4870X2 water block issues

captain_d0nkey · 30 Nov 2010 at 10:29

Hi everyone, my 4870x2 is currently giving me a serious headache, I bought it second hand and immediatly wanted to water cool it, I checked that it was working by playing a few games and I had no problems. I then modded the 4870s heatsink so that I could fit two zalman gwb3 water blocks to it.

Once everything was put together I plugged it into the loop and straight away got serious artifacts I tried cleaning my drivers and installing the new drivers but the card still showed big white lines when in bios and windows.

I put the stock heatsinks back on without the shroud and turned the pc on, artifacts were still showing but because there was no shroud on the card overheated so I shut down the computer and placed a fan on the heatsinks, started up again and everything was fine and games played perfectly.

So thinking it may be the zalman blocks I decided to stop being cheap and bought a full coverage xspc razor water block. Same thing occurred with this block. I loosened the screws around the edges and a bit on the inner gpu screws and it booted in fine and went into windows fine. But 10 minutes into playing a game the pc freezes. The screen goes black and the computer restarts.

So my question is, is this a new problem now such as an old bios or maybe wrong ram voltages?

Also just to let you know when I test default clocks using the custom clock test on CCC it fails. I haven't tested it with stock HSF on.

Thanks in advance

Scougar · 30 Nov 2010 at 10:55

Have you got thermal pads and confirmed that your waterblock is actually making contact with the chips? You can check by doing a test fit with the cooler, and seeing if the paste spreads out properly :-)

(you are using paste right?)

VERY important is that you also ensure you are cooling the VRM chips and have thermal tape on those otherwise you are asking for trouble.

captain_d0nkey · 30 Nov 2010 at 11:24

I've heard that these blocks aren't the best at cooling the vrms, but I've read the temps and I think I recall that the vrms at idle were at 54 degrees. Thing is also I can't tighten the card down completely because it causes artifacts.

When the pc crashed while playing dirt I was looking at the card leds and the 3rd one (overheating led) didn't light up either.

{SAS}TB · 30 Nov 2010 at 11:29

My money is on VRMs - ive had 5 of those cards, 3 being under water and the VRMs are a bugger to keep cool

captain_d0nkey · 30 Nov 2010 at 11:43

another thing is that I noticed that the person I bought the block from had two different pieces of thermal pad on top of eachother for the vrm's (probably trying to remedy the blocks vrm problem) would this cause hinder the blocks ability to absorb heat from the vrms?

JTrickle · 30 Nov 2010 at 13:38

Edit: Didnt realise you still had some of the original card cooler left acting as a heatsink for the anciliaries.

Can you take a picture and post it?
Also, does GPUz not give you temp readings for the vrms?

Edit2:

I've heard that these blocks aren't the best at cooling the vrms,

What do you mean by this? The "zalman gwb3" doesnt give any vrm cooling at all. You need to make your own sinks. I've heard people say they will chop off the rear quater of the card and keep it, but with most ATI designs there is an extra vrm further up the board which usually (but not always) has two heatsink mounting holes nearby. Will have a quick google image search for a 4870x2 bare board piccie.

Edit3:
[Located a 4870x2 half off cooler pic] Ah, you have left the ceramic plate in place? Not sure how you would screw the zalman gpu blocks down unless you drilled out the plate around the 4 mounting holes for each gpu?

captain_d0nkey · 30 Nov 2010 at 14:16

Sorry if I didn't explain clearly,

my first attempt on cooling was with the stock heatsink that I modded so that I could fit the two zalman gwb3's in

my second attempt is now with an xspc razor full coverage block

Both had the same initial outcome (severe artifacts in bios and windows) until I loosened the screws on the xspc razor a bit.

Now it boots into windows fine but when I play any games it locks, screen goes black and it restarts.

Sometimes when it restarts I used to get some red artifacts on the bios picture at the start but I have not noticed this in a while since I have tinkered wit the screws a bit again.

JTrickle · 30 Nov 2010 at 14:21

My bad. Re-re-read your first post and you did say clearly!

So does GPUz give you vrm temps? It didnt used to for my 4850, but pretty sure it did for people with 4870s?

captain_d0nkey · 30 Nov 2010 at 14:34

I'll have a look when I get home, I use everest and I'm almost certain that the vrms were displaying temps of around 54 degrees at idle

captain_d0nkey · 30 Nov 2010 at 19:01

vrm temperatures are showing up as being in the mid 40's when idle but I can't see what temperatures are like when under load because the games keep freezing! It's as if the slightest turn of a screw determines how long it will last and if it will show artifacts or not

JTrickle · 30 Nov 2010 at 20:06

First post you said 10 mins before the crash. Running a game in a window and you could monitor GPUz's temps.

