Strange PC problems - and their solutions!

[Gordon Laing]

Hi everyone, we all experience problems with our computers from time to time, but sometimes the crashes really are quite severe, but bizarre and inexplicable.

These problems can be the hardest to diagnose, but equally the most satisfying to resolve, so this thread is for anyone who's come across a really weird problem with their computer AND managed to fix it. You never know, it may help someone having the same problem.

I'll start the ball rolling with this one:

Several years ago, my perfectly healthy desktop PC suddenly began to act really strangely. Random crashes which completely hung the system, requiring hard resets. I hadn't changed anything on the software configuration, so assumed it was a hardware problem. As all of you know, it then becomes a trial and error process, involving swapping-out various components for spares - if you have them - to see if the problem goes away.

I've seen some nasty crashes before involving over-heating, but this time, the culprit was a memory card which had gone bad. There's no way I'd have known without swapping the cards for spares, and to be honest at the time it was the last thing I thought of. Surely memory doesn't fail, right? Well it does, and after identifying the faulty card and swapping it for a new one, everything went back to normal.

The reason I mention this issue is that it happened again this weekend, only this time with a laptop. Again the laptop had been working fine, and again hadn't had any software updates or installs for some time. But all of a sudden, it regularly suffered from random, major crashes which completely hung it. Basic applications were generally fine, but video playback or skype would result in a big crash after only a few minutes.

Luckily I had a spare SO-DIMM card I was going to use to upgrade another laptop, so swapped it-in just in case. And you know what? The problem completely went away. It was another memory card which had gone bad.

So if your computer suddenly starts crashing or freezing for no apparent reason, I'd recommend giving the memory cards a check. It's now one of the first things I'll check following consistent severe hangs.

Over to you guys now - tell us about your weirdest computer problems and how you fixed them! It could be failed hardware. It could be a cooling device that became clogged-up. It could be a seemingly innocuous software or driver update for what you thought was an unrelated device. It's time to reveal all!

Gordon

[popo]

If you have reason to suspect the ram is faulty, or possibly may be operating outside its limits, give memtest86+ a try. It most easily burns to a CD and you boot off it, where it will run test patterns on your ram to make sure every bit it can access is good. I use it when pushing systems beyond their limit, but also for general checks for example after a new build for confidence. If you find errors then you can try diagnosing if it really is a hardware fault, or possibly an incorrect setting elsewhere.

The most common problem I've encountered is the dying PSU. This has various symptoms, but typically manifests itself when a system that is otherwise fine sometimes wont start up without several retries. I guess this is because the initial surge on boot pushes it just over its limit and doesn't go anywhere. If it should make it start, the load then is lower and it is happy, just. Other hints may be system instability when loaded. I've had this happen on various desktops and even one laptop power brick.

[keystrokesuk]

A common fault I have come across where there seems to be a ghost in the machine is a faulty keyboard. You click and expect a selection to be made but a menu pops up or something else unexpected happens.

When this happens disconnect the keyboard and try just using the mouse. If things are OK with just the mouse you have a faulty keyboard. I have experienced at least 3 reasons for a faulty keyboard;
1) A file or other object is leaning on the keys
2) Some liquid or a paper clip has been dropped into the keyboard
3) The keyboard is actually faulty

On the Mac OS there is a handy utility called keycaps that shows you which keys the machine thinks are being pressed using this is a quick way to check you keyboard is OK.

[Bob Andersson]

..
............................Resolved
Hi folks,

This turned out to be a bit of a saga with several missteps along the way. Rather than just present the solution and appear cleverer than I am here's the full story on the principle that it can sometimes be as helpful to describe what didn't work as well as what finally did work. First I'll set the scene...

My main PC (Windows 7 Ultimate x64) had recently decided to take much longer than usual progressing from the Password screen to the Desktop. Sometimes it would be quick enough but sometimes it might take over a minute. Initial scrutiny of the Event Log showed that Windows was seeing the delay but not providing anything definitive about the cause.

For the full horror story read on or, if you want to find out how it ended, you can just skip to the penultimate paragraph.

