Gigabyte Aero 15 X9 Thermal Testing, AI, Overclocking and Undervolting
So lets take a look at the thermals and see how we can improve performance using the new AI feature, undervolting and overclocking.. In my configuration here, Ive got the Intel i7 8750H CPU, but the X9 is also available with the i9 8950HK.. Otherwise, the other major heat generating component is the 2070 Max Q, graphics and the higher tier Aero. 15 Y9 is available with the 2080 Max Q. Ill, also be comparing the X9 with the 15X, with 1070 Max Q in a future video, so get subscribed for that. Thermal testing was completed with an ambient room temperature of 24 degrees Celsius so expect different temperatures in Different environments. Also keep in mind. There are heatpipes shared between processor and graphics, so a change in one component may affect the other. Ill. Also briefly, talk about Gigabytes, new AI feature that comes with the v9 Aero laptops, so the new ones with RTX graphics.. There are two modes: AI Edge and AI Cloud. AI Edge processes, the data locally to make the best decisions in terms of what to change like fan, speeds and power limits. While the AI Cloud setting will send data to Microsoft and get the best settings for the workload based on other user submitted data, so Im expecting it to perform better.. Otherwise, all other tests had the AI disabled unless stated otherwise. Starting down the bottom of the graph. At idle, the temperatures were on the warmer side, as the fans were basically silent.
. The gaming results were tested playing Watch Dogs 2, as I find it to use a good combination of both processor and graphics.. The stress tests were done by running the Heaven benchmark and the Aida64 stress test. At the same time, as a worst case, scenario., The CPU was undervolted by 0.12v as denoted by UV on the graph and the graphics overclock was with a 90MHz boost to the GPU core. In every single other test. The CPU was thermal throttling at 90 degrees Celsius. As shown by the blue bars, while the 2070 Max Q graphics was thermal throttling in most tests any time the green bar was at 86 degrees. Celsius., The graphics only stopped thermal throttling when the fan speed was manually maxed out and at the top, with the Thermaltake Massive 20 cooling pad in use. I got the best results.. These are the average clock speeds for the same tests. Just shown. Lets start towards the bottom, with the gaming tests., Despite thermal throttling taking place for each of these tests, there are clock speed differences from the different settings.. Both of the AI settings were giving better performance compared to just running the laptop at stock. However, from what I could tell the AI Cloud setting ran with a higher fan, speed which reduced thermal throttling, which is why I think it was ahead of the AI Edge results., Either way with my manual 0.12v undervolt to the CPU 90MHz overclock, to the graphics and With the fans manually set to maximum Im, getting a big improvement over both stock and the AI settings.
, With the stress tests its worth, noting that, with the fan on stock but graphics overclocked, I saw the worst GPU performance so far it was already thermal throttling. So this basically makes it worse until we boost the fan speed. At the top of the graph, with our best case result using the cooling pad. The 8750H was 500MHz from being able to reach the full 3.9GHz turbo boost speed due to thermal throttling.. Despite the thermal throttling its worth, keeping in mind, the gaming performance tested previously was still fairly decent.. These are the clock speeds I got while just running CPU only stress tests without any GPU load and were seeing higher clockspeeds now as theres less heat in the system. Without the graphics contributing. At stock, I was hitting power limit throttling for the first time, rather than thermal throttling., By boosting the power limit with Intel XTU. The clock speed was improved. However, this now resulted in thermal throttling, which is why the undervolt, at the top of the graph, allows us to hit the full 3.9GHz turbo boost speed of the 8750H.. These are the temperatures for the same tests just shown, and we can see that with the power limit boosted, the 90 degrees Celsius throttling was reached while the undervolt improved temperatures by 7 degrees Celsius, while also boosting clock speed.. While it was possible to boost the power limit of the CPU, this was not attempted in the gaming and stress tests shown previously theres no point, as thermal throttling was taking place without it.
