Help: 4940mx overheating

Sorry if this is in the incorrect location. About a fortnight ago I swapped out the CPU in my dell precision to a 4940mx. I figured this would be fine since the computer shipped with this CPU from dell. However upon installation the cpu stayed underclocked at a max of 3.5 ghz on all cores even after having the clock and/or the voltage bumped up/lowered with XTU or Throttlestop. Only after updating the bios and then also using the "dell command power manager" to set the computer to "ultra performance" mode will the cpu go to stock or above. However, the The CPU will jump to 90C+ in a matter of 1-3 seconds when presented with a workload of any more than 20%. I've repasted the CPU twice but it seems like, (I have no objective metric to gauge this) The temp problem is getting worse. The fans will suddenly simulate a rocket launch over something as simple as watching a youtube video. Is there a problem with my CPU? I know haswell runs hot but this seems ridiculous. Could this be caused by the thermal paste I'm using? Thanks.

 

Alright, since I've fought with this CPU for quite a while and am currently the holder of the highest performance to my knowledge I guess I would be the go to guy.

First of all lets figure out which precision model do you have?
What thermal paste did you use??

I can already tell that you have a pretty bad heatsink fit. 20% usage in 1-3seconds 90+C is even for a terrible cooling solution not legit.

 

Hey, thanks for the reply. I have an m6800, and I was using Thermaltake TG-7 paste.

 

Can you remove the heatsink and make a picture of the Heatsink and CPU with thermal paste on?? This will give me a really good idea about your heatsink fit.

 

Sorry for the slow reply! Here are some pics. Tell me if you need anything additional and I'll post it. Also, I know that I have about a gallon of thermal paste on the CPU. The first time I reseated the heatsink I used the correct amount but the second time I was roid raging about my thermals and used too much.

 

Additionally, here are my thermals while idle and under a sort of mild load. I only ran the passmark floating point math test. The screen shot was taken roughly one second into the test.

 

I mean its pretty apparent that the thermal paste is pushed away from the die. The fit itself is rather good, Maybe you should try something thicker like PK3 or ICDiamond.

With that fit even liquid metal is quite valid.

 

Alright, thanks. So you just think that the issue is the thermal paste? I don't have PK3 or ICDiamond on hand but I'll purchase some online now and try it out when it arrives. I'm not sure if I'm ready to do liquid metal yet though just because I'm a baby, worried about damage it might cause.

 

Well as you can see yourself, half the thermal paste is basicially gone.

 

Even Artic MX4 that is easy to find would hold better :/

With your current thermal paste you will always end with over-heating, order some kryonaut and enjoy the result.

 

Sadly I have similair experiences with Kryonaut.

 

Never had pumping with kryonaut, but I had Liquid Ultra dry on me, twice already, on lapped heatsinks, after 1 year and a bit it starts to oxidize and it turns more or less into a compact powder that lets my cpu go into 80-90ºC at idle, on two different laptops.

So, with thermal pastes, YMMV, as usual.

 

Hey thanks to both of you for your help and suggestions. I obtained different thermal paste, however I'm not sure if it's working any better, because after applying the paste, when I turned on the laptop, the cpu started clocking itself down to 2.8 ghz or so under load. It's not overheating at all either when it's doing this, and I don't have any idea how to get it to stop. :/ Do you guys have any ideas?

 

that's odd considering that 3.1ghz is the base clock. You can check with hwinfo64 what the cause for lowering the CPUspeed is.

 

Nothing is really jumping out at me, but I could certainly be missing something. Everything is set to defaults in the bios. Windows power setting is high performance. The first pic is the same test I ran a week ago.

 

Sorry but you're not hitting it with any load on all cores, why would it boost?

You can see that it only uses 1 core. use something like XTU or prime94 to test your CPU.

 

Learn how to use Throttlestop. How to use guide here. Download ThrottleStop from first post.

Run TS Bench (1024M. You find it in main window. Enable Limits reasons and let it be open while you run the test. Remember clear out any varnings after you start it up and before the test. Take screen shoot in middle of the test. With Hwinfo, ThrottleStop main windows, TS bench and Limits reason up in tha background. Need see all info.

As well use High performance in Windows Power plan. Not balanced, own created or the power saver plan.
Lastly... Take screenshoot of all settings from tabs like FIVR, TPL, Options etc from Throttlestop main window.

 

The underclocking actually went away yesterday for no reason, but today it's back. The only thing I've done other than a few benchmarks is use the PC for schoolwork on the internet. As you can see there were no warnings in throttlestop but yet (take my word for it) the FID was 33 since the start of the test, when I should be running at 3.8GHz on all cores. Windows 7 and 10 actually say 3.3GHz is the default running clock speed of the CPU but that may or may not be irrelevant/erroneous considering that it is a 3.1GHz chip.
Pr95 clocks down to a good 2.9GHz and runs at 90C. Seemingly more useful than throttle stop is the XTU readout which says the chip is current throttling. However (take my word for it) even when the chip was running at 3.7-3.8GHz under load yesterday it said that it was current throttling, so once again I dont know what to make of that.

This is just the XTU benchmark running for the sake of thoroughness.
Additionally when it was running at 3.7-3.8GHz under load yesterday it would get up to 96C real quick, forcing me to stop any test, meaning that while thermals are marginally better, they havent really been fixed if I want to run my CPU at stock speed. The CPU did not thermal throttle in any of these tests shown above.

 

I apologize, for throttlestop I forgot to uncheck the disable turbo button....whoops... Today it ran at 3.8GHz before the core thermal throttled about 30 seconds in. The cpu stayed around 90ish C. no other warnings were shown. I really dont know what to do. today the cpu was running at the rate it should in all the tests, but it would just thermal throttle. I dont know why it's thermal throttling and i dont know why its downclocking itself even when its not. I really need help.

 

try disable "BD PROCHOT"

use hwinfo64 and set your fan to maximum, wait 2-3 minute

then run benchmark again to test clock and observe temps

 

Why would i turn off BD PROCHOT if my CPU is already throttling?
I tried turning my fans all the way up and it just took a little longer to throttle @ 3.8GHz

 

BD PROCHOT can cause throttling, that's why you turn it off.

 

Yes, but right now the cpu is throttling because of heat, so if i turn that off, my cpu will overheat and shutdown, or am i mistaken?

 

BD PROCHOT and PROCHOT (thermal throttle) are different things. That being said, it probably won't do anything for you.

 

Tangential, but I was under the impression that PROCHOT was just a trucation of BD PROCHOT. What makes them different?

 

PROCHOT is normal CPU overheating. BD PROCHOT is when something else, like an overheating GPU, causes the CPU to lock at its lowest multiplier (800MHz).