Re: Dell Latitude E6400 - Overheating issues with minute tasks such as playing flash videos for extended periods?

Already changed pipe (just LCD and case has not been changed) in my case caused decrease in temperature by more than 5 degrees. Dust is not the primary problem, but can really help the bad things to speed up.

With E6400 I had problem with sound popping and cracking, it helped to change SATA mode from IRRT to AHCI and downgrade to 8.2 version of Intel matrix drivers.

After UV and cleaning the FAN again (after two weeks I had again 1/4 of fan full of dust) my idle temp is 29 / 38 (fan running at 2800 rpm/ fan off (you won't believe me, but this model can also run without the fan being ON and be very quiet)) and load tmp is 38 cpu and 53 chipset. Keyboard is cold and no slowdowns, even in E-Dock, two monitors and closed lid (It has been running this way during yesterday for more than 10 hours at my office).

I upgraded back to A14, I will see.

mk

btw: We will never buy dell to the company anymore. It's cheap, but ... I already spent too much time reinstalling, calling dell support, exchanging parts etc. I need to work again.

mr dimsum,

until dell finds a cure, you have to cool the system down. Buy a cooling pad, UV the cpu and ask dell to change heater pipe and FAN (it's one piece).

mk

ivka57

mr dimsum,

until dell finds a cure, you have to cool the system down. Buy a cooling pad, UV the cpu and ask dell to change heater pipe and FAN (it's one piece).

mk

:) Great - finally an answer to the thread opener ;-) - Well that's all said than. We did our "duties". When will DELL do? I've sent another e-mail to the contact that handles product escalations. Cross fingers.

JB

I also plan unmount the fan and look under it how much thermal paste is installed.

Today, with A14 I experienced slowdown again (with UV and clean FAN and closed lid in E-Dock). According to Right Mark Clock Utility, CPU clock was 2.2ghz, but throttle was just 1.4ghz. I stopped the BOINC, opened lid and went for smoke.

After I returned, system was back and fast again.

mk

Hehe, that's conspiracy from the tobacco companies =))

ivka57

Today, with A14 I experienced slowdown again (with UV and clean FAN and closed lid in E-Dock). According to Right Mark Clock Utility, CPU clock was 2.2ghz, but throttle was just 1.4ghz. I stopped the BOINC, opened lid and went for smoke.

After I returned, system was back and fast again.

mk

Yes, that's the good thing with RM Clock. You can see the actual core speed and throttle. Did anyone else of you escalated the problem through DELL?

Or maybe we should build our own BIOS. This link describes the similar approach for modifying Award BIOS, dunno if it is applicable with Dell, but it's worth a shot. =) We could modify the termal tables once found. I am mostly scared of some CRC checksums, so the BIOS won't work after modification and installation and you get a nice black brick instead of a overheating computer you have now.

I got today slowdown with the same conditions as yesterday, but today with A03. So the only thing, what is making the difference is the temperature itself (probably on chipset).

mk

I created a detailed report on this problem based on many many hours of troubleshooting. When I posted it to this forum, it was nuked. The sorry excuse that was mailed to me was:

"Your post contains language that we believe reveals non-public information about Dell or another company or person. The Terms of Service do not allow users to post this type of information, which generally includes the names, telephone numbers and/or personal email addresses of Dell employees and Dell Community participants, as well as tracking numbers, customer numbers, order numbers, or service tags."

Of course, I didn't include any such information.

The report clearly exposes Dell's culpability in this problem we're struggling with on the E6X00's.

I'll create a new posting immediately after this one. That posting will contain a link on the web to my report. If you don't see the post, it's been nuked by Dell (and maybe even this one will be nuked as well).

OK, here is the link to the report I've created. I've given up on posting it directly to this forum. The web interface is buggy when it comes to including images and my first attempt to post this info was censored by Dell anyway.

Go to:

http://imsahp.chambana.net/~randall/dell

There are two PDF files there - one is smaller and more compatible, but the screen capture images are less readable/more blurry. The larger, higher quality one has better screen cap images, but is, of course, larger and may not work with older versions of Acrobat reader.

DELL

Your post was deleted by TOS20.

Dear DELL TOS20. I don't understand why you delete postings rather than support your customers. This is unfair as no one hear is impolite in any way. Ppl might sound frustrated after investing hours and hours, days and days into the product you've sold us because it's not running within it's specifications as it should.

