Stability, Orthos vs Memtest86+ vs Memtest

[Rotax]

I'm trying to get stability (100%) but running into some odd results.

E8400 C2D @ 3953 (465 x 8.5) with 1.328 v [loaded]
Asus P5Q Pro (version 0506 BIOS, 05/26/2008)
OCZ DDR2-1000 @ 930 (timings listed below for each test) with 2.16v (actual)

5-5-5-14-50 : memtest86 pass 4x [on test #5], hyper pi pass (512k, 1m, 2m), occt (mix) pass 1h+, orthos (mix) crash, orthos (mem) crash, orthos (cpu) pass for 45+ min

I then used MemSet to change memory timings (via windows).

5-5-5-18-54 : orthos (mem) crash

5-6-6-20-58 : orthos (mem) crash

5-6-6-24-62 : orthos (mem) crash, memtest (windows version, 3 instances of 800MB, 1 instance of "all unused") passed


By "crash" I mean, I get a Windows style error message, yet the app is STILL RUNNING (processing and passing). As soon as I OK the message the program is ended. The "crash" happens about 2-5 minutes in.

I may run MemTest86 longer to see if I get errors, just haven't had the time to let it sit n run.

I do have USB emulation enabled (I read this can cause some weird issues when running certain tests, cache tests). I ran into issues w/ the USB emulation orginaly on the Cache Test within Everest.


Any feedback welcome..


The CPU seems solid, mainly I'm trying to pinpoint this issue 100%..so I can focus on memory, or FSB voltage? or DRAM voltage?

The memory is suppose to be good to 1000 mhz @ 5-5-5-18-54..sooo..

 

[Rotax]

Just ran Orthos (Blend) again and got a STOP at 1h 33m.

Core 0:
Test 1, 560000 Lucas-Lehmer iterations of M212991 using 10K FFT length.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 1 hours, 33 minutes 32 seconds - 1 errors, 0 warnings.
Execution halted.

Unfortunately I have not documented weather its the same core or same test that its been erroring on.

 

[Rotax]

And running memory only I get the "crash", see screenshot.

Yet, the time keeps ticking, and tests keep running... /boggle

Screen Shot

 

[Markfw]

Obviously, you need to reduce your OC, give more voltage/vdimm/vcore, etc... You are not close to stable.

 

[Rotax]

So its not memory, its anything and everything?

And its not "unstable" either with a 3 day uptime w/ random stress testing and game playing... Thus, I'm trying to find the "THE" instability.

I understand having 1 out of 6 apps throw an error is "unstable" but.. was shooting more so for whats unstable..

Pondering FSB...will play more later.

Ty for a reply.

 

[Idontcare]

Originally posted by: Rotax
So its not memory, its anything and everything?

And its not "unstable" either with a 3 day uptime w/ random stress testing and game playing... Thus, I'm trying to find the "THE" instability.

I understand having 1 out of 6 apps throw an error is "unstable" but.. was shooting more so for whats unstable..

Pondering FSB...will play more later.

Ty for a reply.

This is your response to an input from one of this forum's most notable overclocker? :shocked:

Mark is giving you solid feedback. Maybe the fact he didn't use many words should be an indicator to you of just how far off the "mark" you are with your views of a stable overclock versus being at a point of isolating the failing component.

My take on your situation is basically the same as mark's. Your system's errors (as reported by you in this thread) are so numerous in scope of likely suspects that the simple matter is your system is hardly stable and until you back off on your overclock and limit the number of variables in your OC instability there won't be much help the pro's here can give you.

I'd recommend reducing your overclock across the board (keep volts the same, just reduce the clocks via reducing FSB) until you get to the point that you can run Orthos error-free on small FFT for 12 hrs and large FFT for 12 hours. Then increase your FSB slowly (5MHz increments, same volts) while retesting small FFT.

When you reach a point of not passing small FFT error-free for 12hrs then its time to figure out if it is mem (Vdimm) or FSB (MCH) or CPU (Vcore) by changing just one variable at a time and retesting with small FFT and large FFT.

There are numerous guides in these forums on how to know when your overclock needs help, I'm giving you very abbreviated help. Best of luck.

 

[n7]

Run the full Memtest86+ a few passes, not just test 5 (though for initial quick testing that's fine).

I suspect that it's not vDIMM or CPU voltage you're lacking in so much as perhaps PLL or FSB Termination voltage, or maybe NB voltage.


And get Prime 95 25.6 just in case Orthos is the culprit.

Run P95 small FFTs with "Round off Checking" checked under "Advanced" for a few hours.

As i mentioned though, you're likely going to need to mess with more voltages that just CPU & RAM.

 

[VirtualLarry]

If you are using prime95 torture tests, you dont need to check off round off checking.

 

[Rotax]

Originally posted by: n7
Run the full Memtest86+ a few passes, not just test 5 (though for initial quick testing that's fine).

I suspect that it's not vDIMM or CPU voltage you're lacking in so much as perhaps PLL or FSB Termination voltage, or maybe NB voltage.


And get Prime 95 25.6 just in case Orthos is the culprit.

Run P95 small FFTs with "Round off Checking" checked under "Advanced" for a few hours.

As i mentioned though, you're likely going to need to mess with more voltages that just CPU & RAM.

Thank you. I did end up playing with these voltages and having better luck now. 8 hours with Orthos (blend). 2 hrs OCCT (mix).

I have run P95 as well, successfully.


I've actually read through the stickied guide here as well...and no mention on what may cause Orthos to simply "crash", nor other threads I searched through..aside from the obvious, unstable.

After looking into it more, I think the issue is/was with "auto" voltage settings are much higher than default..or much higher than needed for "default" clocks. I found a review someone did checking voltages directly w/ a DMM indicating this.

Thank you for comments and suggestions, I think I'm on the right track.
Thanks for the help guys.

 

[Acanthus]

Originally posted by: Rotax

Originally posted by: n7
Run the full Memtest86+ a few passes, not just test 5 (though for initial quick testing that's fine).

I suspect that it's not vDIMM or CPU voltage you're lacking in so much as perhaps PLL or FSB Termination voltage, or maybe NB voltage.


And get Prime 95 25.6 just in case Orthos is the culprit.

Run P95 small FFTs with "Round off Checking" checked under "Advanced" for a few hours.

As i mentioned though, you're likely going to need to mess with more voltages that just CPU & RAM.

Thank you. I did end up playing with these voltages and having better luck now. 8 hours with Orthos (blend). 2 hrs OCCT (mix).

I have run P95 as well, successfully.


I've actually read through the stickied guide here as well...and no mention on what may cause Orthos to simply "crash", nor other threads I searched through..aside from the obvious, unstable.

After looking into it more, I think the issue is/was with "auto" voltage settings are much higher than default..or much higher than needed for "default" clocks. I found a review someone did checking voltages directly w/ a DMM indicating this.

Thank you for comments and suggestions, I think I'm on the right track.
Thanks for the help guys.

It actually isnt crashing, thats what orthos does when it fails a test.