Help Needed: Orthos + e4500 + Vista 64 -> Random restarts on blend

melbourne720

melbourne720

melbourne720

Associate
Joined
26 Sep 2006
Posts
1,945
Location
West Bridgford
Hi there

I have built an e4500 based system, and am overclocking it in Vista HP 64 bit. Trouble is, after about 3+ hours of orthos blend, I return to the PC to find the machine has rebooted. Spec is

e4500 (cooled with a thermalright 120 Ultra + nexus fan)
MSI P35 Neo2-FR
OCZ 4GB (2x2GB) PC2-8000C5 1000MHz ReaperX
8800 GT
Corsair 520W

Bios is:

CPU Vcore: 1.4V
Memory: 2.1
VVT FSB: 1.4
NB: 1.4
Spread Spectrum: Disabled
FSB: 290 MHz
Multi: 11
CPU: 3190
Memory is 1:1 (underclocked)

I had been doing 30 minute orthos blend runs from stock to 3.2 at every 10MHz on the FSB with no problem.

The Twist:
It runs fine if I run orthos in 32 bit mode!

1. Is this because the 64 bit functionality of the CPU is failing, and I need to up the vcore?

2. Will I have to do 8 hour runs for every clock?

3. And why is restarting the PC, rather than just orthos failing?

4. Should I do a full memtest before proceeding...?

many thanks...
 
When you say that it works in 32bit mode, how long have you run the tests for?

My personal opinion would be to run memtest for a few hours to complete a few passes, I know you say it is underclocked at present but faulty sticks can produce errors even when running below stock.

Also I would run a 8hr small fft with the cpu, at least when you have that result in and the memtest result in it will remove the cpu and ram from your trouble shooting.

From my own experience, having orthos/prime restart the machine and not just throw up an error means that one of the voltages is too low or you are at the max of your chip, seeing as you are at 3.1 I would say its the voltage.
 
Thanks for the reply.

I ran it for 4 hours in 32 bit mode, no errors.

Will do tonight:-

1. Memtest, few hours, a couple of complete passes.
2. 8 hr small fft with the CPU.

Any thoughts on which voltage could be too low?

I'm also reading about OCZ memory in MSi boards causing problems, hope that's not the case here.
 
I would choose the 266 strap first and then if you have no luck try the 333, on my P5K-E I am using the 266 (1066) strap. You normally find that you have to keep raising it the more you raise the fsb.

I wouldn't think you could be pointing the finger to the 64bit OS if the difference between a pass on 32bit and a fail on 64bit is only an hour.

Its a good idea to do those tests like you said, its one way of eliminating things from the equation, I would try 1.4vcore at first but you may need to up it a bit to get 8hrs of small fft.

Not sure about the OCZ/MSI combination, most of the time configuration problems like that would stop the computer from booting and give you loads of issues actually getting it to work. If you have done runs of orthos I am guessing you will be all right with your setup.
 
Last edited:
Update:

Put the strap onto 266.
Memtest for 5 hours, 20 minutes; 8 passes, not a single error.

Result: I think the memory is okay :)

Next step: 8 hr small fft with the CPU.
 
melbourne720 said:
Update:

Put the strap onto 266.
Memtest for 5 hours, 20 minutes; 8 passes, not a single error.

Result: I think the memory is okay :)

Next step: 8 hr small fft with the CPU.
Click to expand...

Good job mate, it's always such a good feeling to return to the machine and see it's still working away without errors :)

I was in a similar situation as you not long back, my cpu passed 8hrs small fft at 3.42GHz and my memory passed memtest but it was failing blend.

In the end it was NB volts set too low at 1.25v, I upped it to 1.4v and all was well, maybe the case with you too... what voltage options do you have for your NB?
 
From 1.1V upwards in 0.025 steps I think. I have it on at 1.4 at the moment, see how it goes.

I hope it was just the strap being 200 while simultaneously trying for 290 FSB that was the issue. Shall see later today when I go and check on it :)
 
This is looking more promising...

10hrsorthos.jpg
 
Good results Melbourne, seeing as you have now passed a nice session of small fft's, have done passes on memtest and are already using 1.4v on the NB it does look like it could have been the strap that was given you issues.

You can clock the fsb higher than what the strap says (i.e fsb can be more than 266 when strap is 266) but maybe your 290 fsb from the opening post was too high for the 200 strap.

You still may be able to use 290fsb x 10multi instead of 266fsb x 11multi on the 266 strap though, I am using 380fsb on my board and that can be done on both the 266 strap and 333 strap.
 
Last edited:
Am retrying at 290 FSB, but only 7 multiplier. This should ascertain as to whether the FSB needs more voltage or the CPU.
 
Last edited:
Fair question, my thinking being...

If it passes at 7 x 290, then I know that the FSB strap (currently at 266) is fine and the northbridge / FSB voltages are fine. Which probably means more VCore is required to get the CPU stable.
 
You must log in or register to reply here.
Back
Top Bottom