, as you can see by my results here:
http://forums.techpowerup.com/showthread.php?t=65333

I'd say both sensors #2 and #3 are at the back of the card, where the VRMs are.

[W1zzard]

then the amd engineers have wrong information

[Widjaja]

So 1, 2 & 3 are not generally set to say 1=core, 2=VRM, 3=memory?

So they are just sensors in random strategic areas of the GPU?
And the location of these sensors are anyones guess?

All I know is 1 & 3 are usually around the same temp while 2 is always higher.

Originally Posted by erocker

They are all off of the GPU.

Well duh

[Wshlist]

I repost: they are NOT all on the GPU ide, since people that had no or bad cooling on their VRM's (at least on the 4850) found that #2 went very high and came down when the vrm was properly cooled for instance, that much is obvious.
Please stop insisting spreading (false) information without having any real info to base it on.
In other news for the girls: you can't get pregnant from toiletseats

[tomcug]

Yes, I had very high #2 temperature without heatsink on VRMs. So I think they aren't all on the GPU die too, because it isn't possible that different areas of the core can differ in temperature by over 40 °C.

[W1zzard]

Originally Posted by Wshlist

I repost: they are NOT all on the GPU ide, since people that had no or bad cooling on their VRM's (at least on the 4850) found that #2 went very high and came down when the vrm was properly cooled for instance, that much is obvious.
Please stop insisting spreading (false) information without having any real info to base it on.
In other news for the girls: you can't get pregnant from toiletseats

i just did a quick test ..
powercolor 4850 with that round zalman cooler and metal plate on vrm
temps in idle in gpuz are 44, 51, 43.5
removed the little vrm heatsink while the system is running
temps in idle in gpuz without vrm heatsink are 44, 51, 43.5

so removing the vrm heatsink has no effect on these temps which means that neither of these 3 is vrm temp

will test with load next

[W1zzard]

load without vrm heatsink:


load with vrm heatsink:


load with vrm heatsink and jammed fan:


Please stop insisting spreading (false) information without having any real info to base it on.

[iamollie]

W1zzard saves teh day!

[Wshlist]

Perhaps it's a case of the 4870 and 4850 being different, or perhaps w1zzard is simply lying because he's having issues with me.
Yeah now go abuse your power and ban me and feel you made a big win, but don't forget to boast about it too.

[Urbklr]

I believe W1zz, but I don't at the same time. I had ran my VRM's passive, and then added a 40MM fan, adding the fan lowered my temp#2 a decent amount. I'm confused

[EastCoasthandle]

Thanks for the clarification W1z. Could you update GPU-Z with the same information as in your screenshots? Also, is there a vrm temp that you could add to GPU-Z? I ask because Everest has indicated a vrm temp (not sure if it's accurate or not).

[tigger]

@Wshlist

If thats what you think of w1zzard and this forum,then why dont you try another forum.

[W1zzard]

wshlist why should i lie ? but if my own testing strongly indicates a certain result i have to believe in what i see. if someone else wants to do the same experiment, be my guest. if someone has any suggestions on what i did wrong or what else could be tested, let's hear it

new gpuz version will have vrm/voltage monitoring on 4870
new gpuz version will have the labels as displayed in my screenshots above

[W1zzard]

Originally Posted by Urbklr

I believe W1zz, but I don't at the same time. I had ran my VRM's passive, and then added a 40MM fan, adding the fan lowered my temp#2 a decent amount. I'm confused

so the fan blew fresh air all over the card?

[Urbklr]

Originally Posted by W1zzard

so the fan blew fresh air all over the card?

No, that's the 120MM's job. That little fan doesn't have enough airflow to effect the Die's temperature at all.