4850 VRM Temps?

[knightlag]

Are these temps safe for the VRM on a 4850? I have a DuOrb Installed at the moment, didn't do the hacksaw thing with my stock cooler, though.




Specs:
E4500, P5L VM-1394, 4850 , 2.0 GB G.Skill DDR2 667 RAM, and I forgot the make/model of my PSU

[W1zzard]

how do you know the VRM temperature? the 3 gpu temperatures are different on-die locations

[Andy_007]

Originally Posted by W1zzard

how do you know the VRM temperature? the 3 gpu temperatures are different on-die locations

I have been told that GPU Temp #2 is VRM.

[W1zzard]

nope not true ..

there are 3 sensors on the die (silicon), they are positioned in different strategic location, like shader core or memory controller. i dont know the exact locations and also don't know which number corresponds to which location. i do know for sure that all three are on the die, so inside of the gpu.

you better believe what wiz tell you, its his software and he knows best .

[Andy_007]

Originally Posted by W1zzard

nope not true ..

there are 3 sensors on the die (silicon), they are positioned in different strategic location, like shader core or memory controller. i dont know the exact locations and also don't know which number corresponds to which location. i do know for sure that all three are on the die, so inside of the gpu.

Thats interesting since someone over at XtremeSystems replaced his cooler but didnt put any heatsinks on the VRM's and Temp#2 was getting over 100 degrees and all the other temps where in the 40's.

[knightlag]

I had always assumed that #2 was the VRM temp. Now I feel silly.
Thanks for pointing that out, W1zzard.

[W1zzard]

Originally Posted by Andy_007

Thats interesting since someone over at XtremeSystems replaced his cooler but didnt put any heatsinks on the VRM's and Temp#2 was getting over 100 degrees and all the other temps where in the 40's.

must be some coincidence .. can you link me to that thread? over 100° isnt that uncommon with the stock heatsink

edit: nm xs is down

[Andy_007]

Originally Posted by W1zzard

must be some coincidence .. can you link me to that thread? over 100° isnt that uncommon with the stock heatsink

edit: nm xs is down

XtremeSystems is currently down but once it is back online i will try to find it for you. It just seems strange that GPU temp, GPU temp #1 and #3 are usually all around the same and GPU#2 is usually much different, My 4850 runs 54/52/66/51 under load at 100% fan stock cooler.

[jmke]

^bump, I'm also interested in knowing where the locations are of the 3 temp diodes, is #2 VRM or not?

[tkpenalty]

The 4870 and 4850 are two cards that use different forms of electronics for filtering. The 4870 uses digital phases, while the 4850 uses mosfets, inductors, capacitor phases. I doubt the 4850 has temp sensors on its "VRM" area; since when did mosfets have temperature diodes intergrated? There are only pins for input out put ground, etcetera, unlike VRMs which have a mini "socket" type thing.

The MOSFETs imho > Digital phases since they don't need that much cooling.

The digital phases, absolutely need a heatsink installed as they would die without them.

Originally Posted by W1zzard

must be some coincidence .. can you link me to that thread? over 100° isnt that uncommon with the stock heatsink

edit: nm xs is down

They are talking about the 4870...

[Widjaja]

Yeah there was a big debate on which temperature was the VRM.
Funny thing is someone disagreed with W1zz and what the program he made was reading the temperatire from.

I think your temperatures are right.

The idle of temperature 2 would not have gone down as much as it did if it wasn't being cooled.
My idle temperatures are nearly twice what your are on the stock cooler.

1:-79
2:-84
3:-80

But this is the way the HD4850 series are with stock cooling.
If AMD saw this as a problem they would not have released the card.
My HD4850 is still running like a champ and is by far the best card I have owned.

[nanohead]

Yeah, but........ something is going on with that other temp for sure. I have water cooled 4850s, and temp 1 and 3 are always cool (28-38 depending on load and room temp), but the other one is like 65-80.

I have nothing cooling the power modulation portion of the card, just water blocks on the GPUs and memory. I had some enzotech sinks on them, but they fell off (a true PITA)

I'm not sure what that other temp is for either. It is possible that it has a calibration problem like the main CPU temp on the Gigabyte MA78G 780G series boards, where CPU temp is always like 85-100C, but core temp is always like 25-35. This is clearly a faulty measurement (I have several of these boards in my family's machines, and they all say this)

I'm gonna rig up some heatsink mechanism for the power section, as there is no way I'm hacksawing the stock heatsinks.

