Monitor causes strangely high CPU usage

[rf13]

see attached

I did not unplug the monitor but the computer is currently running normally.

 

[H2SO4]

Yeah, a definite difference in the DPCs, and it looks like the background interrupt level is lower too.

During the "bad" run, did you feel the level of responsiveness was particularly bad in that block between 50 and 130 seconds? (You can see the DPCs go nuts during that time.)

 

[rf13]

responsiveness was basically terrible throughout but i was really checking - what i can say is that normally the CPU jams at 100% during the bad runs, but this time it did fluctuate from 50% - 100% and jamming at 100% at times. It could have been that the CPU was jammed at 100% in the 50-130 sec block. Bit vague, sorry.

 

[H2SO4]

Delayed Procedure Call. This gets really esoteric. When a bit of hardware (keyboard, mouse, disk controller...) wants the OS to pay attention, perhaps because the user pressed a key or moved the mouse, it does an "interrupt" which results in the OS momentarily parking aside whatever it was doing in order give control to the driver which is responsible for that particular hardware device. The driver is supposed to quickly lodge a request for interrupt-handling work to be performed later, and give back control to the OS so that normal operation can resume. Hence a "delayed" procedure call. This is all at the very lowest levels of software operation.

Sometimes when a hardware device is broken it will "interrupt" for no real reason, and then the OS and the relevant driver have to waste a lot of time needlessly handling those interrupts, the driver perhaps lodging maaaany DPCs to deal with the situation.

Because the broken hardware device interrupts without regard to what is happening at the time, this "handling" behaviour has to happen in what's called an arbitrary thread context - it might be Explorer, or Outlook, or chrome.... or anything else running on the processor at the time, but the processor still gets hijacked to handle the interrupt and the resultant DPCs. To most processor utilisation methods, it looks like explorer/outlook/chrome is responsible, but the problem is really down in kernel mode underneath all that.

Right now, your box looks to me like it might be suffering from that sort of issue. I'm not sure though.

Go on, please do an XPerf run with the monitor disconnected. This is interesting

 

[rf13]

hmm, interesting.

I had another bad episode, so took the opportunity to take this xperf snapshot with monitor still plugged in. CPU was jammed at 100% the whole way through and the computer was particularly unresponsive. I encouraged this state by opening adobe elements which requires the computer to generate a lot of thumbnails of Jpegs and often sets it off.



now have monitor unplugged and will take another snapshot - btw, once the monitor is unplugged it takes about 5 mins for the computer to return to running normally

 

[rf13]

snapshots with external monitor switched off are attached

 

[H2SO4]

Almost beyond doubt, your problem is linked to the number of interrupts being signalled via both processor cores. That second "bad" run clearly shows the machine is spending a sustained 2% inside the interrupt-handler routines, which is very high. In turn, they cause DPC times to sit at a very high 10-20%, and that leads to abnormally high utilisation stats being logged against whichever hapless process just happens to have its tiny slice of processor time when a particular interrupt is signalled.

The contrast with the "good" XPerf run is obvious. Your interrupt levels are basically blips way under 1%, which is where they should be. Same with the DPCs - not even 1%.

Since it takes some time after you plug the monitor in for this to start (I understand there's a gradual buildup), and it takes ~5min for things to settle after the monitor is unplugged, there might be a thermal component to the problem. Whichever hardware device is signalling all these interrupts does so as a function of its temperature, and somehow the presence of the monitor connection is warming it up nicely. That's a total, utter, complete, and unwarranted guess though.

Is the machine under warranty? How bad do you want to troubleshoot this to the nth-degree? There's also a slight possibility that the interrupts are of the "software" variety, but that seems unlikely given the link to the monitor and the gradual onset and decay.

Once you know for a fact these are hardware interrupts, it's back-to-the-hardware-vendor time. Nothing that you can do short of replacing or deactivating the hardware component is going to help.

Probably not the news you wanted

 

[rf13]

ok - thanks very much for your help H2 - it has gone a long way to restoring my faith in humanity.

I have succumbed and spoken to Dell. They are sending a technician out tomorrow to replace the motherboard, so that should get it sorted... will let you know.

 

[bray:way]

can i sugest one thing?

try uninstalling the drivers for the monitor/graphics card and downloading new ones.

this is just a complete and utter guess lol

-Braydon

 

[rf13]

thanks braydon - I have tried a total reinstall, which did not help, and then subsequently updated all drivers to the latest versions, this did not help either.

 

[bray:way]

damn lol
and i just read the post above my last one. it was poeted while i was reading the thread.
good to hear that they are doing that, i thought that they would make u send it to them lol

-Braydon

 

[H2SO4]

No problem mate. I hope the Dell techs manage to fix your machine. If you get a chance to ask them, it would be interesting to hear whether they know of similar issues with that model. They won't necessarily be forthcoming with that info, but they might hint at it if engaged in light conversation

 

[rf13]

the dell tech said that it was a relatively common problem with the graphics card. Apparently, NVIDIA's graphics card has a timing setting, or something, that is not compatible for working with windows. It results in overheating and displays going crazy. He is having to change a whole load of motherboards but said that it was the only problem with the machines. He also said that there had been a court case between NVIDIA and Dell and that he had been told there was information about it on the internet. All I have found so far is this:

Fudzilla - Nvidia, HP and Dell sued over defective GPUs

The tech was confident that the new motherboard would fix my problem, so that should be it for now. Will let you know if anything else goes wrong.

thanks again

 

[H2SO4]

That's great. It's always encouraging when the tech can complete your sentences while you're describing the problem.

It probably is the story you linked to. How many simultaneous nVidia/Dell legal battles can there be?

 

[rf13]

you may be interested to know that despite the assurances of the techie, "the bad" kicked off again last night. It took longer to happen but while watching a movie, heavy downloading as well as backup and restoring, the movie froze and I saw the CPU had jammed at 100%. I removed the monitor, and it solved the problem. Another call to Dell methinks.

 

[H2SO4]

Is there any reason why your laptop would be particularly heat-affected? An exhaust grille blocked by dust, or perhaps no room for air to circulate around it when in its usual location?

Another thing that would be quite interesting to know - does the monitor have to be ON to cause this, or does merely having it connected and OFF suffice?

 

[rf13]

No air circulation probs that I am aware of. I keep it slightly raised off a table to make it easier for the computer to ventilate itself.

Switching the monitor OFF but leaving it connected when the CPU has hit 100% returns CPU to normal. As far as I can remember, the CPU has never jammed at 100% with the monitor connected and OFF, but, I have very rarely used it like that.

 

[TomW]

The 100% CPU usage problem only seems to kick in after I have been using the computer for more than around 30 mins at least. .

Is your GPU using Windows system RAM for it's 128mb of video ram? Is it shared, I mean? To me it sounds like your GPU is over-heating causing your processor(s) to break a sweat after about half an hour. I suppose it wouldn't matter if it's shared-the CPU(s) would start to pull load if the GPU is over-heating anyway... any way to check your temps? I'd recommend a utility, but I'm afraid I don't know of any well enough.

Perhaps try a new video card to see if it still over-amps your cores.

Just a thought...

 

[H2SO4]

If it happens even with the monitor connected but OFF, the very circuitry around the connector becomes suspect.

I'm confident that the abnormal processor util is directly linked to the interrupts, but I don't know what's signalling those interrupts and it would be complex to help you pinpoint that through a web forum. Presumably the Dell tech would know though, if they encounter this problem a lot.

 

[TomW]

Presumably the Dell tech would know though, if they encounter this problem a lot.

I'm not sure the Dell 'techs' encounter this very often over there in India..