In the previous post I was talking about how my CPU overclock has become unstable and that I was trying everything to make it stable again. After many runs of stress tests using Prime95, I settled for 1.4V Vcore in UEFI and started stress testing. No thanks to the huge VDroop this motherboard has, I had to set such a high Vcore. But in reality, the CPU was only getting 1.336V. This is with LLC level 3.
So I let Prime95 run overnight and went to sleep. Unfortunately after about 6hrs of stress testing, it had crashed with a BSOD with code 0x124. It requires more volts it seems. I wonder how I could get by with just 1.320~1.328V the last time. It definitely must have been the lack of AVX.
So I raised the Vcore to 1.405V in UEFI and came to work, letting Prime95 run while I was at work. It's been about 4hrs 15min and still it has not crashed. Seems like I will probably be able to make it completely stable in this go or at least the next. But what I don't get is, even at LLC level 3, there is a massive VDroop. Either everyone else is lying or they have very high tolerance. They don't seem to imply the VDroop is that high with LLC2 and LLC3. I'm not sure if this has always been the case with this board. Without using the fixed Vcore mode, you really cannot see the VDroop. And I've been using the Offset mode all this time.
I just hope that this will be it. 12hrs of Prime stability is good enough for me. This CPU must be a bad clocker. I mean, I bought this CPU second hand but unopened. Someone must have bought a batch of these CPUs and since the ones he tested weren't good clockers, he has returned the whole batch. I guess that's the only way I could have gotten an unopened but used CPU.
I wonder if 4.6GHz would not need that much of a Vcore bump to make the CPU stable. First I will try 4.5GHz, then I will see if it 4.6GHz is even possible while keeping the CPU temps below 85C. That's my temp threshold. Voltage threshold is 1.5V in UEFI which would translate to like 1.4V in reality. Even less perhaps. We shall see. But let's not get too ahead of ourselves. 4.5GHz the current target.
Wish me luck. I definitely can do with some.
Couple of hours later....
I checked TeamWeaver status and guess what? The PC was not online. I thought it had crashed and I messaged my wife telling her to pull the plug again. And I was stunned by her reply. She said he had already pulled the plug. Just confirm that the autocorrect took control of the tense, I asked whether she did that before I messaged her or after. She said BEFORE! OMG! She had thought that I had left the PC ON by mistake and she had turned off the PC. Not by shutting down, but directly pulling the plug. Goddammit! Argh, women!! She should have asked me what to do first if she wasn't sure. I don't think the stress test failed at all. She had pulled the plug about 6hrs into the stress test. Sucks!
Now I have to start from the scratch. :-(
No comments:
Post a Comment