E6400 overheating throttling

Tomorrow, I'll test this for 10min.
What I want to see is throttle percentage when I run this test.

 

I can also give you this but I'm busy this weekend.

 

The more data the better. Some of the new XPS 1645 laptops run just fine while others fall flat on their face and become unusable due to throttling. The updated 130 watt adapter has helped but power still seems to be a limiting factor for some users. For the XPS 1645, it seems that the quality of the power adapters might be responsible for their throttling problems.

Some software like HWiNFO32 and CPU-Z calculate an approximate equivalent MHz number. I found that this approximation is not 100% accurate so I don't bother calculating that for ThrottleStop. I prefer to keep the clock modulation and multiplier completely separate so it is easier to determine exactly where the problem lies. Some Dells use clock modulation first and don't touch the multiplier while others drop the multiplier down to 6.0 first and then will start to use clock modulation after that.

I've found that the most accurate way to determine how much CPU performance you are losing do to throttling is by running two wPrime benchmarks.

Start by running wPrime to get a baseline of how your CPU performs. There usually isn't any throttling when just the CPU is being stressed.

If you run a Prime + Furmark test and find out that the multiplier drops to 6.0 and clock modulation drops down to 75.0%, you could then stop that test and enter those values into ThrottleStop and use ThrottleStop to force your CPU to run at that same throttled speed. The wPrime benchmark scales very linearly with CPU performance so when you run a second wPrime benchmark at your throttled speed, you will have a very accurate comparison of how much performance you are losing due to throttling.

Measure Computer Performance | wPrime Multithreaded Benchmark

 

ThrottleStop is broken...
When I log my system with it,without doing anything, the log report shows that the multiplier vary from ~7.25 up to 8.92 under seconds (about 1/4 of a second) even thought the system is completely idle.

For the test which I'll perform in a few minutes I will exclude Throttle Stop (sorry mate :/)

Everest shows the current Multiplier speed appropriate and current CPU. I already got my laptop to throttle when I had my laptop base replaced (no new thermal paste was applied by the technician), and Everest showed the throttling problem where my CPU was down to 798MHz no mater what I do, unless I use the special BIOS mode, where I override (someone here posted a special code to do to enable this while the system is running).

Dell replaced my heatsink (thermal paste/pad was applied already), and of course the issue was resolved.

 

Ok guys, here are my results:

System Specs:
- Intel Core 2 Duo P8400 2.26GHz (no tweaks nor mods)
- 4GB of RAM DDR2 800MHz dual-channel
- Windows 7 64-bit
- Nvidia Quadro NVS 160M 256MB of memory

Initial system temperature:


FurMark settings: (I went all in with Xtreme Burning Mode)


Prime 95 + Furmark started at 9:40 PM - status:


I got bored... so I overclocked my GPU to show no throttling on my system:

As you can see the GPU is VERY well overclocked. I use Nvidia System Tool (aka: nTunes) to overclock my GPU while being in Windows.
Larger picture (full desktop): http://www.helpweaver.com/tt/GPU_overclock.PNG

11:03 PM - End Stress test - CPU Speed:



11:03 PM - End Stress test - System Temps (GPU was overclocked):


As you can see, with these extreme temperature, I have virtually 0 throttling. As you noticed Everest showed a slight decrease in performance of 5MHz (133MHz if you include the fact that the CPU is supposed to go to 6% overclock: 2394.0 MHz). This is very good, considering that this a laptop.

Immediately after the end of stress test, I used my laptop, and found that there was no speed decrease in any shape or form (the CPU clock got back to its 6% auto-overclock, GPU still overclocked). I played HD 1080p video just fine, and started to play Team Fortress 2 (CPU intensive game) just fine as well, confirming no real throttling issue.

I hope this very long test that I performed confirmed that the Latitude E6400 with the Nvidia solution provide no throttling problem with the BIOS A20.
Therefor on this conclusion, I will keep saying that the problem is due to Intel's bad design, and that the Latitude E6400/E6500 one of the first laptop on the market with this new Intel GPU, did not design it so that the heat-sink is bigger on the northbridge to prevent this issue for the Intel users (lack of space). Did Dell know about the problem? It's a possibility. Probably Dell spent million on the design of the laptop, thinking that the Intel GPU won't be an issue as in heat wise (as the Nvidia solution heats more, and that shows no problem) so now no fix can be done (they won't re-design the laptop to make a thicker model for the Intel solution). Did they knew? Probably, but they probably guessed that the most a business user will do with an Intel solution is play 1080p videos, and not push the GPU with advance 2D graphics and 3D, something that the Intel GPU was not designed for.