In the meantime, I've opened this new thread: http://forum.notebookreview.com/showthread.php?t=348221

I'm going to pass that info now again to our strategic procurement department and the key account manager at DELL. 

Speak soon in the other forums.

JB

---This post has been edited by Dell-Bill B to remove specific violations of Dell's Terms of Service in order to restore it in chronological order. This was done with the foreknowledge and approval of the author, tinkerdude.---

I tried a second time to post the text of my report on this problem, only to once again have it censored by Dell. This time I was told

"Your post contains language that we believe to be harassing, unlawful, or otherwise objectionable. Use of such language violates the Terms of Service, and must be deleted in order to allow your post. Prohibited words and phrases include those in which certain letters have been replaced with alternate characters. Unlawful language may include any statements that might reasonably be interpreted as a violation of the intellectual property rights of any person, including a trademark, copyright or patent."

In my report, I did use words "in which certain letters have been replaced with alternate characters", but is that just a pretext to censor information embarrassing to Dell?

I'm now posting the text once again (third time), this time without using words "in which certain letters have been replaced with alternate characters." I'm curious as to how far Dell will take this censorship.

Again, to see the screencaps and photos I refer to in the text, you'll need to see a PDF version of this report, which is available at:

http://imsahp.chambana.net/~randall/dell

----------

HALLELUJAH! Finally, I've come across others who have suffered like I have with this ridiculous problem.

Only you can begin to understand my pain 8-P

I spent over $2K on an E6500 last fall and have had confounding, frustrating, ultimately infuriating problems with fits of slowness, worse slowness, *dog* slowness and absolutely paralyzing slowness that makes the system categorically useless. In the past couple weeks, I spent hours upon hours troubleshooting this only to discover it was hardware related (specifically temperature related). Of course, I had been blaming Micro$oft, everyone's favorite dog to kick, but it's now perfectly clear to me that Dell's got a major hardware problem on their hands here with their supposedly top-of-the-line business class dockable desktop-replacement (the E6400 and E6500).

I did tons of testing to confirm that the trigger was internal temperature.

Then I talked to three separate Dell tech support folks. They refuse to even acknowledge my test results (because I was using "third party software") and they refuse to acknowledge I have a hardware problem (since diagnostic tests don't fail). I've given up on them for now.

From reading the posts so far it appears that at least SOME corner of Dell is finally coming to terms with the fact that they have a defect of some sort with at least the E6x00 models, but I haven't personally come into contact with that particular corner of Dell yet.

It seems to me that all of us who are working on this should join forces to demand a resolution to this issue from Dell. This could wind up being a huge embarrassment for Dell. The sooner they nail this down, fess up to their culpability and FIX THE PROBLEM AT THE FULL CONVENIENCE OF THOSE ADVERSELY AFFECTED, the better it will be for them in terms of damage control. If they continue to drag it out or try to minimize it or deny it, they will only make things worse.

Let me make one thing absolutely clear - there is NO DOUBT that this is strictly a hardware and/or BIOS problem. That is, it's Dell's fault (perhaps they can spread the blame to one of their component suppliers, but they're the ones making the end sale to consumers). I have "smoking gun" evidence below. Read on.

Let's do the troubleshooting and investigation necessary to expose the truth. Yes, this is all happening on Dell's forum and they could just censor us all, but I hope they understand that would only make things worse for them (yes, I'm making copies of everything too - I encourage others to do the same).

OK, now on to the details:

Here's my setup:

Dell E6500 P8400 (2.26GHz Intel Core 2 Duo, 1066MHz FSB, 3MB L2 Cache) running XP

nVidia Quadro NVS 160M

4GB RAM

Normally docked to E/Port Plus, with lid closed

Dock connections in use include:

  Dual Dell E248WFP monitors (1920x1200)

  100BaseT (to DSL-connected home network)

  Targus ACP50 port replicator (which provides an additional 1280x1024 monitor) via USB

  Audio Advantage Micro sound pod via USB

  Andrea Electronics sound input pod (for external microphone) via USB

VERIFIED: No dust problems inside.

