RAM Troubleshooting suggestions


Recommended Posts

I just switched from 32-bit and installed Windows 7 64-bit, yes I know long over due but never had an issue with 32-bit until I was getting lag after years of usage and it wasn't using 1.2GB of RAM out of my 4GB.

(Yes yes, I know 4GB and 32-bit?! Just never bothered, for the reasons stated above.)

When I installed Windows 64-bit PC's been crashing lately (never crashed with 32-bit for more than three years), with different BSOD messages.

First instinct is to test RAM, and what do you know, only a few sections are faulty.

I have 2 x 1GB Corsair DDR2 CM2X1024-6400C4

and a

1 x 2GB Patriot PEP22G6400EL

I'm aware that it's usually not a good idea to mix RAM but I got extra RAM from a friend and it has worked for years with no problem with 32-bit.

As soon as I switched to 64-bit Windows 7 started crashing a lot. About 6 times in a span of 5 days.

Now here comes the part that completely threw me off.

Even though it showed a faulty section in the RAM (didn't say which stick).

I ran two passes of memtest on each stick individually and they all passed memtest

I was like, well, that was unexpected haha.

Before I do some more testing I'm thinking the following:

Faulty motherboard memory slot which would be a huge coincidence now that I installed 64-bit.

Is dust a possible cause? I moved the RAM around before the install, it's moderately dusty, can that affect the contact? (Random memory of blowing NES catridges came to me as I typed this :D )

Plausible: RAM timings? (Though doubtful since I haven't had a blue screen since the Windows ME days)

That's all I can think of ATM.

I'm looking for other possible causes and suggestions on what steps I should do to troubleshoot this.

For example: Should I test each RAM stick on each memory port of the motherboard?

Here's a screenshot of said crashes: BlueScreenView%20%20-%20%20CWindowsMinidump.png

Link to comment
https://www.neowin.net/forum/topic/1026262-ram-troubleshooting-suggestions/
Share on other sites

I would read your motherboard manual on what it says about RAM placement & pairing.

I would put the BIOS RAM settings on Auto, if they are not.

I would check the seating of the RAM, they can become off, when you tinker around inside.

Windows 7 also has a proggy to test the RAM.

I would test each stick individually.

Excess dust does not promote conductivity.

Even though it showed a faulty section in the RAM (didn't say which stick).

No part of your RAM should have errors.

Replace the faulty stick.

No part of your RAM should have errors.

Replace the faulty stick.

Did you read the second bolded sentence? I tested each stick individually twice, and they all passed the test each time. It only happens when they're used together.

Also, what is the difference between the Windows 7 memory diagnostics and memtest?

Here is another thing you can try before replacing Memory:

Control Panel/System Propeties/Advanced tab.

Performance/Settings.

Advanced tab.

Change the virtual memory to No paging file.

OK, OK, and Reboot.

This will clear out the Page Memory.

After rebooting, go in and change it to System managed size.

Reboot again.

I forgot this tip that I read on the Microsoft site.

Thanks for the interesting tidbit Hum and helpful suggestions! If I recall correctly I had already the BIOS at default but I did it again just in case. I brushed up my NES cartridge blowing skills (ran out of canned air) and made sure to reseat them very solidly.

Ran memtest again and it passed with flying colors! I also ran two passes of Windows memory diagnostics and also passed.

Previous evidence strongly suggests that the RAM was fine. The current tests I did after writing this post points me to believe that some RAM was either loose or not properly mounted or dust or a combination of the two since the first time I ran Memtest within the first two minutes I'd immediately get errors.

Remounting the RAM seems to have fixed the issue.

To be 100% sure I'll see how the PC runs now that every test I ran seems ok. If something does happen I'll come back.

If I do come back, my only question would be... actually I think I'll make a poll.

Here: https://www.neowin.net/forum/topic/1026286-how-many-memtest-passes-is-good-enough/#entry594317914

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • Microsoft's fast coding model MAI-Code-1-Flash comes to Copilot Business and Enterprise by Karthik Mudaliar Microsoft’s recently announced MAI-Code-1-Flash model is now generally available to GitHub Copilot Business and Copilot Enterprise customers. With this support, organizations can have more centralized policy controls and billing while finally being able to use Microsoft’s lightweight, first-party coding model. According to GitHub’s announcement, Business and Enterprise plan administrators must enable the MAI-Code-1-Flash policy in Copilot settings before developers can access the model. Microsoft says that MAI-Code-1-Flash is for fast, iterative coding work rather than the most demanding architectural or debugging tasks. GitHub’s official model comparison page says that the model is great for "general-purpose coding and writing," while it excels at fast, accurate code completions and explanations Microsoft introduced MAI-Code-1-Flash on June 2 as part of a broader collection of internally developed MAI models. GitHub subsequently expanded support to Copilot CLI, the Copilot cloud agent, GitHub.com chat, GitHub Mobile, Visual Studio, JetBrains IDEs, Eclipse, and Xcode, but said support for managed Business and Enterprise customers was still on the way. In Microsoft’s own benchmark testing, MAI-Code-1-Flash scored 51.2% on SWE-Bench Pro, compared with 35.2% for Anthropic’s Claude Haiku 4.5. Microsoft also claimed that the model used up to 60% fewer tokens on SWE-Bench Verified. Do note that these are vendor-run results rather than independent measurements. The model is billed at provider list pricing under GitHub’s usage-based system. GitHub currently lists MAI-Code-1-Flash at $0.75 per million input tokens, $0.075 per million cached input tokens, and $4.50 per million output tokens. For organizations, the main incentive to use MAI-Code-1-Flash is likely to be efficiency rather than maximum capability. A smaller model that responds quickly and limits unnecessary output is quite useful for repetitive agent tasks at scale, especially after GitHub Copilot’s move toward usage-based billing. The "Flash" model is recommended for fast work and not necessarily for huge repositories with loads of context. It's better if teams compare their output with other larger models, especially if they're working on security-sensitive changes and complex, multi-file work.
    • yes AND no the "original" or plain/normal Optiplex 7010 won't be getting any more new firmware updates BUT the Optiplex SFF/SFF Plus {small form factor}, Micro/Micro Plus & Tower/Tower Plus 7010 editions DO get new updates such as this new one   and here are similar guides from the Dell web site for Dell systems: https://www.dell.com/support/kbdoc/en-us/000390990/secure-boot-transition-faq https://www.dell.com/support/kbdoc/en-us/000347876/microsoft-2011-secure-boot-certificate-expiration
    • AT&T has been spying on US citizens with the NSA for decades.. they just know how to keep it more under wraps.. the evil level is still there.
  • Recent Achievements

    • One Year In
      bernmeister earned a badge
      One Year In
    • Week One Done
      Scoobystu earned a badge
      Week One Done
    • Week One Done
      tuben earned a badge
      Week One Done
    • First Post
      OffsetAbs earned a badge
      First Post
    • Reacting Well
      OffsetAbs earned a badge
      Reacting Well
  • Popular Contributors

    1. 1
      +primortal
      462
    2. 2
      +Edouard
      213
    3. 3
      PsYcHoKiLLa
      157
    4. 4
      Steven P.
      72
    5. 5
      FloatingFatMan
      71
  • Tell a friend

    Love Neowin? Tell a friend!