 

Once again, instead of jumping to conclusions, why not post a log file so I can explain to you how a Core 2 CPU works. The algorithm I'm using is not something I made up off the top of my head. I'm following the method that Intel recommends in their November 2008 Turbo White Paper and I've been using that method since then. The method uses high performance timers within the CPU that are extremely accurate and precise. Intel created these timers so software would be able to accurately monitor the performance of their CPUs. The ThrottleStop log file is quite capable of telling me exactly what's going on in a Core i or Core 2 based CPU.

Everest is a great program but the problem is that Everest does not report either type of clock modulation throttling that Dell uses.

Here's an example. For a test, I decided to run 2 threads of Prime95. Here is what Everest and ThrottleStop report.



Both programs correctly report that the CPU is running at full speed on both cores.

In the second example, the performance of this CPU is being severely reduced by the use of clock modulation throttling. ThrottleStop is able to correctly report that there is a problem. Everest completely ignores the throttling and continues to report that the CPU is still running at full speed.



The Everest screen shots you posted are nice but are completely useless. They don't prove anything because Everest does not monitor clock modulation which is the number one method that Dell has been using to throttle their laptops during the last 2 years.

You seem to have already made up your mind that throttling does not exist. Hopefully after the weekend, when HerrKaputt has a chance, he can do some testing and post a log file so I can have a look at some data.

Edit: Here's an example of how Everest handles Intel Dynamic Acceleration.

 

Thank you for the explanation.

I will do again the stress tress with the following setup:
- Prime95
- FurMark with "Xtreme Burning mode"
- Everest showing the temperature
- ThrottleStop showing the Monitor result.

So we are clear that CMod% is what shows the throttling. Meaning if it's at 10%, it means there is a 90% throttle.

 

Pre-test information...
Right now I am stressing the computer... it's been about 20min (I started it before the above post). I am not doing the official test right now, I am just assuring that the test will be done properly.

After 20min, with my GPU overclocked, I notice the throttling. It ONLY occurs when the GPU reaches 100C, when it comes back down to 99C, the throttling is gone.
CPU, and Chipset temperature are not really changing when the GPU bounce between 99C (no CPU throttling) to 101C (CPU throttling).
To be clear: If the GPU is at 100C there is no throttling, but once it reaches 101C, the throttles occurs, and will go back to normal, then the GPU is at 99C. CPU and chipset temperatures are not affected by the small temperature switch from the GPU.

If I turn on ThrottleStop (to block throttlestop) to stop this, then my GPU will go to 103C and stabilize there. I am aware that my GPU is overheating to some extend, but recall that I am pushing the laptop in a most unlikely scenario, where if I game, the CPU and GPU will be at max 100% load for long period of time.

These extrema temperature that my laptop is going through and the throttling that is visible (please note, once more, that my GPU is overclocked like no tomorrow), is completely normal for a laptop.

 

hi there,
what GoodBytes writes there about throttling is what I am experiencing as well: as soon as the GPU reaches ~100 °C, throttling begins and it ends when the temperature goes down.
And that only, if I am running the docking station with a full hd screen attached. running with the internal screen only, there is no throttling for many minutes.

But four(!!!) replacement mainboards where necessary to make this happen. Before, throttling happened much faster.

What I find abnormal is that the GPU temperature while being idle is still at ~70°c while cpu is only at 35°C.
--
hhhannes

 

Clock modulation is an indication of whether your CPU is running at full speed internally or not. When clock modulation reports 100.0%, that means the CPU is not being throttled by the clock modulation method.

There are two problems with Everest. It doesn't monitor the clock modulation status of the CPU but it can also interfere with other monitoring software. I don't recommend using it when testing for throttling.

If you want to keep track of your temperatures then all you need to do is go into the ThrottleStop.ini file and enter TJMax=105 and ThrottleStop will monitor and log your CPU temperatures and if you use GPU=1 ThrottleStop will monitor and log your Nvidia GPU temperatures.

When testing, why not reduce your GPU overclock back to the default settings? You can't do an accurate test if you are experiencing throttling due to heat because you've got the GPU overclocked.

Edit: Here's the Intel documentation that contains information about clock modulation.

Intel® 64 and IA-32 Architectures
Software Developer’s Manual
Volume 3A: System Programming Guide, Part 1
http://www.intel.com/Assets/PDF/manual/253668.pdf

Sections 14.5 and 14.5.3 Software Controlled Clock Modulation explains this and the numbering system Intel uses.