I am seeking community feedback regarding the future direction of the currently stickied Overclocking CPU/GPU/Memory Stability Testing Guidelines thread.
Originally it was drafted with the intent of harmonizing and aligning a baseline level of stability testing for members to fall back on when reporting their 24x7 OC's as well as being a general assembly point of stability testing links for the various components in your system (beyond just that of the CPU).
My motivation for drafting the sticky was born from experience and observation - at the time I was wrestling with mapping out the stable clockspeed versus voltage versus temperature space for my 2600k and I came to the realization that we (the AnandTech Forums) didn't really have a singular location that members could look to and find links to widely accepted/adopted stability testing programs like LinX and Memtest and so forth.
Further, from observation it was easy to see that we had an unnecessary level of friction and angst transpiring between various members owing to claims being made of having 24x7 OC's as verified by differing means of stress testing. It was really becoming an inflammatory topic matter for folks who would use say Aida64 or super-Pi as their stress tester and then report higher OC's as being stable in comparison to the LinX tested rigs of their peers.
Now no two computers will be OC limited for the exact same reasons, different circuits within the different chips will be the weakest link and it stands to reason that no one stress tester program is going to cover all the bases. But it would be nice to have a sort of "agreed upon minimum testing regiment" for establishing one's claims of "my 24x7 OC is x.yz GHz at a.bcd Volts".
One goal of the existing sticky was to avoid making the stress testing protocol overly complicated or tedious such that it would be a turn-off to many new members. We have to keep it simple if we want people to give it a go.
Right now I am once again wrestling with finding the sweet spots for both my i7-3770k and my FX-8350 and I am once again finding that relevant programs have evolved in the past year or two since the sticky linked above was drafted.
Now then - the purpose of this thread, and the vote, is to determine if the community would like to see the existing sticky taken down and replaced with an entirely new OC Testing Guideline thread, or if the community would prefer to preserve the existing sticky and have me basically re-write the OP (at the risk of making some of the existing 150 posts seem odd or out of place).
Furthermore, in addition to voting on the fate of the existing sticky thread, I'd very much appreciate if members would chime in with links and proposal for testing regiments with stress testers.
For example, as we have observed it is not enough to merely say "5 passes of LinX" or "3 hrs of LinX" because we also know the amount of memory used in the LinX testing effects the level of stress the program places on the CPU and the ram.
We also need to keep it simple so the adoption rate remains high, the best testing regiment is not going to be the one that involves running three separate programs simultaneously because few people will actually go to the effort of running such a setup. Keep it simple, good enough to rule out the silly unstable OC's but we aren't looking for procedures that call for the members to have their rigs running LinX for days on end either because then hardly anyone will actually do it.
Originally it was drafted with the intent of harmonizing and aligning a baseline level of stability testing for members to fall back on when reporting their 24x7 OC's as well as being a general assembly point of stability testing links for the various components in your system (beyond just that of the CPU).
My motivation for drafting the sticky was born from experience and observation - at the time I was wrestling with mapping out the stable clockspeed versus voltage versus temperature space for my 2600k and I came to the realization that we (the AnandTech Forums) didn't really have a singular location that members could look to and find links to widely accepted/adopted stability testing programs like LinX and Memtest and so forth.
Further, from observation it was easy to see that we had an unnecessary level of friction and angst transpiring between various members owing to claims being made of having 24x7 OC's as verified by differing means of stress testing. It was really becoming an inflammatory topic matter for folks who would use say Aida64 or super-Pi as their stress tester and then report higher OC's as being stable in comparison to the LinX tested rigs of their peers.
Now no two computers will be OC limited for the exact same reasons, different circuits within the different chips will be the weakest link and it stands to reason that no one stress tester program is going to cover all the bases. But it would be nice to have a sort of "agreed upon minimum testing regiment" for establishing one's claims of "my 24x7 OC is x.yz GHz at a.bcd Volts".
One goal of the existing sticky was to avoid making the stress testing protocol overly complicated or tedious such that it would be a turn-off to many new members. We have to keep it simple if we want people to give it a go.
Right now I am once again wrestling with finding the sweet spots for both my i7-3770k and my FX-8350 and I am once again finding that relevant programs have evolved in the past year or two since the sticky linked above was drafted.
Now then - the purpose of this thread, and the vote, is to determine if the community would like to see the existing sticky taken down and replaced with an entirely new OC Testing Guideline thread, or if the community would prefer to preserve the existing sticky and have me basically re-write the OP (at the risk of making some of the existing 150 posts seem odd or out of place).
Furthermore, in addition to voting on the fate of the existing sticky thread, I'd very much appreciate if members would chime in with links and proposal for testing regiments with stress testers.
For example, as we have observed it is not enough to merely say "5 passes of LinX" or "3 hrs of LinX" because we also know the amount of memory used in the LinX testing effects the level of stress the program places on the CPU and the ram.
We also need to keep it simple so the adoption rate remains high, the best testing regiment is not going to be the one that involves running three separate programs simultaneously because few people will actually go to the effort of running such a setup. Keep it simple, good enough to rule out the silly unstable OC's but we aren't looking for procedures that call for the members to have their rigs running LinX for days on end either because then hardly anyone will actually do it.
Facebook
Twitter