E6400 overheating throttling

Nvidia. I'm getting in the mid 70s GPU with no throttling. Before, I could only get to the high fifties before the computer became useless.

 

Assymetricblog,

On my E6500/NVIDIA system, when it's not docked, it can require GPU and CPU temps in the low 80's Celsius to invoke throttling. Personally, I wouldn't be comfortable that the problem is gone on your system unless you can sustain over 85C for both CPU and GPU (at the same time). If you really want to be sure, I'd get the GPU over 90C as others like Goodbytes and Goldeneye have. Check on downclocking with Performance Monitor (on XP) by monitoring "% of Maximum Frequency" in the "ProcessorPerformance" group (or whatever the equivalent would be for Vista or Windows 7 - somebody else would have to help out on that). There are other ways, but that's the way to do it without installing extra software.

I use FurMark to push the GPU (using the biggest possible window in "Xtreme Burning Mode" of "Stability Test") and Stress Prime 2004 to push the CPU cores (you need to be running two instances of that - use "Small FFTs - stress CPU". Again, run this all at the same time.

BTW, can you tell whether they just replaced the heatsink you had with a different unit of the same kind or is some sort of souped up heatsink? Is that really all they did - replace the heatsink?

Did they have you check for dust clogs before you sent it in?

Thanks

 

Truly amazing. Those of us having this problem couldn't even dream of reaching such temps - our systems would come to an agonizingly slow crawl before we ever got close. Are you sure you're not running any software that overrides the throttling such as RMClock or some other software to undervolt or overclock your CPU? You do seem to be running a very customized system.

OK, well, that's not idle. That's substantial processing. And that is indeed the problem I and other E6x00 users (both NVIDIA and Intel graphics users) are experiencing and talking about on this and other forums - throttling under load - not idle, but doing something at least, not necessarily very much, but something, especially something with graphics. Among the posters to forums complaining about this problem, your experience (where you're not having these problems) seems to be the exception, not the rule.

Well, yes, maybe the Northbridge (which contains the Intel integrated graphics for those systems that have it, as opposed to those with the NVIDIA discrete graphics option). But something or a combination of things heats up "a bit", exactly as you say, but not very much, and in reaction to that, the system is throttled, "acting like it's overheating" just as you say, even though it's not really overheating. That's the problem - throttling as if something is overheating (whether it's an NVIDIA GPU or the CPU itself or the Northbridge or maybe a combination of devices) when in reality the temperatures are not very high at all.

Well, you yourself may not have this problem, but I'm an NVIDIA user and I absolutely do (read my report if you're not convinced). It seems many other NVIDIA users have this problem as well (some with E6400's some with E6500's). For example:

From an E6400 user in this forum:

From this Forum:

On en.community.dell.com from woood77 on 4/10/2009:

On en.community.dell.com from jonataneinarsson on 6/18/2009:

On en.community.dell.com from community_bob on 7/2/2009:

On en.community.dell.com from aloprete on 7/13/2009:

Please, let us all acknowledge that some NVIDIA E6x00 users are also complaining of this problem.

*************************************************************

Many users not only report that they have the problem when their system is docked, they actually report the problem is WORSE when it's docked. I am one of those people (again, this is made very clear in my report ).

For example:

From this forum:

On en.community.dell.com from woood77 on 4/10/2009:

On en.community.dell.com from gegoodwin on 5/26/2009:

On en.community.dell.com from JoeB7 on 6/7/2009

On en.community.dell.com from ivka57 on 3/17/2009:

So to clarify, many users, including myself, find the problem is worse when the system is docked. Let's please not have any confusion about that.

 

I don't have RMClock, nor have tweak the internal hardware in any way, shape or form. Let me clarify something: this is using a FurMark which focuses everything on the GPU, and not on the rest of the system. In real world, like playing a game, the GPU can reach 95C (I'll retest and re-confirm this today or tomorrow, with more details on the other temperatures), and then it throttles because wither the northbridge or CPU has reached it's max temp.

Spinning a globe should not be challenging for the Intel 4500HD which is equivalent to the Geforce 7400M in performance (an average).

Of course, all of our conclusion are not based on scientific facts, so I can argue in any way. All we are doing is smart guesses. But, it based on this forum only, it would seam that people who got their motherboard and/or heatsink replaced, with the Nvidia solution had their problem solved, my guess is that there was a new version of the motherboard or heatsink that was made.
One idea that I have, is that I know that the heatsink touches the metal base at 2 areas. I wonder if that perhaps the base is miss constructed OR the heatsink is different in a way where it doesn't touches and transfer the heat to the base. This is just a hypothesis.

Well you have the 15inch model, and from what was said here, if still hold true (or true at all) the 15inch model of the latitude E, has DDR3 GPU memory instead of the DDR2. NOTE I AM NOT talking about main system memory, I am talking about GPU memory.. the 256MB of it.