However its not sounding like a temp issue to me right now (as you say yourself). Perhaps high temps could have lead to the current damage, but tbh that ceramic plate if it has good contact in the 3 vrm places might be fine as a vrm overall cooler even without the gpu fan. But I haven't first hand knowledge of that card to say for sure.

Unlikely as they might be over a straight card failure, you still have other objects in play.

a) your OS.
b) power supply.
c) Again, get some live temp readings from GPUz as it happens first in case something isnt making decent contact. Idle temps never tell a true story.

Test in another machine? Or perhaps you have an old spare HD and power supply you can quickly stick in?

Would want to start RMAing (assuming the orginal cooler could be put back in a fit state? - or get hold of one off here/the bay?) or else....

OVEN TRICK!!!

captain_d0nkey · 30 Nov 2010 at 21:53

Haha I absolutely went to town on the stock HSF, I don't think I can RMA it anyway as I bought it second hand. I think I've got it working, my thoughts were that there is some breaks in some microscopic solder points, so I loosened the block turned on the comp and let the card overheat and then quickly tightened the block into place.

First time I turned it on was a joke, it booted into windows OK but the screen looked like a NES was running windows 7. I restarted and now it works perfect and is still working after a few restarts and an hour of playing dirt 2 and a bit of bioshock.

Temps look good under load, vrms are around 60 - 65 degrees. I'm praying that this is the end of my problems but I wouldn't be surprised if a slight knock to the pc case would send me back where I started.

Thanks for your help guys, my plan from the beginning was to do a softmod on it and overclock as high as I could but I'm not sure anymore.

captain_d0nkey · 30 Nov 2010 at 22:46

I'm back, dirt 2 is just after freezing on me :mad:

. From the problems stated and very short fix does anyone know what the problem is?

By the way in the game when it freezes I get a black screen, not no signal and I can still hear the game sounds looping over and over again

EDIT:my PSU is a brand new xfx 750w core edition. Could the symptoms given match a bum psu?

Scougar · 30 Nov 2010 at 23:46

Possibly.. but have you upgraded your drivers recently?

Scougar · 30 Nov 2010 at 23:51

Confused... with the stock heatsinks on and the shroud everything works fine?

To me it sounds like A) You are not putting the waterblock on right B) There is no airflow or sufficient cooling for the VRM's C) Mounting on VRM and other chips is not right D) Combination of A, B and C.

Have you put paste on the PLX chip in the centre? Is that getting good contact?

ViRuS2k · 1 Dec 2010 at 00:08

try this before messing with the card again

in your bios bump up your QPI/PCIE voltage up.

captain_d0nkey · 1 Dec 2010 at 10:42

Scougar said:
Possibly.. but have you upgraded your drivers recently?

I installed the latest drivers as soon as I got the card.

Scougar said:
Confused... with the stock heatsinks on and the shroud everything works fine?

To me it sounds like A) You are not putting the waterblock on right B) There is no airflow or sufficient cooling for the VRM's C) Mounting on VRM and other chips is not right D) Combination of A, B and C.

Have you put paste on the PLX chip in the centre? Is that getting good contact?

To be honest because I got over an hours play from the card I'm thinking that the stock HSF may have ultimately given the same problems. My HSF is at the other end of the country at the minute so I can't get to it to test again.

Because I'm getting good temps from vrms does this not mean there is good contact being made between the board and the block.

The plx chip wasn't getting good contact so I placed a thermal pad on it to make up the distance between it and the block.

ViRuS2k said:
try this before messing with the card again

in your bios bump up your QPI/PCIE voltage up.

What voltage do you suggest, I'll give it a go when I get home tonight.

Thanks for your help everyone so far

JTrickle · 1 Dec 2010 at 11:38

captain_d0nkey said:
Because I'm getting good temps from vrms does this not mean there is good contact being made between the board and the block.

If your gpu temps and vrm temps are reading fine, I would forget about temperature. Even if it were high in the past and has led to card lifespan reduction, you are not going to find out.

And most cards just die at a random time due to solder (whatever material it is) breaks. Thats why so many people have success with the over trick. Which is what you are going to have to do once you have decided this is a card fault and not something else.

ViRuS2k · 1 Dec 2010 at 11:53

captain_d0nkey said:
What voltage do you suggest, I'll give it a go when I get home tonight.

Thanks for your help everyone so far

make sure you select the correct voltage not cpu vtt

your default should be somewhere around 1.200v bump it to around 1.290v

that should fix the black screen issue

if not you can always try bumping up the pci express from default auto to 101mhz aswell.

captain_d0nkey · 1 Dec 2010 at 18:13

Can anyone tell me where abouts I can find this option in the bios I can't seem to find QPI/PCIE could it have another name?