HAY GUYS

neus · Dec 3, 2006, 10:00 PM

About 30-100 minutes from startup, the CPU usage jumps to 100% and stays there until other applications start crashing or I lose patience and restart the computer. Even after I turn off all non-Windows-critical processes, the CPU still stays at 100%.

I've updated and ran Kaspersky anti-virus, Spybot Search and Destroy, and Lavasoft Adaware. As expected, all came back clean. I'm fairly certain it is not a virus/trojan/adware problem.

Now, next thing on my head was an overheated CPU. It is winter season and my room has been kind of warm. So I checked the CPU right after restart in the BIOS - 42 degrees Celsius. That seems fine to me.

So now I'm at a loss. I'm willing to try everything besides reformatting this HD because that'd be a pain in the arse

OS: Windows XP SP2
CPU: AMD Athlon 3000+
RAM: 1024 MB

Jam it back in, in the dark.

neus · Dec 3, 2006, 10:17 PM

You see, right now it is behaving - there are no problems. But the next time it happens I'll definitely screenshot it. I do have a HiJackThis log.

HiJackThis:

There's nowhere I can't reach.

neus · Dec 3, 2006, 10:45 PM

I've turned off the indexing service for both of my hard drives.

For the record, during normal operation my memory/CPU graphs look like this:

Normal:

When the computer starts misbehaving, I'll printscreen that and update the thread.

This thing is sticky, and I don't like it. I don't appreciate it.

neus · Dec 4, 2006, 08:43 AM

Originally Posted by RYU

are you sure fan on your pc working?

As far as I can hear, yes. Is there a software method to find out if it is indeed functioning as it ought to be? I would prefer not to take off the case and look inside, but if it comes to that, I'm ready, willing, and able.

Originally Posted by Roph

maybe dumprep? I totally cannot find ways to express my hate for dumprep.

No, sir. No dumprep anywhere in sight - this malady is not a result of an application crashing.

---

Just a while ago, I caught the beast. I was randomly browsing in Firefox, and I noticed things slowing down and I immediately went into task manager. Well, looky here, what do I find?

The running precesses:

The CPU goes nuts:

The processes again:

CPU still going nuts:

---

As I mentioned above, I was afraid I had a rootkit installed. Thus, I ran an app called Rootkit Revealer and this is the scan log. It looks clean to me.

Rootkit Revealer log:

---

My current hunch is that my RAM might be corrupted. Can you recommend me a program that does this?

I am a dolphin, do you want me on your body?

neus · Dec 6, 2006, 10:40 PM

Domino - I've ran memtest for about three or four hours and it came up with nothing.

RYU - It takes me a bit longer to start up and shutdown nowadays.

I've been experiencing this far too much lately and I'm thinking of reformatting my HD. There's damn nothing I haven't tried

Just for kicks though, it happened again a couple of minutes ago and I went into Process Explorer.
None of the normal tasks are using up the CPU time. This is what's happening:

DPCs going nuts:

Apparently Deferred Procedure Calls are taking up 100% of the CPU. This is the cause of my problem and I have no clue how to approach the beast.
If anyone has a suggestion, post it here in the next couple of days but otherwise, I'm formatting.

I was speaking idiomatically.

neus · 03:11 PM

Haha, holy shit Blah I want to make sweet, sweet <3 to you.

I have the exact same hardware set and now I'm thinking that the problem is with my 6800GS overheating. Hm, gonna have to look further into this stuff. I might not have to reformat after all.

How ya doing, buddy?

neus · 11:03 PM

Originally Posted by RYU

try this:

1.open case pc.

2.try to remove all cards installed like"sound,network..." except VGA.

3-turn on you pc without put case.to if fans working will or not(power supply,processor and other fans if installed)

4-try to touch VGA card if hot or not.

4. after WinXP started open Device Manager.

2. Double-click IDE ATA/ATAPI Controllers to display the list of controllers and channels.

3-set none (if available) for both.

4-apply reg file I uploaded.

5-restart you pc and tell me if still slow or not.

I appreciate your intent but I don't think I'll be doing any of those. They seem dangerous and rather unwise. I can check my GPU temperature via nTune (an nVidia app that has all kinds of info on my graphics card). The temperature is well within its operating range.
Also, I'm not about to randomly modify my registry because you've said so - please justify your actions.
Don't take it personally though - I've figured out what my problem is. More on that below~

Originally Posted by SCHWARZE-5

If what RYU posted didn't work. Check and see if its svchost.exe under Local Service is spiking your computer. If that's what it is, just clean unwanted startups with HiJackThis. Clears it up for me whenever I had the problem.

Thanks for that. Yes, I've checked what process is causing the CPU spikes and it is a sub-process/system of the System Idle Process called Deferred Procedure Calls. I'm running six svchost processes and none of them abnormal handles, memory or CPU usage. Furthermore, I've scanned my computer for viruses and other malware (as noted above) with HiJackThis, AVG, Nod32, Kaspersky, Spybot S&D, and Lavasoft AdAware. They all came clean. This is a hardware problem.

