M18x R2 nVidia card turning off screen.

Hi all, I figured that rather than continue to hijack old threads, I should start my own so here goes. Thankyou to @Mr. Fox for the previous informed response on the other threads, I will have some questions you might be able to assist with later.

Before I get into my issue, lets start with my system...

It was ordered in the UK in June 2012, and took I delivery in August 2012. Shortly after, I moved to Canada and now reside in rural Alberta.

This laptop has been sitting on a Cooler Master Notepal U3 Plus its whole life.

The operating system was upgraded to Windows 10 Pro from Windows 7 Ultimate and is currently running Build 20H2. After upgrading I tried to keep all drivers as up to date as possible, including updating the BIOS to A14 which is the most recent option on Dell's website.

I added a Razer Blackwidow keyboard and Naga Mouse in 2013 and I increase the RAM to 32GB in 2017.

I have never repasted any of the processors since taking delivery of this machine.
So onto the problem...

Over the last year, when running on nVidia graphics. the laptop screen would randomly turn off (machine still running) did not matter if I was playing GTA 5 or browsing the web.

Finally I have had some time to sit down and look at it. Based on what @Mr. Fox has already said elsewhere, and some of the things I have tried, I think that either the primary GPU is overheating OR it is failing.

Here is a video of the problem...


A few days ago, I downloaded Unigine Heaven and ran the Benchmark. This is the first Benchmark I took (15/Dec/2020)
https://photos.app.goo.gl/5Wx1Nm2UQBDzaCAB7

Yesterday (16/Dec/2020) I benchmarked again because I couldn't find what I had done with previous day's benchmark.
https://photos.app.goo.gl/JBkxP5pSWZs9hKFG9

It is worth noting that both Benchmarks were taken with the fans running at maximum speed (4500rpm) through HWInfo64

I didn't see the GPU temps when I ran the first benchmark, but when I ran the second, they started at around 30C (Idle), jumping to 45C at the start of the benchmark and climbed to around 60C by the end.

For ****s and giggles, I decided to turn off the HWInfo64 fan control and benchmark without it, well I may just have fried my GPU properly. Temps started again at around 30C (Idle) but jumped to 50C at the start of the benchmark. Towards the end, there was a 20-30C difference between the two GPUs (SLI was enabled) and on Scene 24/26 just as the hotter GPU hit 130C (the cooler one was at 99C) the black screen happened and I was unable to save the benchmark.

Since then my machine has been incredibly unstable. I was able to get into Windows for just long enough to switch to the Intel GPU so I could create this post.

My Father In Law just grabbed me a tube of Arctic MX4 from Calgary (saved me a 2 hour round trip as all I could get locally was Startech Thermal Compund) I was considering kryonaut or similar but I couldn't find any available in store, and Amazon.ca are now delivering to my location in January .

So now I am about to shut down and strip my baby down to the bare bones, while I am at it, I am considering baking the one GPU to see if that increases stability. either way, before I re assemble, I will be switching the GPUs around.

Now for the questions...

@Mr. Fox and maybe others can also comment...
How much of a difference in temperatures should I expect with the repaste?

If the GPU is toast can you suggest affordable replacements? I could buy a new entry level laptop for some of the cards I have seen online.

Will SLI work if there are two different nVidia cards (675m and 765m for example)

I found an old thread about expresscard to eGPU adapters, are these still a thing and how would they compare performance wise to the installed cards - would a 1070 via expresscard be better than twin 675m SLI or is the performance drop from the expresscard prohibitive?

If I end up going the expresscard route later on, I saw something about needing an external monitor, could you not just pipe the eGPU to the HDMI in to still use the laptop screen?

Finally a question for @Mr. Fox specifically. I found your dual power adapter mod thread last night. Do you think that something like that could also help in my situation? I do have the 330W power brick that came with my machine, but looking at my processors, the CPU and GPUS combined could pull 250W peak by themselves (not sure what they are actually pulling since I don't have the test meter to check. Also will repasting reduce power consumption?


Wish me luck.

 

GPU toast, amazing your 675M lasted this long. Many died far earlier. You can put up to GTX 980M SLI and retain all functions or up to GTX 1070 with minimal modification but you will have to rely on SG/optimus going forward.

 

i know its not quite the same, but i dont dare use hwinfo for fan control on my m18x r1 anymore. the few times i did try it, it would randomly shut off and not return control to the systems fan control. luckily due to having the modded bios installed it throttled the gpu once it hit 95c. now i just rely on the default fans and the gpu tops out at 75c. once u change out the gpu and repaste id suggest seeing what the temps are like with ur default fan profiles, i know the r2 also has a different custom bios which might also help things.

 

M18x fan is not controlled by BIOS, it is by EC, we cannot control/modify fan tables.

 

well by control i meant the hwinfo sliders. I forgot it was the EC that controlled it, my mistake, could have sworn the fan behaviour changed though after flashing the modded bios, unless it updates the EC also?

 

Confirmed, primary GPU was toast spent this afternoon stripping it down, swapped with secondary gpu, repasted and rebuilt.

System is stable with SLI off, but crashes frequently when SLI is enabled.,