Laptop Geforce 650m Odd Behavior

[jamesv3]

Hello, so I have a clevo w110er laptop which has been experiencing horrible performance loss, which doesn't make sense to me. When I first got the laptop in 2012 I could play Battlefield 3 on medium settings, now my computer has trouble running 2D action games like The Swindle. (getting 10fps in that game.) I first noticed a sizable performance drop after a warranty repair job (repairs unrelated to any internal components) from the reseller I bought the laptop from. Then yesterday I noticed an odd glitch in the nvidia control panel and decided to do a clean driver install which fixed the problem (deleted registry entries, and all nvidia folders on computer after removing entries from programs in control panel). I re-installed the latest driver for my card, and made sure to have every program closed for the install. When I opened my first game I noticed another large performance drop comparable to the first one. I have run some diagnostics to figure out what is going on.

I also ran dxdiag and Furmark, while gpu-z was collecting log info. I don't know how to attach the texts documents though. Can anyone help me with this situation? Is there any way I can improve my card performance?

 

[Unolocogringo]

Looking at your charts it is overheating and throttling down to save itself.
When was the last time the fans and heatsink were cleared of lint and dust?

 

[jamesv3]

Looking at your charts it is overheating and throttling down to save itself.
When was the last time the fans and heatsink were cleared of lint and dust?

I check it monthly (there has never been significant dust collection in the fan/heatsink, not the problem), according to the manufacturer, expected operating temperature of my card is 60-90 Celsius. The maximum temp during the furmark stress test was 94 Celsius, and hovered around 92 Celsius.

 

[Unolocogringo]

You are definitely overheating. If you look at the GPU-Z chart ,the current temp is 75c, unless you changed the defaults. Which was a short time AFTER you ran Furmark. A short time before that is was quite a bit higher.
If you look at the core clock the GPU had boosted up to 3D performance clocks. Then the clock rate drops. temperatures continur to rise and clock rates drop again. They continue dropping to their lowest Clocks and voltages and stay there.

The only way to clean the heatsink connected to the end of the heatpipes is is to remove the back panel and a few plastic cover parts. It will be where the exhaust vent on the side is located.
Might be time to replace the thermal paste between the GPU and heat sink assembly.

 

[jamesv3]

You are definitely overheating. If you look at the GPU-Z chart ,the current temp is 75c, unless you changed the defaults. Which was a short time AFTER you ran Furmark. A short time before that is was quite a bit higher.
If you look at the core clock the GPU had boosted up to 3D performance clocks. Then the clock rate drops. temperatures continur to rise and clock rates drop again. They continue dropping to their lowest Clocks and voltages and stay there.

The only way to clean the heatsink connected to the end of the heatpipes is is to remove the back panel and a few plastic cover parts. It will be where the exhaust vent on the side is located.
Might be time to replace the thermal paste between the GPU and heat sink assembly.

Ok, thanks, I will do that. I have never taken the heatsink / fan assembly apart to check for dust so that might be the problem like you said. I will also look into replacing the fan as sometimes it will make weird noises.

 

[XxskaterjerkxX]

What the other guy said the thermal paste might be toast and may need some replacing. Your best bet for a thermal compound would be the choices provided in this link with a updated 2015 tests.

http://www.coolpc.pw/?p=115

 

[jamesv3]

You are definitely overheating. If you look at the GPU-Z chart ,the current temp is 75c, unless you changed the defaults. Which was a short time AFTER you ran Furmark. A short time before that is was quite a bit higher.
If you look at the core clock the GPU had boosted up to 3D performance clocks. Then the clock rate drops. temperatures continur to rise and clock rates drop again. They continue dropping to their lowest Clocks and voltages and stay there.

The only way to clean the heatsink connected to the end of the heatpipes is is to remove the back panel and a few plastic cover parts. It will be where the exhaust vent on the side is located.
Might be time to replace the thermal paste between the GPU and heat sink assembly.

Ok so I took apart my computer and opened up the heat sink and found minimal dust collection, it didn't look like enough was present to hamper cooling. Anyways I put everything back together (after cleaning the dust out) and turned on my laptop, immediately fired up gpu-z and looked at the core clocks, and also ran my fan at max rpm. At start, core clock was 835 mhz and 900 memory clock. It looks like every time I start gpu-z my clocks are reset to default level (835 and 900). When this happens, no matter if my fan is at full blast or off, the temperatures rocket to 85 c and then the clock is rapidly reduced to 135 mhz core and 405 memory. This temperature rise happens with absolutely no load on GPU, it seems like just raising the clock on the card (to default levels) is enough to push it over 90 C.

Here in this picture I set my fan to max RPM for a few seconds before I start GPU-Z to make sure temperature is low, and start GPU-Z. Immediately temperatures rocket to 85 C and core clocks drop, temperature stabalizes at 63 C. I completely turn off my fan, temperature rises by a 4 degrees in 30 seconds. Then I turn my fan back on, at full blast, and the temperature barely moves, decreasing 2 degrees after 12 seconds.

 

[jamesv3]

As an update I have ordered liquid metal pads for my GPU (also the CPU) and a replacement fan, I will update the thread in a while with the results. Thanks! Hopefully my card hasn't already been fried.

 

[jamesv3]

Update: I have replaced the old fan with a brand new fan, and have cleaned off and replaced the old thermal paste with gc gelid extreme thermal paste. Looking at gpu-z, the temperature of my 650m chip now stays under 50 C even when running a game. However I am still experiencing the same throttling issue, and I can see the core clocks and voltage shooting down as soon as I open gpu-z. I am not sure why this is still happening. Does anyone know what could be causing this?

 

[Unolocogringo]

It could be a software bug between your laptop and GPU-Z.
Download and try afterburner to monitor temps, voltages, fan speeds and usage.

http://www.majorgeeks.com/files/details/msi_afterburner.html

 

[jamesv3]

Problem solved! It appears that it was indeed the thermal paste not conducting heat properly as now my chip is performing at a normal level and at expected temperatures. The problem that I perceived with the under-clocking in gpu-z disappeared when running a game and all clocks jumped back up to normal values. I guess nvidia optimus was doing that.