GTX480 EVGA Hydro Copper O/C & Temp Test results

[gabe]

Being someone that watercools their motherboard's VRMs, overclockability is second to cooling what is hot in their computer.

Anyway, as far as I know, when an electrical component such as VRMs heat up, they are less efficient at what they do, which is to provide the right voltage to the components they are responsible for. A hot VRM might provide the power "within specification" of the GPU up to their rated maximum temperature, they have to work harder, outputting more heat and becoming less efficient then if they are cooler. Also their voltage output when hot have slightly higher ripple then if they are cool. Of course if these properties add up to reduce overclockability is another issue, but you have to agree when talking about electrical components cooler = better.

And yes, GPU cooling > VRM cooling.

EDIT: OMG Gabe holyninjaeditbatman!

What you say is correct in general. But you need to go into specifics to analize what makes sense and what doesn't. The engineers at AMD, nVidia and Intel, design their products with very rigorous specifications, that are meant to produce a good balance between performance, cost, and reliability (wich is a liability for them). In the VGA business, knowing that the stock cooling solutions are very limited in size and air flow, the engineers have to use components that can sustain high heat without failing. Doing otherwise would extend their liability to unsustainable levels. So, these components are designed to run hot (relatively speaking), and the after market cooling solutions that are available all do enough of a job to extend their shelf lifes much beyond what the stock cooling does. Thus, it really doesn't matter if these components run at 50 or 60or 70. With max specs comprised between 120 and 150 (depending on models), they are so far away from their max operating point that it becomes irrelevant. Any good engineer will tell you that good engineering is SUFFICIENT engineering. Overkill is a waste ;-). We are interested in max O/C, and extended reliability compared to stock, and that's exactly what we are getting here.

This being said, let me finish the last chapter of this round of testing by posting the best stable O/C for this GTX480 under furmark xtreme burn (1hour), and using an absolutely extreme cooling bench:



What is interesting to notice, is that for a 6C drop in GPU temperature (51 down to 45), we gained 6 Mhz in stable performance (less than 1%). Note btw that I tested at 897, and furmark crashed after 30 minutes. What is also interesting to notice is that in order to do so, we went from an MCR220+MCP655 pump, to two MCR320 rads with two MCP5655 pumps. The cost of cooling more than doubled, for a gain of only 6C, and an improved OC of only 6 Mhz.

I thought I would do the above to bring things back into prospective, and allow everyone to make an informed choice when it comes to cooling!