Here are my symptoms:

In short, using my system *even moderately* for even a few minutes can easily cause things to slow down a little at first, then things get worse with increasing sluggishness, any audio will start to crackle at first and then break up quite a bit, as things progress the system becomes less and less responsive, sound will even totally stop, all windows become unresponsive.  Once it gets bad enough, it's not even possible to shut down any processes (let alone shut the system down properly) because the system is virtually seized up (though still alive). In the past, my only recourse was to force power off (holding down the physical power button long enough). A power-cycle like this would offer some relief, but oftentimes the problem would soon reoccur, making the system nearly useless, making me feel like punching a hole in the wall or taking a sledgehammer to the thing (then I remember I spent over $2K on it, so I resist the temptation).

Incidentally, shortly after the symptoms start, CPU utilization as reported by Task Manager generally shoots to 100% and generally stays totally pegged there (even though the processes on my system, again as reported by Task Manager, don't seem to be hogging resources). I discovered recently that this is because the CPU gets downclocked and I wind up with roughly an 800MHz system trying to do the work of a 2.26GHz system - more details below.

I did some troubleshooting along the way over the months I was experiencing these symptoms, but couldn't nail anything down. It was truly confounding. I knew enough not to even bother calling Dell Tech Support back then - I had nothing to go on. I couldn't even reproduce the symptoms reliably - I knew that calling would have been a waste of time.

Then we recently had a heat wave here in the Midwest US with several 90F+ days in a row. I realized that my problems were definitely worse during hot days and that's when I first guessed (correctly) that there was some sort of temperature problem with my system (I normally don't use air conditioning so the ambient temperature would get to around 30C/85F, which is still well within specs).

I followed up on this in detail, finally having a lead on what's been plaguing my system lo these many months.

I downloaded and studied up on DPC Latency Checker and SpeedFan and got familiar with XP's perfmon. I also hooked up a temperature sensor (K-type thermocouple) to my Fluke 179 multimeter to measure the exhaust port temperature (left side of the E6500). Using these, along with Task Manager, of course, I monitored the system while I stressed it various ways (doing video transcoding with Windows Movie Maker, for instance, but eventually settling on a more convenient solution - Stress Prime).

It became crystal clear early on that it's all about temperature. I could create the symptoms by stressing the CPU and/or memory to generate heat, I could monitor the heat with SpeedFan and the exhaust port probe, watching the temp rise and level off. The temps don't need to get very high at all (really anything around 60C or above for the CPU core temps which translates to about 50C or above at the exhaust port will eventually trigger the problems on my system - your mileage will vary). Some minutes later, the trouble starts. The first thing that happens is the processor frequency is cut to 99% (why just 1%, I don't know, and I don't know how this is done, either - but that's what Perfmon reports). That frequency cut coincides with a one-time spike in DPC latency. Then, exactly 30 seconds later, the processor frequency gets cut to 70%, coinciding with another one-time DPC latency spike.  Then the internal temperatures drop like a rock (what you'd expect from the downclocking). But by this time, the CPU core temp doesn't matter much - there's no turning back - because even though the CPU core temps (both of them, that is) fall dramatically,  exactly another 30 seconds later, another huge downclocking kicks in - down to 35% of full speed. From there things go to South pretty quickly. CPU utilization spikes dramatically and soon stays pegged at 100% (Again, what you would have to expect if you're stressing your CPU and it gets downclocked). DPC latency starts rising (like 15 times higher or more) since the system can't keep up as well in servicing DPCs (Deferred Procedure Calls, a form of lower-priority interrupt, I gather). In fact, depending on how high the temp initially rises, *other* system-crippling changes are also made at 30 second intervals that can drag down the system until it's in an absolute coma! And even after the temperatures have fallen to what they were when the system was idle, the downclocking and other crippling downgrades STAY IN PLACE until, and you're going to love this - until I bust out a desk fan, nudge it up close and blast it - after a few minutes, the downclocking and other debilitations are disabled, step by step (one step every 30 seconds again, in fact) until the system is back to normal just as zippy as it's supposed to be.

Have I blown your mind yet?

The smoking gun:

I can even reproduce extreme slowness symptoms using Dell's own Pre-boot assessment diagnostics (which run when you power up while holding down the Fn key). When the system is cool, the test takes 12 minutes 15 seconds to run on my 2.26GHz P8400 E6500. When the system is warmed up by running that same test several times in a row (this is while docked, and docking *absolutely* does aggravate the problem), the test runs get longer and longer. The longest test run I recorded (after running it 7 consecutive times in a warmish room, low 80's Fahrenheit) was 1 hour, 5 minutes and 8 seconds. So totally independent of Windows, totally independent of anything I've installed on the system, using Dell's own built-in diagnostic testing, I can make the thing slow down to less than 1/5 of its usual speed just by running their own test a few times in a row to warm it up (it never *really* gets very warm, actually). I can reproduce this slowdown even when *not docked* using the same procedure, again in a warmish low 80's room, though it's not nearly as bad (in the little testing I did this way, the test ran 23:38 after 4 consecutive runs, which is still about twice as long as it should take).

