A nearly instantaneous stability test?

Discussion in 'Overclocking, Cooling & Modding' started by graysky, Nov 21, 2009.

  1. graysky

    graysky ARP Reviewer

    I may have discovered a VERY quick method to test an overclocked system's stability. I need more data/people to test it though. What I've found is that a Linux installer seems *highly* sensitive to CPU's with vcores that are too low. I don't think this will help you understand if your other voltages are too low (VTT, DRAM, NB, etc.), but I do think that as you begin to arrive at a stable overclock by minimizing your vcore, this may be the thing to use before hours and hours of linpack or prime95.

    The distro is a mythtv-based one called LinHES. You don't need to install the distro at all to test the system. You only need to boot from its live CD. In my experience, if the vcore is too low, you'll see a kernel panic and the live CD just freezes (reset switch will reboot just fine).

    Url to download the iso: ftp://knoppmyth.net/R6/R6.01.00/LinHES_R6.01.00.iso
    (it's 700 megs)

    Here is an example screenshot of what a kernel panic looks like:

    If you do not see a panic, the first screen you'll see is the "Welcome to R6" screen. If you hit "OK" you can then reboot from the next menu with no changes to your system.

    What I've found is that this simple test has been predictive in 3/3 cases that I've used for stability. In other words, it machines passed this test, they also passed 20+ hours of linpack testing and 24+ of prime95 testing.

    DFI BI P45-T2R (BIOS: B245D914) | Intel E6500 @ 10.5x333=3.49 GHz
    266/800 MHz Strap CPU Vcc=1.2875+50 mV NB Core=1.1950 CPU VTT=1.1800
    MD4GX2M2A1066C5 2 x 2 Gb @ 5-6-6-18 @ 1,001 MHz @ 2.100 V

    Vcore above the stock VID of 1.1875 | Result
    +0 mV | kernel panic
    +25 mV | kernel panic
    +50 mV | kernel panic
    +62.5 mV | successful boot

    Now, I will say that the +50 mV run was stable to prime95 for 24 h and also to linpack for 14 h. Again, it would seem that this test is *highly* sensitive.

    What do others think?
  2. Adrian Wong

    Adrian Wong Da Boss Staff Member

    Hmm.. In other words, the VIDs you tested were "stable" while idling in Windows even if it fails the test (shows a kernel panic), just that they won't pass hours of Prime 95?
  3. graysky

    graysky ARP Reviewer

    No, it's the other way around. For example:

    10.5x333 @ 1.2875 V +50 mV = stable to both linpack and prime95, but not stable to this test. Adding an addition 12.5 mV to a total of +62.5 mV does pass this test.

    I'm testing now 10x333 @ 1.2875 V + 25 mV in linpack now which passed this test. I believe it too will pass 24 h of linpack. Running at the slightly higher vcore might be overall-kill since I have shown it to be perfectly stable with p95 and linpack too. I think there are other forms for stability beyond mega CPU usage from p95/linpack, but I don't have a tangible example of them beyond this test.
  4. Adrian Wong

    Adrian Wong Da Boss Staff Member

    Oh, I see what you mean...

    If you are right, then the kernel panic test would be a far more sensitive (and much faster) test of CPU stability than Prime 95 and Linpack.
  5. VictorWolf

    VictorWolf Newbie

    Any Follow up on this?

Share This Page