Testing the Tests – CPU Stability
The following is by Michael Turner, Guest Contributor to AlienBabelTech. He has kindly given us permission to publish his work. As with everything that we publish at AlienBabelTech, the opinions expressed are solely those of the individual writer and do not necessarily reflect the views and the opinions of the rest of the ABT staff.
— Leon Hyman, ABT Senior Editor.
This is my first review here and certainly not the last. I just wanted to start off by showing you some of the procedures I will be going through for subsequent reviews.
As many of you probably saw, I posted a thread hoping for a discussion on how you determine system stability and how you achieve that stability.
As we are hardware enthusiasts, we tend to have high end systems on which we depend on for our uses. I personally believe the days of the “average Joe” computer user are over. Sure we all have a computer of some sorts, but for us enthusiasts, a computer is not an OEM box with proprietary parts and a locked BIOS.
I’d like to discuss my thoughts about what a “stable” CPU means.
We all can agree that a stable computer is one that boots up, and can run whatever you throw at it without a hiccup. It’s understandable that there are flaws in operating system designs that may require (in)frequent reboots. Other than that, it’s a pretty subjective statement. You may be able to play a specific game at a certain frequency, but with another game – it fails.
My personal definition is this:
My computer must boot and provide a reliable and expected performance. It should never Blue Screen of Death (BSOD) or hard lock based on high loads. I should never have to down clock my CPU to play a game or do some benchmarking. Finally, which I believe is the most important; the computer should never have to throttle because of heat.
The last point is highly debatable, and rarely seen. Your computer will probably not BSOD due to throttling, but is it limiting performance and thus conflicts with my definition of stability. The reason why thermal throttling is rarely seen is because it is a by-product of extreme temperatures. These temperatures come from poor CPU coolers, including the stock cooler. As it’s been pointed out many times before – stock heat sink fan combos from a retail processor are never ideal. The reason why I wanted to add this little known fact in is shown here:
The effective clock should be 145×22=3190MHz, but instead I was being downclocked to 2169MHz. I understand that when you’re processing a task that doesn’t fully utilize all cores at 100% the overclock will be “stable,” but we’re not talking about situational stability. We’re talking about complete stability.