• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Freezing- finding the problem

G

gmreplay

Weaksauce
Joined
Dec 27, 2010
Messages
92
I've experienced total system freeze ups recently when my system is under strain in a visually intensive game. I figured the issue was the GPU, and after some testing in Furmark it appeared that the GPU was indeed overheating and crashing.

However, I just got a new GPU tonight (HD 6950), and I still experienced a freeze up in a game. I decided to run a Prime95 test to get a better grasp on what's happening, and almost instantaneously my Worker #1 is failing, with the message "FATAL ERROR: Rounding was 0.49609375, expected less than 0.4." This is very odd, since my system has been totally stable for a couple of years now.

My system is
E8400 overclocked to 3.6 GHz
4GB OCZ Reaper PC2-6400
Radeon HD 6950
Asus P5N-D motherboard
650w Corsair PSU

My bios settings are
vcore 1.325
dram 2.11
ht 1.54
nb 1.54
sb auto
ram 4, 4, 4, 15, auto


Where do I even start to diagnose this problem?
 
Silent Knight said:
Changing timing to 5-5-5-15?
Click to expand...

Interesting. Why would this help?

I increased my Vcore from 1.325 to 1.3625 and ran another Prime95 test for a full hour, which saw no errors. What is the proper test to run for Prime95? Small FFTs, large FFTs or blend? I've looked up some guides, but none were terribly clear.
 
gmreplay, I'm in the midst of an OC myself on a Q9300 and have been doing a lot of research lately. Disclaimer: I am in no way an expert here so I don't have a lot of personal experience to share, but have been reading a lot lately here and around the web, so most of the following info is 2nd hand, but believed to be reliable.

If the instability is recent, have your temps gone up at all? Cleaned out your CPU heatsink lately?

People's opinions on the use of Prime95 vary greatly, and it seems most users have moved on to intelburntest or similar. The general consensus on Prime95 seems to be that if you can run it for 24 hours without errors/freezes/reboots then you have a stable overclock.

> The small FFT test should only stress your CPU and on-die cache, taking RAM out of the equation.
> The large FFT should mostly test RAM, NB, FSB, etc (larger FFTs require RAM to be brought into the equation to crunch the numbers).
> The blend test will stress both CPU and RAM, depending on the size of the FFT being tested .

I like to use the blend test and run it overnight (10-12 hrs) to screen my overclocks- if they can pass that then I move forward. Additionally I try to game for an hour or so beforehand. People have reported running Prime95 for hours without errors yet having problems during games, so Prime95 cannot be relied on completely for determining a stable OC.

I *think* that a single worker failing during Prime95 is usually caused by RAM, but not positive. Probably depends on the size of the FFT at the time of failure.

I have seen posts where it is believed that overclocks can "age." Perhaps this is your issue, and maybe backing off to less than 3.6 GHz will solve your problem. Alternatively, maybe due to age you need more vcore to maintain a stable 3.6 GHz.

Relaxing RAM timing can't hurt, and I have definitely found that I have had to do that to move beyond 400 MHz FSB. Why you would need to do that after years of stability I don't know, but it is another variable to try. You could also run memtest to see if you get any RAM related errors outside of Windows/Prime95.
 
Awesome answer caustic, thank you so much!

I'm going to run intelburntest and I might back off on the overclock a bit, though for now giving the vcore a little more juice seems to have helped. I haven't kept an eye on my temps in years, when I tested the overclock and found it to be reliable and safe. They look to be within safe limits right now, though, even under stress.

How do you recommend I go about cleaning the CPU heatsink? Just blowing compressed air into it? It's an arctic cooling freezer pro, if that helps.
 
gmreplay said:
Interesting. Why would this help?

I increased my Vcore from 1.325 to 1.3625 and ran another Prime95 test for a full hour, which saw no errors. What is the proper test to run for Prime95? Small FFTs, large FFTs or blend? I've looked up some guides, but none were terribly clear.
Click to expand...

When overclocking, including "re-overclocking", the first thing you want to take out of the equasion is the ram/fsb, so you can focus exclusively on the cpu. Then, move on to the cpu next. Then, once you have the cpu/voltages re-locked in, go back to your ram/fsb for the final touches. It's just a matter of having a logical/methodical list of overclocking steps.....to keep things simple (it can be hard enough, as it is).

I personally have experienced loss of overclock, w/ time....2 years in my case. And in my case, it was the ram timings/voltage that needed to be fixed.....my cpu was still ok. For you, it could be either/neither.

.
 
gmreplay said:
How do you recommend I go about cleaning the CPU heatsink? Just blowing compressed air into it? It's an arctic cooling freezer pro, if that helps.
Click to expand...

Yup, those little fins can collect a lot of dust and then reduce the cooling capacity of the unit. But probably would have to be really bad to raise temps enough to cause problems. Depends on how close to the edge your temps are.

And as dekruyter mentioned, try to take notes on the changes you make and then the results of those changes. That way you can logically try to troubleshoot the issue. I use physical notes and have a spreadsheet in googledocs to track my progress.

Good luck.
 
I was hitting 79C with my Vcore at 1.3625, and around 73 at 1.325. I killed the overclock and set the board back to its standard settings, and I'm not not getting any higher than 66C under stress. I wonder if I should just ditch the overclock altogether.
 
gmreplay said:
I was hitting 79C with my Vcore at 1.3625, and around 73 at 1.325. I killed the overclock and set the board back to its standard settings, and I'm not not getting any higher than 66C under stress. I wonder if I should just ditch the overclock altogether.
Click to expand...

Sounds to me like you either need to seriously clean the dust out of your computer or you need to remove the heatsink, clean off the probably super crappy heatsink compound and apply some good quality stuff and reinstall the heatsink.

What heatsink are you using currently?
 
cyclone3d said:
Sounds to me like you either need to seriously clean the dust out of your computer or you need to remove the heatsink, clean off the probably super crappy heatsink compound and apply some good quality stuff and reinstall the heatsink.

What heatsink are you using currently?
Click to expand...

Arctic cooling freezer pro 7 with Arctic Silver 5. Not crappy components, just crappy performance.
 
gmreplay said:
Arctic cooling freezer pro 7 with Arctic Silver 5. Not crappy components, just crappy performance.
Click to expand...

True, but still probably good advice to give the PC a thorough cleaning and remove and re-seat the heat sink if it hasn't been done in a few years. If nothing else it will remove a variable.

Temps are another area where opinions seem to vary widely (what's normal, what's safe, what's too high, how to measure, etc) but IMO that 66°C on stock speeds (stock Vcore?) seems high.

For fun you can do a google search on "E8400 temps"; just take the information you find with a grain of salt as the necessary info to understand the posted temps is not always provided.
 
You must log in or register to reply here.
Back
Top