Do I have a Bad CPU (I7 970)

[Blitz1776]

So I was having alot of issues with BF3, so I went to make sure everything was good (also occosionally a blue screen once roughly every 5 days). I ran memtest on cpu it would constantly fail at around 60-70% (blue screen or restart). It did this even when I tested single memory modules. So on a whim thinking the program was bad I ran Memtest 3.5b instead of 4.0a (suppose to use all cores) and it passed without an issue (haven't finished testing other two sticks).

So the question is, whether the cpu is bad on one of the later cores, or is the program faulty. While I did have an overclock, it still passes prime 95 (oc'ed or not) and I made sure to revert the changes after I was running into issues to diagnose the problem.

If you guys have any situations or thoughts on the matter I would be extremely appreciative

 

[ViRGE]

AFAIK there's nothing wrong with Memtest. If 4.0 says you have a bad system, I'd trust it.

 

[amenx]

Memtest is solely for testing RAM isnt it? So how do you figure the CPU is suspect from running memtest?

 

[Blitz1776]

Virge I was just throwing it out as a possibility since, software can be incompatible, and it might lead to an alternate line of thinking.

Amenx: Memtest is a memory test and it also will test cpu's own memory (as far as I recall). In any case in 3.5v it only runs 1 thread and it will pass a full memory test without issue. In 4.0a it allows the use of multiple cores, and no matter what I did it kept constantly crashing. So naturally I"m led to assuming it's the processor (however rare that can be), or the software.

However I noticed that there is a 4.2 from Memtest86+ so going to give that one a shot instead.
(Passed with no difficulty)

Only thing else I found could be (With a single CPU it is not possible to drive multi-channel memory controllers at full speed making it impossible to detect some types of errors. Bandwidth related errors are usually detected with test #7.)
Course there is also (Detection of cache coherency and interaction errors - Multi core CPUs have logic dedicated to cache coherency and synchronization that is not tested or exercised when testing with a single CPU. In addition congestion and noise from overlapping multi-CPU memory references may expose errors that otherwise cannot be detected.

Implicit testing of all caches and cores - Although Memtest86 is a memory diagnostic, it implicitly tests the CPU and all levels of cache. Version 4.0 implicitly tests all CPUs and all caches.)

If anyone has any input I'm open to it, or maybe thinks it could be a software issue, or something else.

 

[MobiusPizza]

So do you get a crash on MemTest if you revert to no overclocking and default voltages?

Have you try other modes of Prime 95 instead of Blend?

 

[soccerballtux]

odd

 

[Slufa111]

This could be real easy....whats the blue screen code?

 

[BTRY B 529th FA BN]

You also want to be testing with this

 

[Blitz1776]

Yeah mobius it still will crash on memtest 4.0a whether it's overclocked on or not generally around 60-70% completion. In addition on prime 95 I had it test the small FFT's for about an hour or so, to try to get it to trip up even a bit.

Concerning in WinDBG BSOD codes in reverse order with it's reasons for crashing.

11/05 - Probably caused by : memory_corruption ( nt!MiBadShareCount+4c )
Kernel base = 0xfffff800`02800000 PsLoadedModuleList = 0xfffff800`02a45670

11/02 - Probably caused by : kerneld.x64 ( kerneld+4890 ) (aida64)
Kernel base = 0xfffff800`02810000 PsLoadedModuleList = 0xfffff800`02a55670

10/25 - Probably caused by : dxgmms1.sys ( dxgmms1!VIDMM_GLOBAL::ReferenceAllocationForPreparation+2b )
Kernel base = 0xfffff800`0284c000 PsLoadedModuleList = 0xfffff800`02a91670

10/22 - Probably caused by : ntkrnlmp.exe ( nt!ObReferenceObjectByHandleWithTag+22a
Kernel base = 0xfffff800`02868000 PsLoadedModuleList = 0xfffff800`02aad670


BTRY I'm also using your memtest for windows, though from what I"m seeing it's not going to have an issue to pass that, if it does I"ll ammend the post.

 

[Blitz1776]

I'm guessing by the lack of replies, that my original assumption was (sadly) correct ;(

 

[VirtualLarry]

AFAIK there's nothing wrong with Memtest. If 4.0 says you have a bad system, I'd trust it.

I've had perfectly fine systems bug out on Memtest86, the non-plus version. IMHO, it's buggy and should not be used.

 

[Dadofamunky]

I'm guessing by the lack of replies, that my original assumption was (sadly) correct ;(

I disagree. If it passes every other test you throw at it then I seriously would not worry about it. Weird stuff happens with these things. But if it does everything else without a hiccup I do **not** think you would have a case for returning or exchanging the chip "because it fails memtest 4.0."

 

[Blitz1776]

I was hoping for that response that it was software wise, but so many people treated memtest as bullet proof, that it made me hesitant to be insistent on that.