Overclocking postmortem

B

beanman101283

2[H]4U
Joined
Nov 4, 2004
Messages
3,051
So i decided to dip my toes into the wonderful world of overclocking yesterday. I've got an Athlon 64 4000+, Asus A8N SLI Deluxe mobo, and some Corsair PC3200 RAM tested at 2.5-3-3-6.

So i was maybe a little too aggressive? I started out upping the hypertransport to 216 right away along with the memory, trying to get to 2.6GHz. I restarted and immediately started getting file corruption errors in Windows. Restarted, put speeds back at stock, ran chkdsk a few times. It found some errors, but i was able to boot into Windows seemingly ok. Then i restarted and set speeds to 206 trying to do 2.5GHz. Restarted, and booted into Windows okay. Started running Prime95. I came back and the system was locked up. I reset the system and Windows would not restart because \WINDOWS\SYSTEM32\CONFIG\SYSTEM was missing or corrupted. Did a little research and found that that's where part of the registry is. Crap.

I set my speeds back to stock then spent a couple hours researching a way to get Windows bootable again. I finally suceeded, backed up a few save game files, and reinstalled Windows. There was nothing that important on this machine.

So did i push it too hard on my first time? I'm thinking maybe my RAM can't handle the overclock, but once Windows was screwed up i didn't bother changing voltages or anything. It's not rated for anything higher that DDR400. Anyway, feel free to point and laugh. It was my first time. :D
 
i, for one, am not laughing at all. it happens. one way to try and avoid corrupting your windows install is only upping the FSB MHz at a time and testing for stability. another method of avoiding that problem is doing some stress testing with bootable floppy or CD stress testing progs like memtest86.

don't let it get you down too much.

keep backups of your important stuff, take it slow and carefull, and see what you can get out of your system.

it's slow, it's a pain in the arse, but as far as i'm concerned, that's the "right" way to overclock.

all these kiddies who jack up their FSB and voltages all at once, based on what someone else used to get a high overclock, and don't even bother to make sure that the system is prime stable will eventually get what they deserve when they are working on a report or paper and get a serious crash + corruption.

for my present overclock, i ended up using super pi 32M stress test extensively while i was seeing where things would peak out. once i was up around the limit of what my hardware could take (based on how much voltage it was taking to get things stable) i then switched over to long runs of prime95 to ensure that the settings were usable for day to day opperation.

make sure that you have your PCI bus locked, and any SATA drives are not on ports 1 and 2 and decrease HTT multi to 4x

push the FSB 5 MHz. stability test. not stable? up the CPU voltage .025. retest. not stable? back the CPU off, increase the memory voltage .1. retest. not stable? keep the memory where it is and up the CPU .025 again. retest.

repeat for a higher FSB once you have it stable. that procedure should usually get you an overclock.

if that doesn't work, you might have to think outside the box a bit and examine things like you PSU's quality and what kinds of temps you are running.

at the end of the day, there's a good bit of luck in overclocking. i had a DFI lanparty NFII ultra B that would not exceed 200 FSB come hell or high water. people tried to tell me it HAD to be something other than the board, because DFIs automatically overclock. that's a line of BS. every company puts out some boards that OC well, and some that don't OC at all.

if you got a board that won't OC, there's nothing to do but live with it until it's time to upgrade again.

good luck.
 
DFI Daishi said:
make sure that you have your PCI bus locked, and any SATA drives are not on ports 1 and 2 and decrease HTT multi to 4x
Click to expand...
Just quoting the most important part, seeing as you had file corruption.
 
Thanks for the tips! My hard drive is on port 1, and i had never heard that having it there could cause issues with overclocking. I'll move it and give it another try.
 
Just a quick question after all the reading i've been doing today. I did put my hard drive on port 3 and have started again with no problems. When people say that ports 1 and 2 are unlocked, while 3 and 4 are, does that mean they're locked to the speed of the PCI bus? And if i lock the PCI bus, would that make using ports 1 and 2 ok? Since it works fine on port 3 so far i'm not going to change it, but i'm just trying to understand how everything works together. ;)
 
it's a bit of a weird issue, really, and it doesn't affect every board, even of the same make and model.

ports 1 and 2 run of the NF4 chipset natively, and the PCI bus lock does not fixt the frequency that the controller "sees" properly, generating file corruption.

prots 3 and 4 have a hardware abstraction layer that runs through a realtek chip, which somehow allows them to lock properly, when the PCI frequency is locked.

given that the HTT frequency is supposed to superceed the FSB clock when it comes to CPU-chipset, and chipset internal communications, it doesn't make sense that this issue crops up, so long as the PCI and HTT frequencies are where they should be. somehow it does on occasion. :p
 
If I'd had known that last year I wouldn't have been reinstalling Windows twice a month.

My boss would actually make fun of me, I swear he had even made up a sign reading "X days since a clean install. . ."
 
You must log in or register to reply here.
Back
Top