Printable View
I don't know too much about overclocking but I do. I also had a lot of issues getting this dual core up. I basically had to underclock the ram (not sure if thats how to say it..)
edit: this oc should only be about 20% oc, not sure why it jumped up like that.
https://www.sharkyforums.com/images/...2006/02/33.jpg
SuperPI is a rather poor test of stability as 32M only takes me 28minute to run or less IIRC... Its only good for a quick run then using a better tester for those long hauls...
I have even see it before, that superpi 32M pass, but fail when benchmarking in 3DMarks :) So its not really a good program for stability.
What I do is normally drop the memory multi, cpu multi, HTT and raise my voltages to where I want. Usually 1.5v max on my CPU. Then start to raise the FSB by 10Mhz until its unstable. Then backdown until it is stable.
Also if your PCI is not locked, that can cause a problem. On my EPoX board I needed to set the PCI to 67 Mhz to lock it down. My Intel boards have a lock option.
Once I get a stable overclock, I would change the HTT so I dont run to far passed 1000, But I have been stable up too 1100ish. Then start to bring in the memory divider and timings and voltages.
Since I use BH5, I like using 3.0 - 3.6v for vdimm ;P But that wont work for all ram. TCCD rather hates additional voltages. OCZ VX ram liked voltage, but my set sucked.
You can try increasing the Chipset voltage if there is an option, I have seen that help with a higher FSB to help stabilize it.
yeah, i need to play with the chipset voltage and see if that gets me anywhere. i'm prime and memtest stable at 11x228 (2508Mhz, memory 1:1, HT Bus 4x), but not prime stable at 10x250 (2500Mhz, memory 1:1, HT bus 4x). not sure on memtest stability at 10x250 though. i need to test this yet and see if it's the memory. if so, it's RMA time, because they're rated to do 250 at 3-4-3-8 with up to 2.8v...i'm using those timings and 2.65v (more makes them unstable).
not sure what kind of chips are on the mushkin, haven't peeked under the spreaders, but i can guarantee they're not BH5. i have 2 sets of kingston hyperX BH5 which absolutely love as much voltage as i can give them and run 2-2-2-6 at any FSB i can push them to. this stuff will crap out real quick if you try to push them with those kinds of timings
So Colossus what program do you use to test the stability of your system?
Computer Maniac :)
Amen!Quote:
Originally Posted by Colossus
Case in point:
i just finished building my current rig. It passed SuperPi 32Mb, MetaBench, PCMark, HOTCPU, 3DMark 01,03,06 & 06, so I thought it has to be stable. It wasn't until I started encoding movies for DVD burning while doing some work that required heavy network traffic that I would get a BSOD. The BSODs were all device driver failures. It took me almost 3 days of trying new drivers, taking everything out of my case except video card, swapping out video cards, etc., that I found out that my memory doesn't like 1t @250Mhz. As soon as I set it to 2t... no more BSODs.
Point being, that there is no one test for stability. ;)
Absolute Stability -
Loop Memtest #5 for 24 hours
Try Super Pi 32MB on each core
Prime95 (choose the most CPU intensive option) on each core 24 hours
Run 10 loops of 3dmark05
To be honest I don't go all that far. The max I'll run something is overnight or around 8-10 hours.
I'd try 101Mhz for PCIe bus first. Also I'd recently found out don't use the Nvidia IDE drivers. Use the microsoft ones.Quote:
Originally Posted by Almost Famous
How does that make anything more stable ? The 101Mhz PCI-EQuote:
Originally Posted by ewitte
Sometimes if its left at default (on some motherboards, even still) the PCI bus might not get locked. Hard drives can be tempermental when the PCI bus is running out of spec. For the motherboards that are tempermental usually upping the frequence of PCIe causes the lock to kick in.Quote:
Originally Posted by RPG Junkie
Eric
Amen to that! i found out the issue with my old NF2 board, and ever since i have never used the Nvidia IDE drivers.Quote:
Originally Posted by ewitte
A fairly new driver on nf4 was causing my system to reboot all the time. Usually with a "driver" type message or paging error. I'd be able to run prime for days but I would get these strange reboots usually doing the most simple thing. Like clicking the start button.Quote:
Originally Posted by kenrippy
the NF2 nvidia IDE drivers were causing bad burns when doing movies w/ clonedvd/anydvd. it would cause glitches in the movies. it took me a while before figuring it out.
thanks ewitte.Quote:
Originally Posted by ewitte
Been there done all that ...took me almost three days to figure it out.
A simple change to 2t in the memory timings did it.
For the record, I've never had a problem with the NVidia IDE drivers, as a matter of fact, the difference in drive benches between the MS & NVidia drivers are enough to keep me using the NVidia drivers.
I appreciate the thought ;)
Hmm. It does not seem to be an issue here. What kind of cooling do you have on that CPU? I've got mine on water and the temperature stays pretty low. Maybe the memory controller is under too much stress for the conditions presented.?Quote:
Originally Posted by Almost Famous
right now just stock, but I have it @ 250X9 for stock speed CPU until my nw hsf gets here tomorrow (still using stock HSF). Right now I'm at 99f, it gets up to 122f under full load encoding DVDs. It's not the HSF if I'm at stock CPU Mhz, it was a memory timing issue ;)Quote:
Originally Posted by ewitte