Nitro Spin 5 cooling mod/repaste results

[deferentlemur]

I was in the market for a new laptop this July, and wanted something portable, with good battery life, 360-degree screen rotation, and discrete GPU. The Nitro Spin 5 was the only option I found with decent graphics capability, i7 U-series CPU, fully rotating screen hinges, and a reasonable price. I bought one, being fully aware of the thermal limitations and game stuttering thanks to reading and watching a multitude of user and YouTube reviews. The thermal problems were noticeable but did not seem as severe to me as made out in the reviews. I attribute this to driver and bios updates released between the reviews and my time of purchase, I got a well-build unit, and/or I simply was less bothered by it than the reviewers. 

 

In an effort to alleviate the thermal problems, as well as satisfy my compulsive need to disassemble anything that strays within arms' reach, I decided to attempt to improve the heat removal capabilities of the cooling system. I am satisfied with the results, and am providing my process and analysis in the hope that it will help others who are experiencing similar issues. For the short version, skip to the summary at the end.

 

Symptoms:

The CPU frequency would turbo to 3.6 GHz, with an accompanying temperature spike to around 95 degrees C, followed by a reduction of frequency to 1.8 GHz and reduction of temperature to 60-65 C. During gaming or heavy load, the CPU frequency sometimes lowers to 1.6 GHz. CPU throttling as reported by Intel Xtreme Tuning Utility engages within seconds of high load being applied to the CPU. GPU temperatures under load were consistently 70-75 C.

 

Testing:

The core voltage, cache, and graphics offsets were reduced by .100 V. Turbo boost short power max and turbo boost power max were set to 44 W, and turbo boost power time window was maximized at 96 seconds. All settings were applied using XTU. The performance increase was noticeable, however I did not do any testing to determine the battery life and temperature improvements. These changes were applied shortly after purchase, so I don't feel that my subjective observations will be reliable.

 

Intel Burn-in Test was used for CPU loading. Blender OpenCL rendering was used for GPU compute testing, however I can't remember exactly how I ran the test before repasting, so take those results with probably 50% confidence. For idle testing, I datalogged for 5-10 minutes while the computer idled. I do have to mention that MS's background shenanigans (Windows 10) really added an unnecessary amount of complexity to the idle measurements, but I attempted to account for that in the data analysis.

 

I should also mention that I only ran a single before and after test for each category. I did however do many exploratory test runs, and noticed that the temperature stabilizes rather quickly, so I decided that for the purposes of what I was doing a single run would adequately represent the system performance and save time. XTU was used to monitor thermal, power, and current limit throttling, while Open Hardware Monitor was used to monitor everything else due to its logging ability.

 

Procedure:

I began by removing the bottom cover and cooling assembly. The next step was to remove the existing thermal paste. Although only a few months old, the paste seemed rather dry, and came off mostly in flakes rather than smearing off. The next step was to level the CPU and GPU cooler mounting surfaces. I covered the surfaces with permanent marker, and lightly sanded with 1000-grit sandpaper on a small wooden block. The GPU hardly required any sanding. The CPU mount was in pretty good shape, but had minor but obvious deformation where the heat pipes connect. I didn't think to take any pictures. Should have taken pictures.

 

I do want to caveat this for anyone wanting to do the same mod, but hasn't leveled a cooler mount before. I don't know that the sanding will have much or any effect in the long run. It wasn't bad, but I finished it anyway since I had already started. If you do sand the surfaces, just remember that the goal is not to remove material, although that will be a result. The goal is to flatten the surface. Covering the surface with permanent marker provides an indicator as to the condition of the surface, and prevents removal of too much material.

 

I didn't think to have any thermal pads handy for the GPU memory, and since the goo on them was only barely disturbed when I removed the cooler I just left it.

 

The surfaces were cleaned with alcohol before reassembly. Noctua NT-H1 was chosen as the new thermal paste, and was applied as a (small) pea-sized dot to the CPU and GPU.

 

Results:

I ran the tests after reassembly. Ambient temperature for the before and after runs was 70F/21C. Since I left frequency-stepping enabled, I normalized the measurements by comparing CPU package temperature to package power draw. This removes any uncertainty introduced by trying to relate individual core frequency and load to temperature. A second-order polynomial trendline is used to smooth the data points.

