Overclocking causes Windows corruption?

serialb

Diamond Member
Mar 20, 2000
3,107
7
81
I have my 2400+ on Gigabyte 7N400Pro2 (yeah, i didn't know about the dipswitch deal between Pro & Pro2 back then). Fresh XP Pro install and everything was happy until I started messing with the multiplier & FSB.

At some point during my search for my rig's limit, Windows failed to boot and it shows some kind of disk error which I don't remember exactly what it was. Well then, I went back to stock multi & FSB and it actually boot into Windows. However, the system became very unstable and powers off itself in a matter of hours running S@H. The situation gets worse and worse and even Norton AV and Outlook starts giving me critical errors and prompts for reinstallation.

Before I go ahead and do a clean reinstall, which I think is the only option I have. I am wondering if overclocking would actually corrupt system files? If so, is there a safer way for you all OC'ers out there to find the sweet spot? This time, I am sure I'll make some kind of a ghost image before I tweak. :eek:

Thanks in advance for any ideas.
 

maluckey

Platinum Member
Jan 31, 2003
2,933
0
71
To oversimplify, yes.

Overclocking can lead to data corruption. Often times it can be fixed with Chkdsk. Roll back attempts are sometimes successful as well as sytem restore (more successfull).

The best tip is to go only a few Mhz FSB at a time, and stress test for a couple of hours before going to the next level. If you are only tweaking the CPU multiplier, then the same thing applies. Large jumps are asking for trouble.

One last tip is to find the weak link and tune to that. My weak link is RAM. I found that at 202 Mhz FSB it craps out at random. I tuned for 200 Mhz FSB and all went smooth after that. I was forced to drop the CPU multiplier, but the seventy Mhz the CPU lost was made up by the extra FSB speed.
 

Jeff7181

Lifer
Aug 21, 2002
18,368
11
81
If your motherboard doesn't have a PCI/AGP lock as you increase the FSB, the PCI clock will increase with it. Normally it's 33 Mhz... if you go above 36 you'll most likely start to see data corruption.
 

dunkster

Golden Member
Nov 13, 1999
1,473
0
0
Check your memory timings.

Settings such 1T Command Rate or Tras set too low are very likely to scramble an HD.

Hope this helps!
 

pelikan

Diamond Member
Dec 28, 2002
3,118
0
76
While raising the FSB or tightening ram timings boot into memtest86 before going into Windows. Make sure you can pass test 5 and then you probably won't corrupt windows.
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
make sure agp/pci is locked or you will corrupt HDD if pci gets to far out of spec...

Often times the ram getting out of spec will corrupt it as well....XP is usually pretty good at correcting the ram errors but HDD erors can sometimes be unreversable unless you run a repair....


I agee with pelikan. Run the memtest first 1 pass and you should be safe...I think there are a couple other test that are also useful though...
 

serialb

Diamond Member
Mar 20, 2000
3,107
7
81
Thanks for all the above great advice. I'll start making that as a habbit. Thanks.
 

joe2004

Senior member
Oct 14, 2003
385
0
0
While raising the FSB or tightening ram timings boot into memtest86 before going into Windows.
Exactly.
Make sure you can pass test 5 and then you probably won't corrupt windows.
For the issue of corruption one pass is enough.
 

rgreen83

Senior member
Feb 5, 2003
766
0
0
Originally posted by: joe2004
Make sure you can pass test 5 and then you probably won't corrupt windows.
For the issue of corruption one pass is enough.

I think he just meant to go through to test 5, not make more than one full pass. One full pass is always enough.
 

joe2004

Senior member
Oct 14, 2003
385
0
0
I got it. Yes, test 5 is enough for corruption thing. For the stability one needs at least one night of running the whole enchilada.
 

dunkster

Golden Member
Nov 13, 1999
1,473
0
0
When I boot into Memtest86, I go straight to Test#5 (C 2 5 5 Enter Enter). Any memory settings that will run a minimum of three error-free passes of Test#5 have always been stable in Prime95 or any other program or game. It's usually the most critical test, and several passes takes only a few minutes.

Similarly, running Prime95 for more than 20 minutes is a waste of time. I've never played a game that was unstable on a system that is Prime-stable for 20 minutes.
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
Originally posted by: dunkster
When I boot into Memtest86, I go straight to Test#5 (C 2 5 5 Enter Enter). Any memory settings that will run a minimum of three error-free passes of Test#5 have always been stable in Prime95 or any other program or game. It's usually the most critical test, and several passes takes only a few minutes.

Similarly, running Prime95 for more than 20 minutes is a waste of time. I've never played a game that was unstable on a system that is Prime-stable for 20 minutes.


That means nothing!!! I have primetested 12 hours and could not encode a Divx movie, run TMPgenc, or render in autocadd...I raised the vcore a notch and worked in al of them....

Don't get so complacent....
 

serialb

Diamond Member
Mar 20, 2000
3,107
7
81
Originally posted by: dunkster
When I boot into Memtest86, I go straight to Test#5 (C 2 5 5 Enter Enter). Any memory settings that will run a minimum of three error-free passes of Test#5 have always been stable in Prime95 or any other program or game. It's usually the most critical test, and several passes takes only a few minutes.

I have quite a bit of errors in test #4. Sometimes, the error wouldn't show up during the first pass. Is this normal?

I already have my RAM running on SPD. :(