Since my later university days, I've had the fortune of sampling some fairly highpowered laptops when using them for work. However, performance laptops (also known as gaming laptops) are not without their qualms, especially compared to the quite robust desktop gaming pc's. If my blog/website had had any sort of significant following I would probably have used it as a sounding board for my (in my mind, well founded) greivances. Because even though consumers are generally well cared for in Europe, sometimes things do slip through the cracks.
Case in point, my earliest highpowered laptop from DELL, the fateful m17x. This machine caused me a lot of headache back in the day, and I'll spare you the details and present the abridged version. The machine contained both an integrated video card as well as a discrete one. Switching between the two required a reboot. Before you toss up your hands and laugh at this decision - consider that this choice might - theoretically - not have been that silly. Running in 'energy conservation mode' most performance GPU's still suck a lot of power, so while including a secondary integrated card may have made the whole machine heavier and more expensive, theoretically it would yield the best of both worlds. A truly energy efficient integrated graphics card for longer lasting battery sessions, and discrete graphics cards to really yield performance. The reality was that the machine required custom DELL-made nvidia drivers to function properly. Over the course of the machines support life, guess how many drivers DELL issued?
When considering that using SLi (which the machine supported) in any game requires specific driver support, you can quickly see how lacking this triple driver release schedule was. If my memory serves, the three drivers were released over the course of 12 months. Meaning after that, any game released around a year later would not make use of SLi with DELLs mandatory drivers. I ended up having to install the nvidia reference drivers on an older machine just to extract the SLi profiles and then inject them into the DELL drivers, thus making it SLi capable with newer games. When Nvidia then changed the SLi profile format, I was forced to also manually modify the profiles prior to injecting them. The M17x also had a host of performance issues and would lock-up somewhat sporadically (while using DELLs homegrown drivers of course). Three hardware revisions were made, and while a few customers were lucky enough to recieve free upgrades, DELL in Denmark seemed less inclined to help.
The moral of the story is, if you buy a high-powered laptop, make sure that the producer stays close enough to the Nvidia or AMD reference cards to use the native drivers, rather than rely on the hardware producers themselves for driver updates.
Fast forward a few years and I sit here with an aging Origin PC EON17-SLX whose performance has been waning. It sneaks up on you. Games just don't seem to run quite as well anymore - and I really wish I had made some performance measurements the very first day I got the machine to get a comparative baseline. Alas - I did not, so I can only show the current before/after state of things. So let's dive into it.
Before you replace your thermal paste
You really ought to spend a little time to figuring out whether or not thermal paste is the source of your performance problems. All my research hasn't really led me to a conclusive answer so far. Some people seem to think "every two years" is the right answer, and others say it should last "forever". Do yourself a favor and look into how your GPU's are performing before doing anything at all. I use HWiNFO myself, but any free sensor monitoring tool will do. Set it up to monitor both the temparatures and clock frequencies of your GPUs. Here are some of HWiNFO's readings during a recent play session of Shadow of Mordor.
The main GPU is in red (on top), and the secondary GPU is in gray (on the bottom). I've annotated a few things in the image above to make things more clear. The green lines indicate the GPU's maximum clock frequency (on boost), the brown lines show where the temperature is 90 degrees celcius (and clocks down), and the two purple lines approximately indicate where I entered/left the pause menu in the game. Notice how the main GPU struggles at a near constant 90 Degrees clocking down in frequency in order to not meltdown. Only in the pause menu does the machine get to catch its breath. As I leave the pause menu (around the 2nd purple line from the left) you can see the main GPU clocks back up to its maximum frequency of 757.7 Mhz for just a little bit, before the GPU immediately hits 90 degrees and once again clocks down. This is a clear cut case of GPU overheating. If you see something similar when taxing your machine, then replacing your thermal paste might help.
Note: Just because the clock frequency is at maximum, it does not mean that your GPU(s) are performing at maximum capacity. Aside from running a graphics-heavy game, another useful tool is FurMark, which renders some very processor intensive graphics.
Replacing the thermal paste
You can find a ton of step-by-step guide on how to do this online, so I'll just stick to some of the specifics in regards to the Origin Eon17-SLX unit, along with a few things I learned along the way.
Above you can see the backside of the laptop along with some of the tools I used to do the job. Arctic Thermal Paste along with the branded cleaning solution I'm fairly sure you can find a much cheaper replacement for.
Here's the opened laptop, with battery removed. In addition to unplugging your laptop when doing any maintenance, removing the battery is also a good idea. A few guides will suggest wearing various anti-static wrist guards or other safeguards. While I'm sure charged static can lead to damage on electric equipment you may be servicing, its something I personally don't take any precaution regarding. But I also do not seem to gain any static charge while in my working environment, so if you do seem to attract a charge, make sure to discharge it prior to working on your laptop, i.e. touch something made of metal.
Having removed most of the pre-existing thermal paste using the cleaning products and a significant number of coffee filters, here's what the CPU looks like with a bit of thermal paste applied. A few things to note here. When cleaning either the CPU or GPU, coffee filters are often recommended as a replacement for a lint-free cloth. To my understanding you can use most any cloth for the first few passes, as long as you reserve the final cleaning passes for either a lint-free cloth or a coffee filter. Many sites will suggest using a particular method to apply the thermal paste. I'd suggest having a look at one of the many cool YouTube videos convering how thermal paste tends to spread to get a good idea of which method to use. Note, some methods (like the dot method) are probably better for beginners and tend to do the job well.
Here's a good shot of what your GPU should look like once you've completed the cleaning process. Nice and smooth. The mobile camera had a lot of trouble focussing on the chip itself as it tended to reflect all the light that shone on its surface.
Here's a small amount of thermal paste applied to the GPU. I'd say it might be a bit more than optimal, but after having completed this process about 3 times, I'd say it's good enough. To the best of my knowledge, the worst thing that can happen if you apply too little, or too much is that the unit fails to cool properly, resulting in an emergancy shut-down. While this isn't exactly ideal, it does not result in any permanent damage to the GPU/CPU as far as I know.
Having the replaced the thermal paste approximately three times to ensure I got it right, I'd love to say that performance was much improved and things are better than ever. Except they're not. I've definately noticed and improvement, but not as much as I had hoped. The machine still overheats, and I'm beginning to think it's just hardware fatigue setting in. The only thing left would seem to be the fans, and while they aren't blaring away, they certainly aren't silent either. The vast difference that I am monitoring between the master and slave GPU would seem to indicate that the master GPU simply fails too cool properly anymore. Perhaps due to internal reasons?
I've considered switching the GPUs to see if the same situation presents itself and if it really just is an issue with one of the GPUs, or perhaps the attached heatsink, but given that the hardware is nearly 3 years old by now and likely to be replaced soon, it hardly seems worth the effort. That being said, I may very well opt for a non-SLi laptop in the future. The extra weight and heat does come with a cost it would seem.