HD4870 VRM/GPU Temps

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
Recently I removed the stock cooler on my HD4870 to install the Thermaltake DuOrb. I wasn't looking for lower temperatures, merely a quieter cooler with some sort of LED lighting, as my Shuttle was looking very dark. I used all the heatsinks that came with the DuOrb for the cards memory and put some on the VRM chips as well, but I'm still having issues.

After playing Crysis Warhead for 5-10 minutes the computer crashes. My system is/was fully stable previously, so it's not related to my CPU overclock. After running Furmark for a few minutes my GPU temp is climbing into the mid 90's and my VRM temps are all up around 120c. I sort of expected higher GPU temps (within reason) because the fans on the DuOrb aren't overly fast, but the VRM's are running quite a bit hotter then with the stock cooler.

Is it more likely that the 120c+ VRM temps are responsible for the crash in Crysis or the 90-95c GPU temp?

What I was going to try tonight is removing the VRM and memory heatsinks and installing the red metal plate from the stock cooler that normally makes contact with/cools the memory and VRM chips, and then put the DuOrb back on as well.

I'm hoping that it's merely an issue with the higher VRM (and possible memory) temps and that the stock cooler plate fixes things because I really like the DuOrb but I was looking for some feedback/input in the mean time.

What temps have you guys seen on the GPU/VRM's and still remained stable?
 

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
I just re-ran Crysis Warhead and at the time of the crash the last reported temperatures (in celcius) reported by GPU-Z were:

GPU Temp.(DISPIO): 112.5
GPU Temp.(MEMIO): 120.0
GPU Temp.(SHADERCORE): 116.0
VDDC Slave #1 Temperature: 103.0
VDDC Slave #2 Temperature: 104.0
VDDC Slave #3 Temperature: 101.0

The thing that strikes me as odd is that all temperatures had been recorded at the same or higher temperatures before the crash, so unless it was just the sustained temperatures causing the issues I dunno.

I'm gonna try removing the ramsinks and VRM heatsinks and putting the stock cooler plate back on. The DuOrb blows more air over the cards PCB then the stock cooler did, so theoretically the VRM/Memory temps should only improve if the plate is installed under the DuOrb and at least then I can narrow down what is going on.

I'm still looking for input on the topic but we'll see what that plate does to help.
 

rincewind

Gawd
Joined
Sep 21, 2004
Messages
888
I just re-ran Crysis Warhead and at the time of the crash the last reported temperatures (in celcius) reported by GPU-Z were:

GPU Temp.(DISPIO): 112.5
GPU Temp.(MEMIO): 120.0
GPU Temp.(SHADERCORE): 116.0

VDDC Slave #1 Temperature: 103.0
VDDC Slave #2 Temperature: 104.0
VDDC Slave #3 Temperature: 101.0

VRM temps are okay (if a little high for Crysis, you have Furmark temps there!), but the GPU itself is getting WAY too hot. I'm pretty sure the DuOrb is supposed to be inadequate for the 4870, and this looks to prove that. Stock cooler would be better!
 

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
I'll still try adding the plate for the memory/VREG chips to see if dealing with that heat helps the GPU, but if my GPU temps don't change I might have to go back to the stock cooler.

I don't have room for the TRad2 with a fan in my shuttle, but are there any other coolers that work well with the HD4870 that only dualslot including fans?
 

Digital Viper-X-

[H]F Junkie
Joined
Dec 9, 2000
Messages
14,390
those temps are wayy to high for crysis!
the VRMs are furmark temps too(stress test) and the GPU is way too high man
aftermarket cooling should net you 50-60 load temps maybe 60-70 on smaller coolers, not higher!

I'd suggest taking apart the ATI cooler using the bottom plate for your mem / vrm( i did it with my twinturbo) and using either the stock heatsink with a 120mm fan strapped to cool the plate + sink or the duo orb then, it sounds also like the heatsink didn't get seated properly it shouldn't let your 4870 get to 120
 

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
I'm pretty sure I had decent contact, it might just be an issue of the DuOrb not handling the temps well enough.

Looking around I'm having a hell of a time finding a dualslot cooler that sufficiently cools the cards VRM chips as well as the GPU/Mem.

If adding the stock coolers plate back on the card under the DuOrb doesn't help I'll probably just go back to the stock cooler entirely and purchase some LEDs or a 4" CCFL to make up for the DuOrbs lighting.

Is there any way to stop the stock fan from reving up to full speed on startup and shutdown?
 

rincewind

Gawd
Joined
Sep 21, 2004
Messages
888
Looking around I'm having a hell of a time finding a dualslot cooler that sufficiently cools the cards VRM chips as well as the GPU/Mem

Any which provide enough clearance for the Thermalright HR-09 mosfet cooler (Type-2 with the angled heatpipes) ;)
That thing is reported to keep VRM temps way down.
 

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
Sadly I do not have room for the thermalright mosfet heatsink on the card in my Shuttle with the Ultra-120 eXtreme right next to it. :(
 
Top