Hello!
I have a problem with my ATI5850 reference PCB:

My Accelero s1 rev2 heatsink +12cm fan and a VRM-R4 heatsink can take all the temps under 55° with no problems.

But there is a VRM that no heatskins can cool down, i'm talking about MemIO, the VRM that manage the memories (that's what an internet site says)

as you can see is not in a good spot, the not-reference PCB is better (5850 non reference is 5870 reference)



that's what i did:



i cut the original cooler so i can use the accelero, the vrm-r4, and the last piece of the original one to cool that VRM



there is a thermal pad between them



now the GPU is @54° under load, the VRMs are @50°, but that small VRM still reaches 70° while playing and 125° under extreme furmark burning test, with a Tamb of 27° and all my case fan at 1200rpm (5x 12cm in an Antec p180)

Do you have a solution? Is there a cooler i can use?
I'm cutting an old NB cooler for that VRM but is not easy.
Thanx!

How are you getting the temp of that VRM?

I'm also interested as to how you get that temp, in GPU-Z I only get 7 temps;
GPU Temperature
GPU Temp #1
GPU Temp #2
GPU Temp #3
VDDC Phase #1 Temperature
VDDC Phase #2 Temperature
VDDC Phase #3 Temperature

I have a very similar setup to you, reference 5850, Accelero S1, Thermalright VRM-R4 with a 120mm Nexus blowing onto it, I only have one case fan though, not five. Instead of cutting up the reference cooler I cable-tied an old RAM-sink onto the "bastardo" chip (see pic).



All the temps I get in GPU-Z are fine, and have no problems with stability in Furmark or games. Are you having stability issues or are you just being cautious?

I put an AC Twin Turbo Pro on my Sapphire 5850, and used the original stock cooler metal backplate as my heatsink. It's the ONLY heatsink I'm using for the non-gpu-itself chips.

GPU-Z is reporting (after running a few days, and idle in Win7):
GPU Temp: 45C
GPU Temp1: 45.5C
GPU Temp2: 47C
GPU Temp3: 44.5C

VDDC Phase 1: 54C
VDDC Phase 2: 53C
VDDC Phase 3: 54C

After 20 min in Fallout 3, all those temps only go up a few degrees.

Hello, sorry if i'm writing my reply so late, i tried a lot of thermal pads to solve this problem, let's start answering:

@FJC and BPD:
in GPU-Z
GPU#1 is the Core
GPU#2 is MEMio
GPU#3 is arnyalo
VDDC #1,2 and 3 are the VRM "in line"

i can say this because Everest shows me the name of the sensor, so i compared the temps.

@BPD:
as i wrote, the small VRM reaches over 70° while playing and my system crashes @~95°, i tried with furmark extreme test with max AA.

after removing the original pad i used an adesive Akasa thermal pad between the VRM and the original Sapphire plate and it lower it's temps of ~10°, but it still reaches 80° during benchs.

i'm sure if i use a little cooler like your i'm going to burn it, but i'll try when i come back home

months ago a bug of MSI After Burner increased my VRM voltage to 1.6 for a week, did it damage the small VRM?

is the same problem of the old 4870 LINK

An user of another forum has the same issue that i solved by half, using a VRM-R4 for VDDC#1,2,3, but GPU#2 and 3 are still high... LINK

Hmm, I don't know what to suggest then. As I said, I have the same VGA cooler as you, 120mm nexus @ 1000rpm blowing on it and one case fan. My temps for GPU#2 are FurMark-80°, gaming-60° (maximum over 3 hours). Just with that silly little RAMsink on it, and no stability problems.

today something happened, i used an EK Water Blocks Thermal PAD between the sapphire plate and MemIO instead of Akasa AK-TT12-80 Thermal Adhesive Tape, now it's temp can't raise over 60° under Furmark fullscreen 1680*1050 AAx8 Extreme mode with a 890/1130 overclock...

i hate this card dude, i hate it

Stop using Furmark to record temperatures and live happily, that's what I did.
Crysis gets my 5850 highest temps around 20° lower than Furmark or OCCT. If that game doesn't warm my card up that much, I'm pretty sure I'm fine, as no man could create a game that will do it!