So this appears to be incontrovertibly Dell's fault. Would someone like to confirm by running this same test? Just power up (preferably while docked), with the lid open. Hold down the Fn key while you power up. Test will start automatically. There is a point early on where you'll have to hit a key to move on (you'll have an audio cue when this is required).  Then just let the thing run. I started my stopwatch when the Dell logo power-on screen clears. Most of it consists of memory tests. Watch it run the whole way through the first time (again, for my 2.26GHz system, it takes just about 12 minutes 15 seconds on the first (cool) run). Then when it finishes, record the time, hit the physical power button (should turn off right away), and once again hold down the Fn key, then power up. Start the timer again once the Dell power-up logo disappears. If you do it in a super cool room, I'm not sure if you'll reproduce the problems if it's not docked. Would be interesting to find out, though.

Now, detailed evidence from my testing under XP:

Below is a screen cap of my system before I start a test run. Again, I have 3 monitors, so the screen cap (using Print Screen key) is pretty big. On the top monitor, I'm running Task Manager full screen. On the twin widescreen monitors on the bottom, I'm running:

Top level, left to right: Perfmon, SpeedFan, Stress Prime and Windows' clock.

Bottom level, I'm running DPC Latency Checker in a super-wide window that spans both twin monitors.

As you can see, the system has been sitting for a while and things are pretty steady-state. DPC latency is well under 500 microseconds (DPC Latency Checker), CPU utilization is well under 10% (Task Manager), Processors are clocking at full speed (pink traces in Perfmon). GPU temperature (nVidia) is reported as 62C, CPU core 0 is hovering around 38C and core 1 around 44C (SpeedFan). I have a couple other traces going in Perfmon having to do with DPC, just for kicks. Stress Prime is not engaged yet ("IDLE"). There are two instances because you need two to fully stress test a Dual-core system (though I'm only going to run one instance the first time around).

So here we go. I'm about to start a single instance of Stress Prime with settings that will bring the CPU to around 50% utilization. I'm going to go very easy on the stress testing this first time around. Ambient temperature is 27.4C (as per Fluke/thermocouple). Exhaust port temperature is 45.8C (measured at the bottom center of the port, roughly the hottest spot, again as per Fluke/thermocouple). Here's a photo of the exhaust port probe:

The following capture was taken 1:02 after starting Stress Prime (Check the "Duration" statistic in the Stress Prime window to confirm and in all these screencaps, check the Windows clock to see a timestamp for comparison). Note that CPU utilization is about 50% and the Core temps are taking off. Exhaust port temp is 48.7C.

The next capture was taken about 2:30 after starting the test. The core temps are starting to level off. Everything is still hunky-dory. Exhaust port temp is 50.7C.

OK, here's where the fun begins. Remember, I'm only stressing the CPU itself (not anything else) and it's only at 50% utilization, but 10 and a quarter minutes into this test, the CPU gets downclocked (by what, I don't know). The screencap below shows CPU core temps have leveled off at around 62C and 72C. Kinda warmish, but far from alarmingly hot. GPU temp is not appreciably affected and is essentially immaterial (we're not stressing that after all). Perfmon shows the dip to 99% (Perfmon updates its display by moving that vertical red line to the right). (Why such a small downclock to begin with? I have no idea.) Note the spike shown by DPC Latency Checker - that telltale spike is a marker showing exactly when the downclocking occurred. You'll see quite a few more of those telltale spikes in this report. Exhaust port temp is 53.1C.

The next screencap shows that another, much bigger downclocking, this time to 70% of capacity, occurred 30 seconds after the first (the telltale DPC latency spikes marking the downclocks are 30 seconds apart, since the samples are taken each second). (By the way, the little jumps you see in CPU utilization correspond to me doing the screen captures).

OK, the next screencap is where the hammer really comes down - processor frequency is cut to 35% - which would be somewhere around 800MHz for my system. Again, this happens right on schedule, 30 seconds after the last downclock. Exhaust port temp is 52.6C.