My first thought was to run "CHKDSK C: /f /v" and "SFC /ScanNow" from an elevated Command Prompt but no errors were found and startup performance remained unchanged.

Then I used cCleaner to clear out some unnecessary startup programs, also to no avail. By the way, Microsoft's own "System Information" app is worse than useless at telling you what runs on your system at Startup.

Given that this is recent behaviour the obvious candidates are software updates (Microsoft or third party) or an impending hardware issue. I discounted the latter because once the Desktop is up and running the computer can run as normal all day without further issues.

Anyway, I then "hid" the problem by applying a Registry "band aid". I added the 32 bit DWORD value "DelayedDesktopSwitchTimeout" to the key "HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\System" and set its value to 5 (seconds). This fix is described in Microsoft's KB940452 for a slightly different issue but it worked for me in that I got to the Desktop in a reasonable time. As for why the band aid works, there are a number of Google results when searching for DelayedDesktopSwitchTimeout. None of them seem to relate directly to my problem but when one reads that even something as trivial as setting the Desktop background to a solid colour can induce a thirty second delay in progressing through the "Welcome" screen one wonders exactly what sort of hallucinogenic substances the Microsoft programmers may enjoy during working hours.

The trouble with band aids is that you don't know whether the wound has healed until you remove them so a couple of hours later, and after an hour of vigorous exercise which pumped an extra red corpuscle or two through the little grey cells, I decided to use cCleaner's "Cleaner" tab at it's default settings, apart from disabling the removal of my Firefox cookies, and it cleared out 484MB of "stuff". Then I removed the "DelayedDesktopSwitchTimeout" value from the Registry and rebooted. Restarted a couple of times and the transition from password entry to Desktop display took about 12 seconds. Success, I thought, until this morning when it took well over a minute to get from Password entry to the Desktop.

So it was back to the Event Viewer to try and see what was going on. I created a Custom View to show me all Critical and Warning events. No critical events were showing up but I was seeing a lot of "Error 0x80070002 occurred while verifying known folder" events (ID: 1002) at each start up, those errors first appearing back in September. These errors were confined to subfolders of "C:\Windows\system32\config\systemprofile\". Firing up Windows Explorer and enabling viewing of System Folders allowed me to drill down and find that, indeed, none of the folders being complained about actually existed. My next port of call was the Registry where I found matching references to each missing folder under two "FolderDescriptions" keys, one for 32 bit and one for 64 bit. Bravely, and with the benefit of too little knowledge , I decided that these Registry entries needed removal so that's what I did, including removal of all references to those entries but not before I created a manual Restore Point. On the next restart the computer progressed from Password to Desktop in good order so I though "problem solved". Mistake!

Another look at the Event Log showed the same error messages so it would seem that Windows has these folder paths stored for use when the registry entries are missing. A System Restore to the manually created restore point undid all the registry deletions I had made. Then I did the simple thing (finally!) and just created new Folders to match the ones that the Event Log was complaining that Windows couldn't verify. The only extra step I took as I created each one was to assign ownership to the Administrators to match the ownership of the parent "systemprofile" folder - I'm not sure whether that step was necessary but it seemed prudent.

By the way, cCleaner is an excellent tool but like all sharp objects it should be handled with care and lots of thought - "measure twice and cut once". Similarly, making changes to the Registry should only be done if you are 100% confident in the changes you are making.

