For the record I don't like prime95 i don't think it's a fair test. It is extremely unlikely that I'll find a program that uses all four cores 100% running 4 threads each. It's just not a real world test. Even with all of the video editing/converting I do I don't think that happens. Most programs just aren't written to do it.
Well said, Rigit, I couldn't agree more. Its quite aggressive for personal use. Running prime95 for a two hour marathon has the potential to put huge restrictions on an overclock
and in many cases this will limit what your system is capable of in the real world. These virtual limitations, if you will, imposed by the aggressive nature of the test leave what is called a grey area -- and that grey area can really be thought of as the elephant in the room. Or the Chat room, as in this case. Many, many times, the first answer to a new thread is:
"Did you run prime for two hours to make sure your overclock is stable?" And let me be clear, that's a great question but the term 'stable' is qualified here with both "two hours" and "prime95" and that's where we start running into trouble; you see for some overclocking enthusiasts prime95 can be thought of as the quintessential instrument for measuring reliability. Without a pass, your overclock is useless, unfit and unreliable. And if you dare question it, you are not only questioning the program, but also questioning the person behind the name. You are calling into doubt the very essence of what that individual believes in and holds true when crafting the perfect overclock. What these folks forget, is variety, the spice of life, the simple point that not everyone who overclocks requires the same treatment. Not every user is running F@H 24/7. Not every user is a hard core gamer playing Crysis 3 for hours on end. Not everyone is using a PC in a resource intensive way and many of them will never, ever peg all CPU cores and subsystems for 2 hours straight.
In reality, many users will never need the stability that prime95 demands. Case in point: I have seen a handful of overclocked systems fail in blend mode after 30 minutes of testing, yet maintain a record of 100% uptime for years, even during intensive use. How can this be?
Because the reliability point measured in prime95 is much higher than what most users will ever be able to throw at it. Indicating there is a grey area between real world stability and that which is dictated by Prime95. And as a result, perhaps people are misguided into setting overclock limitations prematurely. This means they've lost out, for the sake of an over-aggressive reliability test. I want to push my system to the absolute limit of what it can handle in the real world, not in the world of virtual reality. I test stability by
actually using the computer day by day, in the real world. In a very basic analogy, this entire issue of reliability can be compared to a sea trial for a new ship. You take it through countless tests of engine power and weather to determine
if it is fit for its purpose and use on the open seas. You certainly don't throw a hurricane and a rogue wave at a 50 car ferry at the same time and expect it to stand tall. You tailor or harmonize the sea trials and reliability tests towards what that boat may be expected to encounter during its lifetime. Same thought process can be applied to reliability and how it is measured on a PC.