The next screencap shows what's transpired a few minutes after the downclocking. The system is definitely running slower, but it's not serious - it's still quite usable - just annoyingly slow. DPC latency (which can have a big effect on perceived slowness and preclude smooth audio or video playback) is not really too bad. But the thing is, check out how the CPU temps have gone back to what they were at idle. But the system is still downclocked to 35% of capacity. What's up with that? It's bad enough that my $2K system has been turned into a dump 800MHz box for no good reason, but now that the temps are back down, it's still poky slow because the CPU remains downclocked. Exhaust port temp is 48.3C.

The next screencap just shows more of the same. Temps way down, but the system still downclocked. This is 20 minutes after it was crippled down to 35% capacity. Exhaust port temp is 46.4C. 

At this point, I start blasting the system with a small desk fan to force the issue. Here are photos of the fan setup I used. Note the white eraser I stuck underneath to help with air flow.

Photo from other side:

So this next screencap, taken about 3 ½ minutes after I turned on the desk fan, shows the first upclocking, back to 70% capacity. Note the telltale DPC latency spike. Exhaust port temp is 42.3C.

And exactly 30 seconds later, the screencap below shows the next upclocking step, back to 99% capacity. Exhaust port temp is 42.7C.

And, once again, exactly 30 seconds later, we're finally back to 100% CPU clocking. Exhaust port temp is 43.5C.

So there you have it, something in Dell's design or some defective component or a bad BIOS table or something is progressively downclocking my system when it's got no business doing so. And then *leaving it downclocked* until I blast it with a desk fan. So the big questions to me are:

What the heck is so desperately wrong with CPU temps around 70C that the CPU speed has to be cut by 70%?

What on Earth is up with keeping the CPU speed throttled even after temps return to the 40's C and the system is hardly stressed at all, even at 35% of its normal operating frequency?

Why in the name of Beelzebub do I have to bust out a desk fan to get back my 2.26GHz system back from the abyss of 800MHz?

Now, if you think that's bad, let me just say you ain't seen nuthin' yet, because what you just saw is innocuous compared to what happens if I'm doing real work with my system. Again, allow me to demonstrate:

Once again, starting out steady state. As before, DPC latency is well under 500 microseconds (DPC Latency Checker), CPU utilization is well under 10% (Task Manager), Processors are clocking at full speed (pink traces in Perfmon). GPU temperature (nVidia) is reported as 62C, CPU core 0 is hovering around 38C and core 1 around 44C (SpeedFan). As before, I have a couple other traces going in Perfmon having to do with DPC, just for fun. This time, I'll use both Stress Prime instances and really try to stress the CPU and RAM too in order to generate more heat. This actually isn't even a worst case scenario for heat, but it will totally hoark my system nonetheless.

OK , ready to start. Ambient temperature, again, is 27.4C. Exhaust port temperature 45.5C. Screencap below is just before I kick in the Stress Prime runs.

Screen cap below is 1 minute 17 seconds into the test. You can see the CPU temps will settle a few degrees higher this time around. CPU utilization is pegged at 100% but the system is still zippy-responsive (really, this system is great when it's not crippled by Dell's screwup, whatever that turns out to be). Exhaust port temp is 50.6C.

And here we are about 3 minutes into the test with the first downclocking (it's not quite yet registered by Perfmon - I snapped it a little too soon, but the DPC latency spike marks the time of the downclock to 99%). Exhaust port temp is 53.6C.

And, right on time, 30 seconds later, comes the downclock to 70%. Exhaust port temp is 54C.

And now the downclock to 35% right on schedule. Exhaust port temp is 53.3C.

But this time, it's not over. No, this time my system is crippled even MORE. For the next 2 and a half minutes, every 30 seconds exactly, some *OTHER* kind of downclocking or other crippling debilitation is apparently inflicted on my system (as evidenced by the telltale DPC latency spikes and the step-by-step worsening of the DPC statistics. It's apparently not enough for the CPU to be downclocked to 35% of capacity. No, other parts of my system must be sabotaged. Downclocking various buses perhaps? FSB? RAM? It's a mystery to me, but by the time I took this capture, my system was absolutely unusably dog slow. All this despite the CPU temps dropping like a rock, just like last time. Go figure. Exhaust port temp is 48.1C.