It seems as though that last step has finally solved the problem. The last four restarts have got me from Password to Desktop in about twelve seconds each time and the Event Viewer is no longer complaining that it can't verify the previously missing folders. Of course those manually created folders aren't "special" insofar as their Properties are concerned but as they don't contain any files, and probably never will, I'm not going to worry about it. If tomorrow morning's start up is also normal I can relax. If it isn't then I'll be back to the drawing board, albeit with a few less event warnings to worry about, and I'll be adding to this already over-long post!
.
.
Oh dear, here I am again after another long one. The frustration is that these long startup times are intermittent so I can never be sure when the issue is sorted. I cleaned out the /Windows/Prefetch folder this evening: too early to be sure but I don't think that has helped. Next stage will be to uninstall Kaspersky's ISS, reboot and then reinstall it. If that doesn't help then I'll be uninstalling NVidia's driver and installing the next version which ahas just been "qualified".
.
.
Well, I'm back again so you know that everything I've done so far was rubbish! Well, not completely but it didn't solve the problem. But I had a major clue a couple of days ago when I decided to upgrade to the latest (qualified) NVidia graphics driver. When I do a manual graphics driver update I like to create a temporary Administrator account to avoid issues with my very complicated Desktop as the screen resolution changes. So after creating the temporary account I did a few Restarts into that account both before and after the graphics driver update and I was still seeing a slower than desired progression from Password to what was now a very simple Desktop on the temporary account. That was a big clue as there is very little major software that loads up during start irrespective of the User who logs on. Suspicion now fell squarely at the door of Kaspersky ISS 2011 but first I deleted that temporary account.

With Kaspersky ISS 2011 now firmly in the spotlight I uninstalled it completely (except for the Activation code) and rebooted. Voilà, straight from Password to my Desktop in about three seconds and I did another Restart just for confirmation. As I still have over six months to go on the Kaspersky licence I then reinstalled it. Installation seemed to go well and, as expected, it then asked for a Restart to complete. Again all went as expected and I wasn't interested in timing things at this stage with Kaspersky still settling down. Once at the Desktop I initiated a Kaspersky update. That went well enough and once it was flagged as "Complete" I did another Restart. Password to Desktop took about a minute after which Kaspersky blared out that its databases were obsolete. Bizarre given it had just Updated. Anyway, another Update and a manual Restart and things seem to be more normal though I'm puzzled as to why it seems to be so busy over the Internet immediately after Password entry as I don't allow it to check for updates during the first eight minutes after startup. The net result is that Kaspersky is on a very shaky ground with me on the credibility stakes right now - certainly if I see my Password to Desktop time misbehaving again it's out of here never to darken my door again. And when the licence runs out renewal is certainly not a given.

With hindsight maybe the solution should have been more obvious but there again why should I immediately suspect a program that has worked without particular issue for a long time? Reading the earlier parts of this post again I think the only significant mistake I made was in trying to edit out those "FolderDescriptions" registry entries. As for the rest, while they weren't actually relevant to the problem some of the actions I took may even have done some longer term good. In any event I certainly hope that my journey may be of help to anyone else who starts to see the same problem.

Bob.

Final Update: Kaspersky ISS 2011 is now removed - guess why! I've decided to continue the ISS conversation in the new thread Which Security Suite/Anti-virus package do you use?.

Update: Like all intermittent problems one can never be sure a solution has been found unless there's an obvious smoking gun and it turned out that the fix above nudged things in the right direction but the problem did come back to haunt me. However I'm hopeful that the latest solution, namely to change the Bonjour service from "Automatic" to "Delayed Automatic" may have finally put this issue to bed.

[Bob Andersson]

Hi folks,

Ever had Windows refuse to connect to the Internet after boot-up?

It started happening to me some while ago, not often but enough to be annoying. If I looked at the basic network information and connections I found that whenever this happened I'd see an extra "Unidentified Network".

Temporary workarounds included rebooting or disabling and then re-enabling the network adaptor. I finally got fed up with this and went Googling and came up with the post here. The post describes the fix under Vista but it worked equally well for me with Windows 7. There are two relevant posts, both of which need to be read, and as I can't find a way of linking to individual posts here are the two posts that helped me out.

I have the same problem with Vista and 2008 Server.

The issue is that the ROUTE table contains a phantom persistent default route 0.0.0.0 to one of the network interface cards. This activates the "Unidentified Network" (Public) every time the network connection becomes active. Sharing permissions are turned off.

My solution is to use the Windows Command Screen as Administrator and remove all default routes.

