Memtest86 3.0 VS. Sandra Memory Burnin

[stompah]

Rig specs:

8k3a+
1700+
Spectek 256mb pc2700 (beleived to be cas2.5)
WD 30g 5400 RPM
Windows XP

For this discussion all timings will be at the MB's 'Turbo' Stettings.


When at 180fsb and 2.8v to the memory Sandra's Memory Burn-in program crashes Windows and causes a reboot. This happens within 30 minutes of testing.

When at 180fsb or 185fsb and 2.8v to the memory Memtest86 3.0 finds zero errors.

Which program is the one to follow? Talking to a few people have led me to ask different questions.

1. IFMU, Whats the difference between a memory testing program and a memory burn-in program? What method does each use to test or burn-in the memory?

2. William, How does one program find errors when the other doesnt? I guess this question is an extention of the second question above.

And my own question. Exactly when can I call my memory stable? I can run Memtest overnight with memory @ 185fsb with no errors... But its likely if I run Sandra's program I will crash. Now the extra variable here is Windows XP itself. Could XP inherently have a flaw in it which will conflict with Sandra?

 

[Xevuhtess7]

ive been using memtest86 3.0..... never tried sandra's memory burnin. since sandra's tests dont involve "checking for errors" like memtest does, i'd assume that most of the crashes peple experience when their oc is too high might just be windows instability? i have no idea, im just throwing ideas out. i think that memtest is a better way to test because it does not involve windodws in any way. it is just my gut feeling though..... mabye we should try to get sisoft to make some kind of "error detection" like prime95 and memtest86 have, that would be really cool.

 

[schismspeak]

use prime95 and run the "torture test" under options overnight, if it gives errors your system is not stable, and just cause memtest doesnt give any errors does not mean it is stable it must be stable under windows because that is the enviroment you will be in anyway. download prime95

 

[stompah]

no errors at 180fsb with prime95.

 

[schismspeak]

well then i would just expect something is wrong with sandra, incompatibility maybe, anyway which version are you using of sandra? if your using 2003 try 2002 or the other way round and see if it crashes. And now you say at 180fsb no errors y did you go down from 185 to 180 does it give errors at 185 in prime95 or...??

 

[stompah]

2002

I am just jumping around with things. I am trying to find the fastest fsb with turbo timings and no extra voltage.

 

[FIZZ3]

It sounds somewhat rare that Sandra would crash, yet Memtest does not. However, it's possible that you're not running all of memtest's tests... so that a particular kind of failure remains unchecked, which in turn by pure chance causes Sandra to fail. In case you have not tested this, you could set memtest up to include all tests and then loop it overnight.

As for Prime95; this program is more of a CPU stability test, but also very good for memory if you set the proper amount of ram to be used. I use Prime95, memtest86 and 3DMark for my stability testing myself, where 3DMark is more of a formality and the others the true tests. I do think you need both btw.

 

[stompah]

an extra .10v solved this problem, didnt want to but I guess I had to.

 

[hipro5]

Try to run ONLY "test 5" of the Memtest-86 for half an hour and see what will happen......In test 5 your ram WILL have the MAX temp on it's chips......IF it can handle it ,then you'll be OK as for ram temp......In the meantime , touch your ram with test 5 of Mmetest-86 with your finger and then touch them(the chips) with any other program......

 

[Xevuhtess7]

when i use memtest the only settingsi change are i set it to do all tests, and set it to use all memory.