Windows 7 Memory Test - Inaccurate Results

I had a new system build (Gigabyte GA-G41M-ES2H, 4GB Corsair xms2 DDR2 6400, Windows 7 Professional OEM 32-bit) that began showing the Blue Screen of Death about 2 weeks after it went into service.

As part of the troubleshooting process, I loaded the Win 7 repair disk and ran a memory check. When the system rebooted and the memory test started, I received a "Hardware Problems Detected, contact the system manufacturer" message. I let the memory scan finish; no memory errors were detected.

I swapped out the motherboard with an identical model and replaced the hard drive, just to be on the safe side. Once again, BSOD kept popping up 1-4 minutes after booting!

Second Win 7 memory test on the new board. Same "Hardware Problems" message; once again Win 7 found no issues with memory.

Now for the punchline: Since I had replaced everything but the memory, I took out the Corsair and put in 2x2GB Patriot 6400. I'm typing this post 45 minutes into rebooting the PC with no more problems.

The moral of the story: Do not trust the memory test results from a Windows 7 repair disk. In my case, it said the memory was fine and the hardware was FUBAR. In reality it was the exact opposite.

Four hours of my life I'll never get back... :tas

 

Thanks for the heads up man! I always use memtest86+ to test memory, available on main page here. Was your original memory on the QVL?

 

This was one of four identical builds for a client, all using the same type/model of Corsair memory. Since this is (hopefully) the only one that crapped out, I'm calling it as a bad memory module and getting an RMA from Corsair.

As for QVLs, Murphy's Law of Memory states the place you buy the mobo from will not have any brand or model of vendor tested memory in stock. rolleyes

My experience with Gigabyte's 775 and AM2/AM3 boards (around 20 in the past year) has been that they'll work fine with whatever brand Fry's has on sale that week.

 

I just want to confirm the worthlessness of the Windows7 memory test. For months I've been fighting occasional crashes, sometimes several a day. I eliminated most hardware, KNEW rolleyes the RAM was OK, since it passed the Windows 7 memory test. Yesterday I ran across mention of the new memtest86 v 4.0 and figured I might as well try it, though I expected it would pass. Well, it failed and indicated a single bit was stuck on '1', couldn't be written to '0'! A few minutes swapping RAM sticks isolated it to a bad one and I'm now running with 4 GB instead of 8. I'm sure glad I didn't replace the motherboard, which I had tentatively decided would be the next step.

Thanks, Microsoft, for the nasty ride. If you can't write a memory tester that actually finds the errors, just give it up!

 

I just came across this thread so sorry for joining so late. I must point out that NO memory tester is perfect, especially software based testers, and therefore you should not rely on just one. Note that I have had RAM pass Microsoft's testers (the downloadable version and Windows 7 version), DocMemory and Memtest86+ but still fail when installed. I have had RAM pass all tests, run great when installed individually, but fail once paired with another stick that also passed all tests.

That said, I have never had any of those software based testers report RAM as bad that was really good. That is, if they report the RAM is bad, then it is bad. In that respect, they are like Power Supply testers. They cannot prove a supply is good, but they can prove if it is bad.

As noted above, swapping sticks is about the best way to verify if RAM is good or not. There are hardware devices for testing RAM that are excellent and very reliable, but the investment is not worth it for most users.

 

Digerati, I agree there is no such thing as a "perfect" memory tester.

My point in this post (which is validated by WallaPatrick's response) is the Win 7 memory tester is especially bad. If it had merely given the memory a false "pass" score this would have been excusable. What set me off about it was the fact it instead diagnosed the motherboard as bad - not once, but twice in a row.

This happened about 45 days after I purchased the first board. Had I not purchased the board from a local supplier that offers a no questions asked in-store replacement during the first year, it would have cost $10-$15 bucks for return shipping and a few weeks of wait time to get an RMA through Newegg. If I had done the RMA through Gibabyte, it would have been 6-8 weeks (been there, done that). This would have been worse since the client would have been p*ssed during the weeks the PC was out of service.

Windows 7 Memory Tester is definitely a FAIL. An false passing grade or a "Windows Can't Figure This Out" answer would not have triggered the hassles of replacing the motherboard

 

Well, then, by the same criteria, Memtest86+ and DocMemory are a FAIL too. There are plenty of posts out there where folks have tested with MemTest86+ and found no errors, yet the problems were resolved only after replacing the RAM.

I am sorry you had problems with Windows Memory tester, but you merely illustrated the problem with all software based scanners - whether it be memory testers, or malware scanners, no one tool is perfect.