ROUTE DELETE 0.0.0.0

This removes the default routes from all network interfaces. You must then reapply the necessary default routes to your interfaces (Do not use address of the network adapter as the default). If you have an isolated network (without any router, firewall, or gateway) with more than one computer -- I do this in a development laboratory -- is to choose for default address for Vista, 2008, Win-7, to be the IPv4 address of any other attached network interface.

I don't know why Microsoft deciced to make the default routes 0.0.0.0 persist in the routing table instead of just being bound to the network adapter. The network stack used by XP, 2000, 2003 systems do not have this problem.


I hope this helps!

In response to the question "how do I reapply the necessary default routes to the interfaces":

Windows Vista

*** Open the "Network and Sharing Center" - there are several ways to do this.

Control Panel (Control Panel Home)
...Network and Internet
......View network status and tasks

Control Panel (Classic View)
...Network and Sharing Center

*** Examine the properties of the Local Area Connection

Connection "Local Area Connection" View Status

...or

Manage Network Connections
...Local Area Connection

*** Select Properties

-> Internet Protocol Version 4 (TCP/IPv4)

IF YOU GET YOUR IP DYNAMICALLY VIA DHCP
...The DHCP (Dynamic Host Configuration Protocol) server
...should be connected to the same switch or hub as your
..."Local Area Connection" interface adapter.

IF YOU MUST SPECIFY YOUR OWN IP ADDRESS
...Choose an IP address for your "Local Area Connection"
.........i.e. 192.168.2.101
...Specify a network mask (i.e. 255.255.255.0)

...Enter a default gateway IP address:
.........This is the IP address of your Internet router
.........or network access firewall.

.........If you are on a closed-isolated network, choose
.........any available IP address of something on your
.........local network (i.e. printer) BUT NOT THE SAME ADDRESS you use for the "Local Area Connection"

I'd advise against applying the described fix unless you are comfortable with such low level fiddling with your OS, particularly if you don't have Internet access from another computer which can serve as a lifeline should you have trouble re-establishing your network connections. At the very least before you issue the "ROUTE DELETE" command have a good look through the procedure described in the second quote, open up the various dialog boxes described and make a note of the way your machine is currently configured so you can enter the required information when you need to do so later on.

Update: Reactivation of this thread (next post) reminds me to post a further development. The fix described immediately above seemed to work for a while but the problem came back. BUT, since I've set the "Bonjour" service Startup Type to "Automatic (Delayed Start)" -- it was "Automatic" -- I've not had a recurrence!

Bob.

[popo]

Had a funny I think I've finally narrowed down now...

Today I'm using my new desktop as main desktop finally. During the build and setup, I did dabble with a little overclocking, but as a high end CPU there didn't seem to be any "free" headroom without turning up the voltage. Now and then I'd get a windows error requiring a reboot. There was an oddity in that all the usual stability tests never gave any error (prime95, linpack, memtest). I gave up and decided to run at stock, but I soon realised with prolonged load (many hours) there could still be lockups. This worried me as I've not overclocked anything now.

Cutting in to the suspect answer, I ran my high performance ram under its spec, and haven't had a crash since. It was XMP profile rated as 1600 CAS 8 1.5V. I've backed off to the fastest JEDEC setting of 1333 CAS 9 and all seems well now. Maybe I can tweak it up some more, but I doubt it'll be worth the time it'll take to prove. I did tell myself in the past to stop buying "performance" ram and just get decent branded middle of the road stuff... but you tell me that when I'm shopping!

[Citruspers]

If you build a VPN server to bridge into your network (instead of reserving a dedicated subnet), make sure you enable promiscuous mode on the virtual switch in VMWare ESXi, otherwise you can't ping outside of your VPN server.

[lagnificent]

Had a funny I think I've finally narrowed down now...

Today I'm using my new desktop as main desktop finally. During the build and setup, I did dabble with a little overclocking, but as a high end CPU there didn't seem to be any "free" headroom without turning up the voltage. Now and then I'd get a windows error requiring a reboot. There was an oddity in that all the usual stability tests never gave any error (prime95, linpack, memtest). I gave up and decided to run at stock, but I soon realised with prolonged load (many hours) there could still be lockups. This worried me as I've not overclocked anything now.

