How Does My Power Supply Impact Overclocking?

Yesterday I was having a conversation on IRC with a bunch of guys that are very much into building desktop gaming PCs and of course many of these enthusiasts focus on overclocking. I was asked how much a PSU came into importance when we are overclocking our system in terms of overall stability.

As mentioned above in the intro, someone wanted to know just how much impact a PSU could have on system stability when overclocking a system. I am going to paraphrase a bit here, but this is what was asked.

One thing I've always wanted to know, is there an actual difference a good PSU makes in terms of stability when overclocking a PC? I am curious as to how much ripple and voltage regulation actually has to do with overall hardware system stability.

In my head, common sense tells me,"A lot." But then again, what the hell do I know, so I reached out to Paul Johnson. Paul is our resident PSU Engineer. He is the guy that writes all those PSU reviews and does the testing. When it comes to power supplies, Paul is one of the smartest persons on the planet and also extremely passionate about all things PSU. His passion for PSU information is the main reason he writes our PSU reviews. If you have not read any of his reviews, you are missing least if you want to "geek out" and get a lot of tongue-in-cheek humor along the way.

Now please do not read this as a highly technical answer to the question posed. While I have done some editing on this, it is still pretty much in its original form as we was being emailed back and forth, off the cuff. There is a lot of simplification and generalization written below in order to make it a "quick" answer. This is more of a "conceptual answer" than a "theory and application" answer.

The answer to your question is that it depends on how well managed your heat output is; how long you want the hardware to last; and how far towards the edges of the ATX12V specification you want to be.

Ripple/noise have what I would call a moderate amount of impact on overclocking per se in the short term unless the values are massively out of specification. However, the closer you are to a clean output the less work that is done by the downstream VRMs, MOSFETs, and capacitors. Less work means the less wear, the less heat, etc. that these components experience. The less heat these downstream components experience, the less heat you have to manage in the system, which calculates to longer component lifespan. The longer you are able to push the boundaries of those components (potentially) before you cross that line where those fail or the effects of those stressed components on a GPU or CPU of increased voltage to those components to overcome the increased resistance (it is somewhat of a viscous circle) that causes heat problems in the CPU/GPU itself. That all said, in reality, the ripple/noise values are more a long term damage issue due to the better quality motherboards these days. (This is of course subjective depending on how "long," "long term" is, but not in the few minutes you are trying to hit some peak OC then back off to a lower value.) But your poor hard drives....oh the data!

Voltage regulation can be a much more obvious problem. All of the things above apply to voltage regulation since voltages are "stable" (or so we hope). Not seeing the same kind of change rates we see with ripple/noise, we don't have a cyclic problem but a continuous one where we are stepping down a voltage (12v mostly) to something much lower and in the efficiency side of things we are seeing losses in the form of heat. So, again, all of the issues you have with components on your motherboard or your video card heating up, and the stability issues that comes with overclocking. The second issue is, "Are we in specification or not?" All PSUs have some sort of base set point for each voltage in that particular unit due to tolerances and screening of components. It is rare that the set point is nominal (12v, 5v, or 3.3v). So, the regulation becomes immediately important because if your unit is near one of the specification limits, or you just have really poor regulation, and a load causes the voltage to exit the specification limit, you may exceed what your VRMs can handle. Or you maybe where thsoe can handle it but those are unable to provide the voltage called, and then that is it. So, you want a unit with a good of voltage regulation as you can get so you don't deviate anymore from whatever its set point within the spec range is than is necessary so there is less stress on those components downstream.

All of that said, how much does all this affect stable GPU and CPU overclocks in a scientifically quantifiable way? Holy god the number of confounding variables! The only thing I would reliably say and not end up eating a hat without access to a TON of programmable power sources and CPU and GPU cooling combinations is that less ripple is better and tighter voltage regulation is better than loose.

I hope that helps.

While far from the normal data collection and testing we do on PSUs, I thought it would be a great note to pass along to our readers as well as generate a good open conversation about this topic.

Article Image

For all this, I have decided to give Paul his first HardOCP Silver Award! Why not a Gold? Well, you might already know how picky Paul is when it comes to handing out Gold awards, and I figured he would gripe at me for being too easy on him if I dealt out a Gold! big grin