Alienware 15R3 - Re-pasting and Overclocking Experience

Hi Everyone,

I thought that I'd share my experience with my Alienware 15R3, since I am done with all the overclocking today

The laptop was ordered on 30 September 2016 and arrived on 25 October 2016 with an i7 6820hk, GTX 1070, 4K screen, 512GB SSD, 1tb HDD and the Killer 1535 card.

As soon as I received the laptop, I noticed a few major issues: heavy overheating of the CPU (it was hitting 90°C already from 3.6-3.7 GHz with stock voltage), overheating of the GPU (also throttling and hitting 80-90°C overclocked to +200 core and +300 memory), and uneven core temperatures (over 25°C difference between cores!!). It seems that I was quite lucky with the 4K screen though because I only had some minor light bleeding (mainly at the bottom left corner), and it is actually only visible at 100% brightness.

Therefore, I decided to contact @Mobius 1 who re-pasted the CPU and the GPU for me with liquid metal (1g Grizzly Conductonaut was enough for both CPU and GPU). He was very friendly and we are still discussing everyday, so I would highly recommend him.

After the re-paste, I noticed a few things: temperatures were around 30°C lower for the CPU, the fans were silent, overclocked GPU temps were kept at around 60-65°C, and most importantly all cores were approximately at the same temperature (more details below).

Then I decided to overclock the CPU with Throttlestop 8.4 (for multipliers and voltage) and intel xtu (to remove the power limit). Each overclocking setup was validated with 10 hours of OCCT larget data set and 10 hours of OCCT linpack. FYI, getting it stable with linpack usually required +5 to +15mv compared than standard OCCT with large data sets.

Here are the results of the CPU OC:

- Stock Clocks (2.7 Ghz w/o turbo boost) stable at -170mv core/cache (avg cores temp difference: 2°C)
- 3.8 GHz stable at -115mv core/cache (cores temp difference: 3°C)
- 4.0 GHz stable at - 30mv core/cache (cores temp difference: 4°C)
- 4.1 GHz stable at +15mv core/cache (cores temp difference: 5°C)
- 4.2 GHz stable at ~1.30v (cores temp difference: 5°C and I'm not using offset anymore above 4.1 GHz)
- 4.3 GHz stable at ~1.35v (only tested for 1 hour though with all fans at the maximum because I didn't want to risk it with ~1.35v - cores temp difference: 6°C)

To be completely safe in heavy gaming, you can always increase your OCCT stable values by around +5-10mv but it should not be needed.

I believe that I was very lucky with the quality of my CPU. Cores temperature differential went down from over 25°C to around 2-5°C depending on the voltage. I don't know if it's possible to get 4.4 GHz stable because it would require over 1.40v which is something that I am not willing to test. The best compromise seems to be 4.0 GHz or 4.1 GHz for a daily gaming setup.

After I was done with the cpu, I moved on to the GPU OC (1070 GTX) with MSI afterburner 4.3. The GPU was perfectly stable with +230 Core and +300 RAM (i.e. 8600 MHz RAM), which is higher than what I could do before re-pasting. However, there seems to be a bug with the Micron vram which requires a firmware fix (see here: http://www.guru3d.com/news-story/ma...for-geforce-gtx-1070-due-to-memory-issue.html), preventing it from overclocking higher. FYI +240-250 core also seems to be kinda stable, but I got a crash after 4 hours of Heaven Benchmark. GPU temperature stays around 65°C at +230 MHz core 8600 MHz ram.

Firestrike tests are around 19k for the graphics score and 16k overall score on the highest overclocking settings.

Overall, I am quite happy with with the whole process The laptop is finally almost up to the expected quality of an Alienware.

However, there is still a major bug in the bios which has to be fixed: fans take too much time to adjust to the temperatures, and it's not unusual to reach 85-88°C above 4.1 GHz for a few seconds before fans are activated (and then the temperature goes back to below 80°C). This bug is valid for all bios versions (1.0.2, 1.0.6, 1.0.8 and 1.0.9) and needs an emergency fix from Alienware.

Finally, Alienware should also release a GPU firmware fix for the Micron memory ASAP, similar to the one released by all the major 1070 GTX cards manufacturer - see my other post here: http://forum.notebookreview.com/thr...bal-issues-gpu-firmware-update-needed.800219/

Thank you for reading my review, please let me know if you have any questions

 

Thanks for the information! I seem to only be able to get around 190-200 overclock on the GPU core before hiccups, but I am able to get +600 on the memory clock. Have you tried higher with the memory clock?

 

It appears that our GPU has completely opposite behavior. Mine will crash if I go above 8650-8700 MHz (which is +325-350 in afterburner since it is DDR memory), but the GPU core is stable without any problems at +230 or +240 overclock after 12 hours of heaven and gaming. In fact, I am even considering adjusting the memory to 8600 (+300 in afterburner) because it seems to be the optimal for the benchmark scores. Are you able to get your 1070 gtx stable at 9200Mhz memory (i.e. +600Mhz in afterburner)?

Please note that before the liquid metal re-paste, I also had trouble going above 200-210 MHz for the core.

 

I think the problem now with AW is that they're too conservative with fanspeed, other than that the heatsink is excellent.

 

Agreed, the laptop (CPU and keyboard) should never get so hot because fans are too slow to react and to spin. Let's hope that it will get fixed with a bios update asap.

 

I have been gaming and running plenty of benchmarks around +550 on the memory clock and it seems to work perfect. Whether or not that's a sweet spot that offers optimal fps is unknown but I dont get less performance than without overclocking so I guess it's good. sitting at 2090mhz core Clock and 9000mhz memory. This is stock no repasting no nothing. I hit the lottery with thermals.

 

Subbing for future finding. Might be returning mine for an upgraded model.

Sent from my VS995 using Tapatalk

 

I slightly updated the values (by a couple of mv) to reflect further testing.

 

This is weird, +600MHz verus +300MHz is a pretty big difference for the memory so in fact I am wondering whether Alienware has put some GPUs with different memory brands. On the other hand +230MHz vs +200MHz core is still kinda within error margin of chips and also due to liquid metal re-paste.

Edit: it looks like we got Micron and Samsung memory available on the GPUs and unfortunately I got Micron Micron will usually not go above +300 while Samsung memory can go +600.