HD6850 + Accelero S1 rev2 What am I doing wrong?

[fokfokfok]

I'm trying to replace the stock cooler on my gigabyte HD6850 with a S1. Unfortunately I get very high temps (CPUID and GPU-Z):

Idle = 50
furmark = 103 in a few seconds and I close the program to not damage the gpu. Similar results with Heaven DX11.

To solve the problem I did this:
1. replace the thermal paste with Gelid GC2. Temps went down one degree.
2. Mounted a slipstream 800rpm fan running at 12v. Result temps went down 5 degrees.

I've tried to touch the cooler when the gpu was supposed to be 100 degrees and it doesn't feel hot at all. More like 30 degrees.

When disassembled the card after my first try I could see that the thermal paste was distributed evenly on the gpu, so I think there is contact.

I didn't test the temps with the stock cooler mounted.

Any ideas?

Thanks

[frenchie]

How much paste did you put ?
Does the base of accelero get hot ?

Is it a new Accelero ? Is it modded ?

[fokfokfok]

I applied quite a bit paste because it was kinda thick so it was hard to make a thin layer.

I have to remove the fan to see if the base get's hot. I will try that tomorrow.

It is a new Accelero and the only thing I modded was one of the small heatsinks.

Maybe I should try some other thermal paste. Maybe Arctic Silver 5?

[Cryoburner]

I don't think the type of paste will likely make much more than a few degrees difference. If you're using too much though, it could significantly reduce heat transfer. It should just be a very thin layer, enough to fill in the imperfections between the two surfaces. Using the edge of an old credit card might help to make a thin, relatively even layer.

[Ratty]

I know when fitting an Accelero S1 to a 5XXX card there is a problem where the spacers are too large preventing the GPU die from coming into contact with the base of the Accelero. This is due to the height of the die being reduced by a few mm in the 5 series cards (as opposed to the 4 series). Here is a good example I found on Google showing a clear gap when the thermal paste is removed:

http://farm3.static.flickr.com/2313/451 ... 4a42_o.jpg

I am not sure if this issue is present with the 6 series cards but could be worth checking? If that is the case you can either cut the spacers to size or get rid of them all together (if you do this be careful not to over tighten the bolts).

[lodestar]

The 6850 is quite a hot running card, some reviewers have recorded a idle temperature of 40C+ and load temperatures approaching 80C. Arctic have an add-on 2 x 80mm fan Turbo module for the Accelero S1 and I suspect that its use is mandatory with a graphics card like yours. As an alternative, if it will fit, a 120mm PWM fan of your choice attached with cable ties plus one of those GPU small->standard PWM header adapters (such as this one http://www.quietpc.com/gb-en-gbp/produc ... -vga-cable) could be worth considering. Using software like MSI Afterburner you could then create a custom profile enabling the 120mm PWM fan to be quieter under idle. If you picked a PWM fan like the Scythe SlipStream SY1225SL12LM-P it would still result in substantially less noise than the stock cooler. Probably below the noise level of existing case fans making it effectively silent.

[flemeister]

Double check that the heatsink base is making contact. You may need to remove the plastic spacers so that it makes firm contact. It's a common issue with the Accelero S1.

[fokfokfok]

Thanks. It was the spacers. I removed them at got good temperatures. The joy was short though. I tried to run furmark and the gpu broke at 90 degrees Pretty weird since it worked fine at over 100 before. Maybe it was already damaged from the short periods with heigh temperatures.

Now I got vertical bars when I'm booting and windows will only start in safe mode and there are small vertical lines all over.

What a waste. Now I'm gonna get a 5770 passive, which I should have got in the first place.

[mentawl]

Did you have some decent cooling on the VRMs? Furmark is particularly brutal to them (indeed I'd hesitate to use it for testing an unknown cooling solution), and it's possible those have damaged the card by getting too hot.

[fokfokfok]

I'm not sure how the WRMs look but I mounted the small heatsinks that came the S1.

[ntavlas]

I think the vrms got damaged by excess heat too. I doubt the gpu itself would get damaged by running at 90-100 degrees for a few minutes. My explanation is that during the first test, the gpu throttled due to high temps. In the second test, the gpu was cool enough to run at full speed requiring more power and causing the mosfets to overheat...

If you want to cool a powerful card passively, the safest (and most expensive) solution is to use a dedicated heatpipe cooler for the vrms. Thermalright makes such, compatibility is rather limited though.

[fokfokfok]

That would be good explanation. The other test I made I close furmark after a few seconds while the last one it ran for half a minute.

I think I'll just buy a passive 5770. I don't play games much these days anyway.

What really bothers me is that I was a bit careless. I actually had a fan in front of the card that wasn't running. I thought I would be safe if I just had my eye on the gpu temp.

[deimos3428]

Sorry I'm late to the thread.

It might be too late, but check the heatsinks on your VRMs. I've found one of the S1's heatpipes tends to knock the bottom-right VRM sink off during mounting. You need to offset that sink about a mm to the right so this doesn't happen.