Prime Tests Failing - Need Help

Hello. I hope someone can help..

Machine Specs:
Core2Quad Q9550 E0
ASUS PQ5-SE2
4GB Kingston Hyper-X PC6400
EVGA 9800GT 512MB
CoolerMaster Elite 335 case w/ 120mm Antec intake and exhaust fans
Arctic Cooling Freezer Pro 7 rev. 2 CPU heatsink
Corsair 550W PSU
Windows 7 x64

I can run Small FFT testing overnight without a failure.
This test ran last night for over 9 hours and it was still running this morning.

If I run Large FFT a random core will error out without the first few minutes and usually display the error: "ILLEGAL SUMOUT".

When a Blend test is ran it also fails without the first few minutes. An error "Rounding was 0.5, expected less than 0.4" comes up.

My BIOS settings are all on AUTO currently (this is my first homebuild and I just finished putting it together).
The only thing I changed after some research before my build was changing the CPU Margin Enhancement from 'Optimized' to 'Performance Mode'.
I read in these forums and others that this 'Optimized' setting was causing problems for people with a P5Q and the Q9550.
I should point out however that I ran these tests with this setting on 'Optimized' and 'Performance' with the same results in Prime.

My core temps are very stable at idle and during testing.
Idle core temps are 29C-30C.
During testing they reach 40C-41C.
My core voltage (according to CPUZ) is 1.112V during testing.

Does anyone have a suggestion as to what may be wrong here?

Thanks in advance!

Have you checked your RAM settings? A failure with your RAM can cause a failure on Prime95.

I recommend checking your settings in your BIOS for your RAM and make sure it matches the manufacturer specifications. Run MemTest86 for seven passes to make sure you have not faulty RAM. Retest Prime95 and see what happens.

I would try 1.8 to start and see what happens. If that doesn't work, try 1.9 for the voltage.

I reran Windows Memory Diagnostics with voltage at 1.8 and 1.9 - both failed on me within 1 of 2 test passes.

So, I went into my stash of computer stuff and put in a different pair of DDR2 PC6400 RAM that has traditionally worked for me no problem, reran the test and that also failed but on pass 2 of 2.

Yes, I did. Still failed Windows Mem Diagnostics as well as the same Prime failures..

That might be it..
I shutdown, shuffled the memory to the other slots (I'm using two DDR2 2GB sticks), changed my BIOS settings to AUTO (for the time being) and ran the Win Mem Diag and it went through without any errors.

I just booted into Windows and will run a Blend test in Prime while I do a few things around the house and see what happens.

Update:

I had to go to Micro Center to return an IDE-to-SATA convertor and while I was there I bought some new memory:
8GB of OCZ DDR2 PC6400 CL4.
(Got them for $69 each kit of 4GB- figured it was a good deal with the prices of DDR2 right now)
I picked up some CDs too so I could burn MemTest and try it with NEW memory.

Got home, burned MemTest86 to a CD and ran some tests with that.

Immediately errors were coming up one after the other.
I was on test #2 and had 100+ errors all ready.
Remember, this is new memory too..

This was with timing and voltage set to AUTO.
I went into the BIOS and set the timings and voltage to the manufacturer specs and still same result.

After several different BIOS adjustments and testing individual slots and such, I think I may have found the solution (or close to it)..

I changed the Northbridge voltage from 1.1v to 1.2v and thus far I've run MemTest86 with only 2 errors.
Those errors popped up on Test #5.

For the sake of documentation, the errors are:

Test- 5
Pass- 0
Failing Address- 001d1bb7518 - 7451.7MB
Good- 00008001
Bad- 00000001
Err-Bits- 00008000
Count- 1

Test- 5
Pass- 0
Failing Address- 001d5bb74f8 - 7515.7MB
Good- 00008001
Bad- 00000001
Err-Bits- 00008000
Count- 2

Right now, MemTest is on Test #8 and has been running for over 55 minutes.
Those are still the only two errors that have come up.

I'm feeling much better about things now.
I will try to change a few more settings to eliminate those errors.

Glad to see you discovered a solution & your system is running stable now (for now on hopefully!!! )