11-22-2018 04:59 PM - last edited on 03-06-2024 07:43 PM by ROGBot
11-22-2018 08:12 PM
11-23-2018 12:25 AM
JustinThyme wrote:
The entire premise of the LLC and how it works does result in overshoot especially when going from loaded to unloaded. Raja did a write up here some time ago and explained things much better. The object is to get the highest OC with the lowest Vcore and lowest LLC.
Im not too Familiar with the VRMs on your board and where they stand in the ability to react to load changes.
My general process, right or wrong is start with LLC of 3 and push the Vcore as far as I dare (depends on which chip for amount) and bump up the multiplier on the CPU first until it becomes unstable then bump the LLC to 4 and try again. Still unstable bump to 5 and thats as far as I go. If still unstable Ill back off the clock until stable then move on to the uncore than to the memory. Lather, rinse, repeat. Der8auer has a nice OC guide that the basic principles apply to pretty much every board and is a great starting point.
https://rog.asus.com/articles/guides/video-guide-der8auer-overclocks-rog-rampage-v-edition-10/
11-23-2018 01:49 AM
11-23-2018 02:26 AM
HiVizMan wrote:
I personally like to have some vDroop. It is a design feature that has value at so many levels. The most important is the health of the CPU itself.
Personally I feel too much emphasis is being placed on you passing an artificial stress test. If your system does what you use it for 24/7 without any problems then it is stable mate. All passing a few hours of cooking your CPU with prime confirms is that it passed that run, no guarantee that it will pass again. I just don't see the point of cooking my CPU like that.
11-23-2018 08:08 AM
11-24-2018 06:17 AM
HiVizMan wrote:
I hear you mate.
I think running RealBench 2x with pass is enough for a day to day rig, yes.
A 0.03 Vdroop is certainly acceptable.
As to the image you linked, it shows a graphic representation of the different stress tests as determined by a particular set of parameters. Are you asking if one stress test is more relevant than another and I would answer that each stress test has all the value that the user of that stress test places on the test. So by that I am saying, if you belong to a group that gain bragging rights by how long they can run a particular stress test continually without failure then that stress test has great value to you. I know folks who run Prime for 4 days continually as a test of stability. It makes them happy so it has great value to them. To me very little.
I have personally seen folks do that extreme CPU stress testing and be freaked out when their system failed the first time a flash video was played, or even more embarrassing that their system would not resume from sleep. Your system is stable if it does what you want or need it to do without errors. Simple as that.
I benchmark. I overclock the shyte out of certain hardware. I build systems for financial institutions that want 100% stability. Real world usage is what determins fit for purpose not some artifical test. Rog RealBench written by a ROG member called Nodens is all about real world stuff and real world applications. That is why it gets my vote each time.
Enjoy mate.
11-24-2018 07:58 PM
11-25-2018 02:35 AM