Quote:

That's a surefire way of screwing your computer up entirely. Seriously, no advice is better than this.

Again, I appreciate your intent but it would have helped me (and most everyone else reading this) if you justified why it was bad advice. But thanks anyway.

---

Now, here's what I've discovered so far. Following the clues from Blah's thread, I came across this excellent discussion in the Sysinternals forums which pointed me to the RATTV3 program which basically logs DPCs. Using it, I narrowed down my DPCs to a single video driver - VIDEOPRT.SYS.
I further researched this driver against my hardware set and figured out that a large number of people complained of this very same problem (random CPU "storms" - periods of insanely large number of DPCs all related to that nVidia video driver). The common thread between all of these people was the graphics card: nVidia GeForce 6800GS.
As soon as the card was removed or disabled, the problem went away. This was the consensus across numerous official nVidia driver updates. No matter the driver set (and believe me, I have tried a lot of them myself, including the newest ones released 7 days ago), none would fix this problem. I believe it to be a poorly designed graphics card.

Breakdown of the problem:
1. A system intensive task raises the GPU temperature inside the 6800GS
2. The video driver, VIDEOPRT.SYS, starts throwing DPCs at the CPU
3. The longer the stress on the graphics card is continued the more DPCs are sent and the more CPU time is eaten.
3a. You can find out what process is using up your CPU via the app Process Explorer. You can find out what part of your system is causing Deferred Procedure Calls (DPCs) via the app RATTV3.
4. When the video card is disabled (there is a way to do this via the control panel in Windows) or removed, the problem goes away
5. Solution: so far, I've come up with: get a different graphics card. The 6800GS is getting dated right now and it'd be a good idea to update if you have the cash. If not ... I'm sorry

I read through the entire discussion on Sysinternals above and I found no software solution to the problem. Many people recommended getting a new graphics card as they were reasonably cheap now and I believe this is the course of action I'll take. There was also talk of the 6800GS being incompatible with the nVidia K8/K9 motherboards but I didn't look much into that.

I've understood it to be a fault with my graphics card. I've even tested the idea empirically - if I leave the system idle, the DPC process behaves and stays reasonably low. As soon as I engage a video or a game - tasks that are very GPU intensive - the DPCs start rising and given time gradually rise to 100% of the CPU.

I've heard a lot of people switched (for a reasonably low price) to an nVidia GeForce 7900GT and encountered no problems. I suppose it is a time to upgrade my graphics card.

I believe the reason I didn't discover this earlier is that I hadn't placed the card in a heated, winter environment before. This is my first winter playing with it and since it is on the top floor of the house (hot air rises up) and since I keep my room rather warm, the card has started breaking down.

Thanks for all your replies.

If someone else has this problem in the future, I would advise you to ditch the 6800GS as soon as you can. If you are absolutely broke, give these third-party drivers a try: link. Though they are far less recent than the latest ones published by nVidia, I have found them to be much more stable in terms of this problem. I have noticed them being buggy with some games though so take your pick - restarting the computer every random amount of time with the nVidia drivers or keeping it turned on but playing selective games.

Phew, that was quite schpiel. I want to be thorough because if someone looks at this in the future, I'm sure they will be quite relieved to get the information they want in a single place. I know I would have preferred reading this one post to 35 pages of that Sysinternals discussion

FELIPE NO

neus · 06:36 AM

I am thoroughly sorry to bump a thread after a month of inactivity like this but I've discovered a way to fix the entire problem and I feel that ought to be documented.

If you have the problem I've described - a CPU spikes to 100% because of Deferred Procedure Calls (DPCs) thrown by the driver of the nVidia 6800GS - the easiest way to fix is it to simply leave the PC to do something GPU intensive for about ~5-10 minutes.

The easiest way I've found of eliminating the CPU spikes is to simply open a movie as soon as they occur and leave it playing for 5-10 minutes.

Edit: Visual proof. Pardon the crappy mspaint job.

The DPCs, and consequently the CPU usage, gradually fall down to 0-1% and everything goes back to normal.
The reason I've not discovered this before is because it goes against one's instinct. When I play a game and everything randomly slows down, the first idea is to quit the game and see what the problem is.
In fact, the correct course of action is to simply keep playing and the problem will go away.

[speculation]
I believe the problem lies in the GPU temperature. Once it reaches a specific temperature, it starts throwing DPCs at an abnormally high rate, but once the temperature falls or rises, the problem goes away. I believe this to be the case because playing a movie would only make the problem worse - ie. the GPU hotter. Similarly, I've yet to encounter this problem when the air around the PC is cold.
Remember, the 6800GS has a range of 0*C to 80-some *C, if I recall correctly, and these storms usually happen around 40-50*C. [/speculation]

For what it's worth, I'll try to sage this as in 4chan so that the thread isn't bumped to the first page - the info would only be relevant to people searching on the internet.

What, you don't want my bikini-clad body?