In any case, I truly hope that everyone pushes and bug the hell of Dell for this. If one person complain, it will be put under the rug, if enough people complain, then a big fix will come. Complaining on the web is kinda useless, as all it does is give a hope that a Dell BIOS programmer sees this, faces with the same problem on his system and talk about it to the manager who might ignore it. Sending it to Dell and pushing it, will ensure that the proper manager gets these things and plans a fix with it's team of engineers to have it fixed.

[edit]
I don't believe in exceptions in computers, for such thing. A possible hypothesis into the problem is that Dell Canada did something that is possibly different from other regions. Sadly, we are not enough of each region to come to such possible conclusion. Who knows, maybe Dell Canada uses a different motherboard chipset then from Dell U.S. Heck U.S is so big that maybe it even separated in segments.
[/edit]

 

OK. You can now add me to the list of those who are affected by this bug.

Yesterday I was running some tests for this thread and I disabled the throttling in RMClock (I normally have throttling set to 100% as a legacy from my E6400 throttled itself if run from a 65W PSU).

Today my computer has been feeling sluggish (OK, I am running a virus scan in the background). Finally, I remembered I had disabled the throttling in RMClock so I had a look on RMClock's monitoring page and, sure enough, the CPU had been throttled to 939MHz. The two CPU cores were at 40C and 47C respectively.

So I re-enabled the throttling in RMClock (select only the 100% value which means no throttling) and immediately the CPUs got down to do some serious work and started to warm up. The ACPI temperature (as reported by HWmonitor) has risen from 53C (throttled) to 64C (dethrottled by RMClock).

I would have expected to have noticed these problems during the original benchmarking of my E6400. However, I didn't, and I wonder if this problem was added in during a more recent BIOS update. I stick by my earlier theory that someone put this rule in to try to minimise the thermal cycling of the nVidia GPUs and thereby reduce their failure rate, without realising that the BIOS would also apply the rule to the integrated graphics versions of the E6400.

John

 

It doesnt make sense to me. From what I have read is the throttling issue mostly affecting those with the Intel version where the temps is lower than those with the Nvidia card. I have temps over 80c on the ACPI without any throttling. I use HWmonitor for temp readings and use the resourcemonitor to see the CPU. I also use I8kFan which always report around 2548 Mhz. And I dont have any lasting "event 7" in the system log either.

This might be a problem to verify, I wonder if all those who claim they have the problem really have event 7 that have lasted a long time (or until restart) so it can be verified were the problem is. Some maybe have other problems, I have had an overheating GPU. And remember, the Nvidia GPU have "normal" working temps up over 90C, perhaps is the Intel version more sensitive to heat?

 

Yes, what I see from your description and the screencaps (thanks for those) is consistent with Software-controlled Clock Modulation (AKA On-Demand Clock Modulation, ODCM) kicking in due to this bug we're (mostly) all experiencing. I see that your performance state (P-state) was *not* degraded since your FID (CPU clock multiplier) was still at its maximum of 9, so it seems RMClock was apparently preventing degradation to a lower-performance P-state. However, since you had throttling (ODCM) override temporarily disabled in RMClock, the bug was biting you that way. It's clear that ODCM hadn't kicked in all the way, though - you would have had even worse performance in that case. If you had been monitoring in RMClock as the degradation was kicking in, you would have seen something like what I've attached (I simulated your situation on my system to create the screen cap). If you trained a fan onto your system instead of re-establishing ODCM control in RMClock, you would have (eventually) seen a staircase going up instead of down.

Is it possible, do you suppose, that you've had the bug all along? At least one person has gone all the way back to BIOS A03 without eliminating the bug. See:

http://en.community.dell.com/forums/p/19247293/19510422.aspx#19510422

Interesting theory - can you tell me what you're basing this on? Is there some indication that the Quadro NVS 160M is prone to such failure?

Thanks

 

That's a nice graph that illustrates what is happening. The question is why?

I got my E6400 on 28 August last year and BIOS A03 did not come out until mid-September, by which time I had RMClock controlling the throttling because of the PUS issue. I have BIOS A02 if anyone wants to try it.

what faith would you have in nVidia having fixed the packaging problems when you read articles such as this?

John

 

So as long as RMClock is overriding the throttling, users won't experience any slowdown associated with this bug? What is the passive thermal management users would sacrifice? Is it possible to replace this with RMClock thermal management? Basically, with RMClock, would users be able to avoid this issue altogether?

Thanks for all the help you and everyone has provided regarding this problem.

 

I reported this problem and the problem with the fan running at too high speed to Dell a couple of weeks ago. I kept asking them to escalate the problem and in the end I got the answer that Dell are working on at least the throttling problem and will come out with a BIOS update later that will fix it. Since I hardly ever push my E6400 to the limit where it will start throttling I will not do anything on my own until the BIOS update arrives.

I just hope that they will fix the high fan speed problem in the BIOS update also since that is a more annoying problem for me.

Örjan