 

Figure 1: Idle Package Temperature (C) vs. Package Power (W)

*Note the excessive noise of the data points. Upon observing the load and frequency data, it seems to me this noise is caused by the cooler temperature lagging behind the frequency and load oscillations induced by the background activity in Windows.

 

Figure 2: Intel Burn-in Test Package Temperature (C) vs. Package Power (W)

 

Figure 3: Package Temperature (C) vs. Package Power (W) under GPU Load

*Note: As stated previously, I have reduced confidence in the absolute comparison on this graph, I must have rendered a different scene. This graph is still however included because it displays trends consistent with the other tests.

 

Analysis:

Testing after the cooler modification and repaste reveals consistent performance improvement at high CPU power states. One interesting observation is the downward concavity of the plots while using the NT-H1 thermal paste. This suggests the thermal properties of the paste make it more effective at higher temperatures, which agrees with the observation of slightly worse performance at lower CPU power, as compared to the stock paste. Take this merely as speculation, as I have not done any research or testing to identify the cause, and am focused more on the results.

 

Subjective Observations:

Using the stock thermal paste, when the CPU was initially loaded it would turbo to 3.6 GHz, and the temperature would spike to roughly 95 C. Within seconds, the CPU frequency would reduce to 1.6, 1.8, or 2.5 (MAYBE 3.0+) GHz depending on the situation, and temperature would reduce to 60-70 C. XTU would report thermal throttling at this time. Behavior during gaming was similar, with accompanying intermittent stutters in gameplay.

 

After the repaste and cooler mount leveling, when the CPU is loaded, the frequency will raise to 3.6 GHz and temperature steadily rises to 70-80 C. Turbo frequencies of 4 GHz (3992 MHz) have now been observed. The interesting observation is that the CPU now maintains a higher temperature under load than it did originally. However, it does so while maintaining 3.6 GHz, whereas originally it would only maintain non-turbo frequencies. Gaming performance is similar, with the CPU typically maintaining 3.0+ GHz while the GPU is under load. Additionally, no throttling whatsoever has been observed (according to XTU) since the modification has been completed.

 

GPU temperatures have shown a moderate improvement, lowering from roughly 70 C to 60 C under load. There has been no observed difference in frequency range or GPU performance.

 

Summary:

With the CPU and GPU cooler mounts leveled and high-quality thermal paste applied, the maximum CPU temperature has lowered by 15 degrees C. The CPU temperature under load has increased by roughly 10-15 degrees C, however thermal throttling is no longer applied. This allows the CPU to maintain 3.6 GHz under load, versus 1.6-2.5 GHz while thermal throttling. Low load/idle temperatures have increased by roughly five degrees C, although fan activity at idle has not increased. GPU temperature under load has decreased by 10 degrees C. No GPU performance improvement has been observed, although the additional thermal headroom presents a greater opportunity for overclocking.

 

Recommendations:

The cooling system improvements have resulted in a tangible benefit and in my experience have been worth the effort. For performance improvement, replacing the thermal paste with a high quality paste should be sufficient. In my case, the cooler mounting surfaces were not deformed enough that I feel they would have had a significant impact on the modification results.

 

For fan noise and battery life improvements, this modification has not provided noticeable benefit. Fan activity remains similar, although with significantly improved processing power. The best way I have found so far to reduce fan noise is to restrict the CPU from achieving turbo frequencies.

 

[genexis_x]

I like this. Detailed info on the CPU results. And showed how bad is the stock paste in laptops nowadays.

[deferentlemur]

Thanks. I tried to make it thorough enough to be of value to those with experience in cooling mods, but not so much to discourage newcomers from trying to get the most from their system.

 

You raise a good point about the paste, I was completely surprised by how bad it was. I previously repasted a 1.5 year old 6200u and a four year old 4700HQ, and the paste on neither one of them was as bad as this stuff which was only around three months old.

[genexis_x]

5 hours ago, deferentlemur said:

and the paste on neither one of them was as bad as this stuff which was only around three months old

I had a 10+year old Dell laptop where its thermal paste is dry as hell. Still, the temp difference is like around 5C only after repaste.