We only need to look at boosting the power limit if were hitting power limit throttling, but that wasnt the case there. To demonstrate how this translates into performance. Ive got some Cinebench CPU benchmarks.. While it was possible to get full CPU performance in Aida64 with the undervolt. Only in Cinebench I still had power limit throttling at the default 52 watt TDP, so I had to boost it a bit to almost hit 1200 just a little under what Id normally expect from an unconstrained. 8750H.. Here are the GPU, only clock speeds, while under a graphical only stress test at stock and with a 90MHz, overclock applied.? These are the temperatures from the same tests, no change with or without the overclock, and not reaching thermal throttling at 86 degrees Celsius, as we saw in the combined CPU and GPU loads earlier.. So how do these performance boosts actually translate into games? Ive tested with the exact same Windows, Nvidia and game updates, installed. Far Cry 5 was tested using the built in benchmark at 1080p, with ultra settings. Down the bottom. Ive got the results purely from running the laptop at stock.. Just above this is with the fan set to full speed, which gives us the second best performance., The AI Edge and AI Cloud settings are giving better performance compared to stock with the default fan, but simply setting the fan to full speed was enough to provide better Performance as it helps reduce the thermal throttling.
. Finally, at the top, once the CPU is undervolted by 0.12v and the graphics overclocked by 90MHz on the core and 400MHz on the memory with the fans at full were getting the best performance, not surprising. I think manually tuned settings will always win.. As for the external temperatures, where youll actually be putting your hands at idle, it was in the low to mid 30s warmer up the back and cooler over the WASD keys.. While gaming, the palmrest and touchpad, were still fairly cool, the middle of the keyboard was fairly warm, though getting to 50 degrees. The WASD keys are still sitting at around 30 though. Similar deal with the stress test. Running. The back gets to the high 50s as hot air is exhausted below the screen, but I dont think its an issue as you dont rest your hands there. As for the fan noise produced by the laptop Ill. Let you have a listen to some of these tests. At idle. It was almost silent and then, while gaming and under stress test with the fans at default, it was actually very quiet, much quieter than most other. Similarly, specced laptops, Ive tested. With the fans maxed out. It was quite a bit louder now about the same as many other laptops Ive tested and as we saw earlier, the fan boost, improves, CPU and GPU clock speeds by a fair bit.. Overall, my Aero 15 X9 is a hot machine, although at stock the fans arent actually that loud and as weve seen performance is negatively affected as a result – and this is with the i7 8750H and 2070 Max Q, the Y9 is available with i9, 8950HK and 2080 Max Q so Id be interested in seeing how that goes in the same body.
. I didnt personally see much improvement with the AI feature either. As we saw before, I was able to get much better results with my own simple tweaks, though, of course most people probably wont be doing that. Compared to running at stock with AI disabled. I was seeing some improvements with the AI, but better results were had by simply maxing out the fan, which is something anyone can easily do.. I think in future, as different workloads are learned, it may improve over time. It could just be that the AI isnt currently optimized to best handle the workloads that I was testing, especially the AI Edge setting, as I presume that learns from my behaviour, whereas AI Cloud has more data to work with.. These differences in performance shown arent hard and fast rules. There are different factors which will vary results, primarily the temperature of the room, youre running in application of thermal paste, and even the specific hardware which comes down to the silicon lottery.. You may not be able to undervolt or overclock your hardware the same as me. It depends on the chip and its specific power requirements, so dont just blindly copy my settings and do some testing to find out where your stable point is for best results.. It may be possible to further improve temperatures by swapping the thermal paste. However, as this is a review unit Im not able to try that., If I change the paste, the next reviewer will unknowingly report incorrect thermal results.
Due to what Ive done. I have swapped. The thermal paste on my personal Aero 15X, although I didnt personally see a difference there and many people who have gone to liquid metal have also noted minimal changes. So it may just be the case that the cooler isnt sufficient to dissipate the heat, regardless of the contact material between the chips and heatpipes.. In any case, undervolting and raising the fan, speed is much easier for anyone to do and as we've seen it did improve performance in this laptop.. Let me know how much of a performance boost youve found by undervolting your hardware and what you thought of the improvements here and dont forget to subscribe for the full review of the Gigabyte Aero 15 X9 laptop, as well as future tech.