Linpack?

Quote:
The IBT uses the Linpack test (but I heard that it's not just the Linpack. I heard that it's different, or more stressful. Eh. I dunno yet.)

Anyway, it's a safe test, but the temps do get about 10-15°C warmer. Not only that, but it's far more stressful on the CPU, and thus can kind of show you how stable it truly is. I had the same problem you had (right down to the last detail), and I found that, even though my CPU was stable at 1.304V in Orthos with Small FFTs for 12 hours, I had to increase the core voltage to 1.320V for OCCT'S Linpack test. I ran the Linpack test for over an hour, and then stopped it while simultaneously saying "ok now it's stable".

So, as long as your temps aren't getting really insanely high, then I say it's 100% safe. But do be a little more cautious than you are with the Small FFTs test.

Quote:
This damage occurs when people don't know what they're doing and allow the temps to get too high. There is too much mis-information out there about overclocking and stress-testing because so many people go ahead and do it without first doing their homework. If everyone did their homework first, then almost nobody would damage their hardware.

Quote:
Not to be rude or anything, but stop being so paranoid. As I said, it will be fine as long as you know what you are doing. Watch the temps. And if they're going like way above 70, then find a way to reduce them and continue.

But with your overclock of 3.87GHz, I'm guessing your voltage has to be at least 1.40V. Regardless of where I was going with that, I say that all you need to do is increase the voltage. After all, 3.87GHz is a huge overclock. Nay, it's massive. So, the Linpack is showing you that, technically speaking, your CPU isn't 100% stable.

And again, stop worrying. Don't be afraid. Just be smart.

Quote:
Quite true. This is pure fact, not opinion. Nice post.

Quote:
Agreed.

You want your computer to pass at least 25passes of 1: maximum ram and system tested.

I do 25-30 passes of Intel Burn Test, 2 hours of OCCT and if it passes both, I do prime95 for 12-24 hours.

Intel Burn Test linpack testing is more efficient in terms of overclocking because it will tell you VERY quickly how much vcore your chip will need at a given clock.

good luck with your chip.

Jam3s

It's OCCT website where it states that their Linpack test is the same as IBT.

Quote:
I do not really agree with the statement that no one even uses their comp for 100% for even an hour. This is simply not true if you fold, specially if you use the SMP client.

What I think is that every test has its own use:
- Want to know if your cooling is ok? --> Linpack 15 minutes
- Want to know if your Memory is Ok? --> Memtest86+ 6h - 12h
- Want to quickly know if your CPU is "quick stable"? --> Prime95 small FFT 30 min
- Want to know if your overall system is "quick stable"? --> OCCT 2h
- Want to know if your system is "rock solid stable" --> Prime95 6h - 12h + OCCT 3h - 6h

Quote:
I guess I forgot to update my Sig Rig, but I just replaced my Q6600 3.87Ghz OC with Q9650 currently at 4.12Ghz @ 1.30v. I am also running 8GB of OCZ Reaper 1066s, OCed to 1100mhz @ 2.24v. I have passed so far:

6 hours of Small FFT P95
3 hours of OCCT 3
3 hours of OCCT 2 (Just in case)

I will be testing more of OCCT 3 for another 6 hours tonight after I go to bed.

The highest my temps go right now either with OCCT2 or Small FFT is 53c under load, so my temps are not the issue. I did do another run of LINPACK after setting my Vcore to 1.32v, and I ran it for 20min, and I saw my temps go to 60c under full load, so once again, my temps are fine.

My paranoia is coming from a issue I had with my main P5Q-Deluxe acting funny with my new CPU. It would not pass any test other than Small FFT no matter what I did. So, as a final check, I replaced my motherboard with a BNIB P5Q-Deluxe I had laying around, to find out that everything is working properly now. I guess I got a little scared because I didn't want to stress this motherboard to a point, where it was going to do the same thing. I am not saying my OC damaged my other board or anything (could have been defective from the get go), but...

How does any one of us know for sure how much it DOES degrade our stuff, running stress tests that are designed for stock CPUs with stock settings (IBT?). Even with P95 and OCCT, it stresses our system redundantly, meaning to a point where we will never use in real life. What does LINPACK prove that other stress tests don't? (It's a question, not a comment.)

Quote:
No wonder OCCT's Linpack test can be toggled between 32-bit and 64-bit (well, not on my system. The 64-bit checkbox is disabled). This isn't to say that I wasn't assuming or guessing that it was the same Linpack test that IBT uses, because I was. But, I never said anything because I was never 100% sure. I'm glad you posted that!

Quote:
4.12GHz with a Q9650? Whoa. Your system must be blazing fast!! Nice.

Anyway, since the temps are fine, then the core voltage (or some other voltage) probably needs to be increased a tiny bit more.

As far as the difference with Linpack, I think I read that it uses calculations that are far more complex and "difficult" for the CPU. But, rest-assured, stress-testing doesn't degrade our stuff much at all. What degrades our stuff the MOST (far more than stress-testing) is having it overclocked on a daily basis. I was told a few times here on OCN that CPUs left at stock are designed to last ~10 years (approximately). But, they also said that with the way some of us typically overclock them, we should only expect 2-3 years out of them at the most, and then at that point we should only expect stock settings to work the best until the end of its life. However, most overclockers replace their CPUs before 2-3 years pass anyway, so it doesn't matter one bit. For example, now that I'm really into this (and I wasn't into this at all up until around March of 2008, and didn't build this rig until around late April/early May), I know for certain that I will replace the CPU within 365 days from now.

Now, I grant you that anyone who keeps their CPU for 2-3 years is definitely wise to do regular stress-tests as a way to keep checking for degradation. However, it's my guess that these tests should only need to be performed like maybe once every 3-6 months. This way, they can prevent data corruption due to instability. Not only that, but they can know when it's time to begin shopping for a new CPU to kill, err, overclock. hehe

So, don't think of it as simply killing the CPU completely. Think of it as very, very slowly and gradually killing it's ability to overclock.