And yes, it gets worse. By the time I made the screencap below, my system was basically in a coma. As an example, one thing I have to do to make these screen caps is save them to a file from Paint. Just to bring up the "Save as" dialog box in Paint takes about 56 seconds (that is, once I select the "Save as" menu item, I have to sit there watching Windows construct the "Save as" dialog box, bit by bit - I can see each little window component get incrementally drawn in - the outer border, the title bar, icons, text boxes, etc. It's really rather ridiculous. And again, this isn't even the worst I've seen it. Exhaust port temp is 46C.

The screencap below is about 12 minutes later. CPU temps have even dipped below the levels I started at, but the system is still crippled to be unusably slow. Note that perfmon couldn't even record stats for a short interval because of the debilitated state of the system. At this point, I cranked up my desk fan in order to recover. Exhaust port temp is 45.1C.

The next screencap is about 9 minutes later. You can see how the fan brought temps down a few degrees C, not really that much, but enough that whatever is crippling the system in steps started reversing its actions. You can see the telltale DPC spikes every 30 seconds which indicate upclocking or some other change somewhere in the system. The spikes I'm talking about are near 1, 31, 61, 91, 121, 151, 181 and 211 (guessing on that last one, it's not obvious). The capture is right at the point where the first CPU upclock kicked in - back up to 70% capacity. At this point, there are 2 more CPU upclocks left (to 99%, then full capacity) and then the system will be back to normal. Exhaust port temp is 41.5C.

The capture below is 30 seconds later, showing the CPU upclock to 99%. Exhaust port temp is 42.6C.

The next capture shows the final CPU upclock to full speed. Exhaust port temp is 43.8C.

Below is a screencap about 4 minutes after the final upclock that restored my system to normal operation (you can still see the last DPC latency spike that marked the final upclock). At this point I still have the desk fan trained on the system. But the stress test is still going. What do you think is going to happen? If the fan was off, of course, the system would go into downclocking-induced seizure just like before. But since the desk fan brings it out of that coma, you'd expect that perhaps if I leave the fan going, that may prevent the seizure in the first place. Exhaust port temp is 48C.

And below is a screencap from about 15 minutes later. Things are running pretty warm (though far from too hot) with the CPUs at about 67C and 75C. Even though it's been going 15 minutes like this, everything's still at full speed and the system is just as zippy-responsive as it's supposed to be. This is what one should expect of this system, it seems to me. Exhaust port temp is 48.6C.

I've already mentioned the big questions - to review:

Why cripple the system when the CPU core temps are only averaging around 70C?

Why keep the system crippled even when temperatures return to baseline (and even fall below that)?

Why do I have to dedicate a desk fan to my system to restore it from being comatose or prevent it from getting totally hoarked in the first place?

More questions directed at the possible cause:

Is there some additional (perhaps defective) temperature sensor that's causing the system to cripple itself?

Where is the logic used to decide on crippling the system? Is that what's defective?

I encourage you all to try these tests yourself. They are a bit time-consuming, I know, but we need to ramp up the pressure on Dell. Or we could just resign ourselves to pairing up each E6x00 system with its own portable external fan for the rest of its life and hope for the best 8-P

I'm about to call Dell again to move the issue forward. Ratchet up the pressure, that's what it's all about. No need to get hostile - do it politely. Hostility is counterproductive - be polite but persistent.

Some observations from previous posts.

People running Vista, XP and Ubuntu report this problem.

People with both integrated Intel and upgraded nVidia graphics report this problem.

And looking back, I see at least one other person duplicated slowness with Dell's built-in pre-boot diagnostics.

Another observation:

Now that summer is here, more and more people will experience this problem. Dell better get crackin'.

With all the respect to your research tinkerdude, i already posted this on previous page of this thread =)

The Temp4/5 sensors activate the downthrottling at 55C. You can override the internal downthrottling (at least partially) with RM Clock.

(Which does not mean Dell is liable for all the time we wasted on this + the censorship is driving me angry.)

Absolutely the same problem here on a Dell E6400. I've been beating my head against a wall for months trying to deal with this - I get it after I do a big compile task. Since it was most visibly a graphics slowdown, I've been upgrading the drivers (I have it with NVIDIA graphics) as soon as new ones come out, but nothing solves it.

After reading this thread, I have the same problem. To solve it for now, I have a 6' desktop fan blowing into the DVD drive slot, with the DVD drive and battery removed. It certainly has solved the issue for now, but completely useless as a laptop!

It's clearly a poor design on the cooling side. When I compile, I notice the graphics core temp rises 10F, even though its not doing any graphics work.