It would be nice to understand the mystery of that other temp though. WiZZard, I'd be willing to test or probe. Ya got any debug code

[W1zzard]

Originally Posted by nanohead

It would be nice to understand the mystery of that other temp though.

no mystery .. reseat your waterblock, it is not making good contact with one of the edges of the core. is your waterblock even designed in a way to cool the whole die? or just the center of the die?

[nafets]

Regardless of the electronics and circuitry used between the HD4850 and HD4870, they both use the same RV770 GPU core. So whatever temperatures Gpu-Z is polling, they are going to come from the same place for both cards. It's already been established what temperatures 1, 2, and 3 are (Display IO, Memory IO, and Shader Core). If you're seeing a large discrepancy between the temperatures, chances are very good that the VGA heatsink/waterblock isn't making proper level contact with the GPU core. This problem has been seen numerous times now, here on the TPU forums...

PS: I may be horribly wrong about everything I just said, but it's my educated opinion...

[W1zzard]

Originally Posted by nafets

Regardless of the electronics and circuitry used between the HD4850 and HD4870, they both use the same RV770 GPU core. So whatever temperatures Gpu-Z is polling, they are going to come from the same place for both cards. It's already been established what temperatures 1, 2, and 3 are (Display IO, Memory IO, and Shader Core). If you're seeing a large discrepancy between the temperatures, chances are very good that the VGA heatsink/waterblock isn't making proper level contact with the GPU core. This problem has been seen numerous times now, here on the TPU forums...

PS: I may be horribly wrong about everything I just said, but it's my educated opinion...

nope you are exactly right

[SpookyWillow]

as people have said, the 4850 does not have a VRM temp diode. the 4870 on the otherhand does and that temp can be monitored with one of the latest beta versions of everest.

[W1zzard]

Originally Posted by SpookyWillow

as people have said, the 4850 does not have a VRM temp diode. the 4870 on the otherhand does and that temp can be monitored with one of the latest beta versions of everest.

gpuz 0.2.7 can read those vrm temps too

[SpookyWillow]

dont have a 4870 so i never noticed, only what i read elsewhere :P

[tkpenalty]

MOSFETs have better temperature tolerances compared to 4870's VRMs... therefore higher temperatures are less likelier to cause damage. Atm waiting for Gigabyte to release a non VRM 4870, because VRMs SUCK. May someone enlighten me why they have to use them?

[Darknova]

Originally Posted by W1zzard

gpuz 0.2.7 can read those vrm temps too

Really? All I see is GPU Temperature, GPU Temp.(DISPIO), GPU Temp(MEMIO) and GPU Temp.(SHADERCORE) and VDDC Slave #1-3 Temperature. Which is which? :S

[W1zzard]

everest vrm temp is one of the three slaves i guess

[kfahn]

Just so that everyone knows, I was having this issue of getting temps around 45 degrees C for gpu temp 1, and 100+ for 2 and 3 on my 4850, and after reading this thread, I decided to look again at the way that the sink was seated on my gpu. It turns out that the lower .5cm was exposed below the die. I am using the Accellero Twin Turbo, and after looking at it, it turns out that you can adjust the alignment of the sink, so I lowered it to cover the whole area surrounding the die, and it lowered my temperatures from the high 100's to maximum of 70.

If anyone is interested in getting the Twin turbo, I can recommend it, my temperatures under furmark went from 64/95 to 38/58 with the card at 700/1145 overclock. None of the other gpu temperatures ever go over 80 degrees under load. The fans area at 60%, which is completely inaudible.

[jmke]

keep in mind that with the latest Catalyst 8.8 the system keeps Furmark at bay, lowering temps
http://en.expreview.com/2008/08/26/a...it-run-slower/

[kfahn]

Originally Posted by jmke

keep in mind that with the latest Catalyst 8.8 the system keeps Furmark at bay, lowering temps
http://en.expreview.com/2008/08/26/a...it-run-slower/

That is quite interesting, and I was completely unaware of it. That aside though, Furmark still manages to stress my GPY by about 5 degrees more than ATI tool has ever been able to, so I think that I will stick with it. Any word on a newer version of Furmark that might sidestep this "upgrade"?