Cutting in to the suspect answer, I ran my high performance ram under its spec, and haven't had a crash since. It was XMP profile rated as 1600 CAS 8 1.5V. I've backed off to the fastest JEDEC setting of 1333 CAS 9 and all seems well now. Maybe I can tweak it up some more, but I doubt it'll be worth the time it'll take to prove. I did tell myself in the past to stop buying "performance" ram and just get decent branded middle of the road stuff... but you tell me that when I'm shopping!

I'm not an overclocking guru (never do it myself), but reviews of Intel processors on sites like Anandtech, Tomshardware, Hardocp have shown that when the i7 debuted, alot of exisiting highend memory didn't play nice at the time due to voltages/timings.
Nehalem
http://www.anandtech.com/show/2658/7

With your processor, pretty much the whole northbridge was integrated into the CPU, to the point where Anandtech said to pretty much leave the memory and timings alone, and just increase the multipliers (if it's a K model).
Sandy Bridge
http://www.anandtech.com/show/4083/the- ... 0-tested/3

[popo]

The old i7 warning was on the high voltage modules around at the time. Most are now running at the standard 1.5V, including the ones I got at their rated speed.

This article at Xbitlabs compares the impact of ram speed in various applications. It ranges from nothing to perhaps 10% in non-synthetic benchmarks between the two settings I'm using (ram rated vs CPU safe). I did overlook the CPU is only rated to 1333 so 1600 ram is in a way pushing it beyond its rating. Side note: I went for the performance 1600 ram as at the time there was no significant cost saving going to branded slower grades. Still, the system is stable and I'm not going to tweak it any more for now. I could probably push it a bit further, but I'm not going to spend days fiddling to get a few % if that average difference.

[popo]

Wow, has it been so long? The stability issue I mentioned before, only now have I made some headway into it. The curiosity was that the system would appear fine for a long time, and only have a wobbly perhaps as rare as once in two weeks. I did try reducing the ram speed, timings, stress testing, adjust cooling, replace PSU. Nothing helped. In desperation I got a hint to try an extra long run of memtest, and on the 10th pass there was a single error. And that was the only error in 20 hours of running. So whatever I'm chasing, at least I have a possible way to catch it now.

My first test is to go back to the ram rated speed but I've set a vague performance setting on my mobo from the default of "turbo" to a mere "standard" and see if that helps. The reported timings didn't change all else being equal, so it must be some hidden tweak somewhere. However I wont know for a while if that works or not...

[popo]

I'm not far from throwing my desktop PC in the skip (after removing HDs )... I found and corrected an occasional memory glitch by manually setting slacker timings than the mobo pics. Over 24 hours of memtest clear, and a few more hours repeating the specific subtest that caused it. But my screen glitch is still happening... I never had proof the two were related, but I'm way out of ideas now. This desktop has been a pain for a while and the more I use my laptop the more I think I should just use that as my main system. Convert the desktop to my winter heater and sell it on in spring when it warms up again.

[Bob Andersson]

Hi popo,

Sorry if I've missed it but what screen glitch?

I've had the odd blackout on my screen for ages and thought little of it as it only lasted a second or two. But occasionally it lasts long enough to generate an nVidia "screen driver not responding" error from Windows so I did a bit of searching. Firefox is always running when the PC is on and turning off hardware acceleration in its Options dialog seemed to help but not completely. I've now also turned off hardware acceleration in the Flash player (right-click a Flash video and select Settings). Not sure yet if it will completely cure the problem but neither Firefox or Flash seem any slower so I can wait and see without issues.

Bob>

[popo]

Bob, that's sounds like the one. I have suspected I only see them when I'm using the PC. Sometimes, but not always, either FF or Live Messenger will crash when that happens. Both use flash... I suppose I could throw furmark at the GPU and see what happens but right now it has new ram fitted and is undergoing yet more memtesting. In the past I have eliminated it as a specific gfx hardware issue by swapping with another model but it kept happening.

Side note, I probably should have started a new thread for my problems since I'm lacking on the solutions part for this thread...