[colecampbell666]

Weird. I came back from an hour of sledding, and it works perfectly, as my sister found out when she turned it on. It must have overheated when I overclocked it and is now cooled down enough to function. I reset the clocks to default to be sure. Thanks for the help, as always!

Ya, when you get problems when overclocking, try a reboot, then if that doesn't work, shut it down and wait for it to cool, then try again.  How about put an aftermarked cooler on that GPU?  say, one with a fan.    Then try again.

It's got a small fan, and I'm not about to spend 20-30$ on a four year old GPU.

[jr2]

Hmm, k, then just take the fan off to clean the dust from under it... ok, then try to maximize airflow to it.  Put a fan in the back of your computer, blowing out, say.  Or move a card down one slot to get it away from the GFX card.

[colecampbell666]

I see what people mean about OEM stuff being crappily cooled. This PC originally had only the heatsink fan, which actually may be good (CoolerMaster)

[jr2]

Ya.. I like having two fans in my PSU, the obligatory fan on the CPU, a fan on the Northbridge, a fan blowing out the rear, an intake fan in the front, an intake fan in the side (preferably, have a tube extending down to just above the processor fan, but I haven't done that yet...), cooler fans on my HDDs, a cooler fan on my GFX card, and heat spreaders on my RAM sticks. 

[CP5670]

The heatspreaders on memory are essentially just there for the looks. They don't actually contribute to cooling at all.

[Herra Tohtori]

Well, depends on how much they increase the heat dissipation area of the memory chips' surface and how good the airflow around them is.

What they definitely do is that they even out temperature changes of the memory chips, making the increase in temperature a little slower - assuming the thermotape attachement is of any good quality. Also, if they sufficiently increase the dissipation area, they do also increase the rate of heat removal. Again assuming the thermotape connection between the chip and the heatspreader is any good.

Of course, on default clocks they aren't necessary at all...

[CP5670]

The thing is that their surface area is not much higher than that of the memory chips on their own. The thermal interface doesn't transfer heat anywhere close to perfectly, so the heatspreader often actually increases the temperatures (although only by a degree or so). Many hardcore benchmarks remove the heatspreaders altogether and run a fan over the bare chips when pushing the memory up.

[Herra Tohtori]

Just googled some images about heatspreaders. You're right.

I thought they were something similar to the small 3rd party heatsinks taped on top of GPU memory chips, the ones you get along (for example) Zalman's GPU coolers...

If you could fit them between the RAM units, a bunch of them taped on the memory chips might work better than the heatspreaders. But in most cases it would be most beneficial to just ensure a good air flow around the bare RAM chips.

[jr2]

Hmm... ya, I'd prefer something like the 3rd party GPU sinks, but they wouldn't fit as most compys have their RAM stacked so close together its not funny.  And, if you want to do some calculations, figure out the surface area of the RAM chips themselves (the little black squares), those are the things that get hot.  Now, figure out the surface area of the heat spreader.  The heat transfers from the black chips to the spreader.  The spreader is supposed to then be a larger surface from which to remove the heat.  Perhaps it doesn't work that way, but that's the theory.  How about, instead of just a fan over the RAM sticks, you put spreader on and a fan??    I'll try that next, I think....

[CP5670]

Well, it needs to be larger by enough to offset the inefficiency of the thermal paste and the fact that the chips are no longer getting direct airflow, which is rarely the case.

Small heatsinks would be ideal, but I haven't seen any that are specifically designed for memory and are small enough to fit in the gaps between two sticks. Of course, normal people (i.e. not benchmarkers who run SuperPI all day ) usually don't need to worry about it anyway since the temperatures are only an issue at unusually high memory speeds or voltages, higher than what it takes to max out most processors right now and far beyond the point where you would see any improvement in real programs.

[colecampbell666]

What about the grill spreaders on OCZ Gold and Platinum?