Picked up a new laptop: Gigabyte Aero 15x v8

[Albuquerque]

For reference, I'm coming from a Lenovo Y460 "budget gaming laptop" from 2010 -- Allendale i5-520M, 8GB ram, Radeon 5650M, and a 240GB Vertex 3 SSD. It's been a fine laptop, but there's a hairline fracture in the display cable which makes the device "wig out" when the screen is angled a certain way.

Ordered this on Saturday, arrived at my desk today: https://www.gigabyte.com/us/Laptop/AERO-15X--i7-8750H#kf I picked up the 8750H + GTX 1070MQ + 144Hz FHD + 512GB NVMe model, and the first two hours of tinkering with it have been good so far

Obviously I'll pop in here later with some thoughts from loading up Steam and grabbing some games, but I'm pretty happy with how it's been reviewed. There will always be opportunities for improvement; the common complaints are the touchpad needing some driver work, and the fans making noise when you get it all loaded up and hot. If that's all they have to complain about, I'm pretty happy already

Anyway, I'm loading up ThrottleStop and MSI Afterburner so I can undervolt the CPU and GPU (respectively), which seems to be the very first steps everyone takes on the new gaming rigs.

Also curious to see how this fares against my i7 3930k, 1TB Sammy Evo 850 and 980Ti...

 

[Malo]

Wow that looks amazing. It's incredible how thin they can get the form factor now with considerable hardware. Of course sustaining clocks might be an issue, if be interested to see how it fares with maintaining high clocks.

 

[Albuquerque]

Yeah, thermal throttling is one of those things you have to keep an eye on.

Much to that point, I spent several hours in ThrottleStop's FIVR controls last evening to undervolt the various functional components of the CPU. I used the IntelBurnTest tool to spot check compute performance and thermal output as I made voltage changes. Interestingly enough, you can see the calculation rate slow down as the voltage drops below a certain threshold. With the lower voltage comes less total package power, which doesn't necessarily result in lower heat, but instead higher attainable / sustainable boost clocks.

Sure enough, -160mv @CPU core, -50mv @ cache core, and -50mv @ IO core the CPU found another 200MHz (+2 multiplier) of sustainable boost clock for a total of 3.6GHz with all 12 logical cores under continuous load. I ran another 100 loops of the IBT "extreme torture" test to ensure the lowered voltages didn't have longer-term stability issues; the CPU came up to 81*C with the fans operating around 50-60% and finished in about four hours. Average clocks stayed at 3.6GHz until the end of the test.

Tonight will be GPU torture.

 

[Cyan]

hot stuff! What an amazing machine. How did you manage to undervolt the GPU? My i7-7700HQ 1050Ti laptop doesnt allow me to do that when I use MSI Afterburner. I am using default settings for both the CPU and the GPU 'cos I am not always running intense tasks and when I do I turn the laptop cooler on. I always have the laptop sitting on the cooler, it's just that I don't use the cooler all the time, of course.

 

[CSI PC]

hot stuff! What an amazing machine. How did you manage to undervolt the GPU? My i7-7700HQ 1050Ti laptop doesnt allow me to do that when I use MSI Afterburner. I am using default settings for both the CPU and the GPU 'cos I am not always running intense tasks and when I do I turn the laptop cooler on. I always have the laptop sitting on the cooler, it's just that I don't use the cooler all the time, of course.

In general the Nvidia voltage is locked on laptops as you probably realised.
But have you tried the Core Clock Curve option and raise the frequency earlier relative to voltage axis in MSI Afterburner while keeping power limit at 0 and current-ish version (4.3.0 onwards but not sure which version had 1050 voltage compatibility added)?
It is meant to be accessible by <Ctrl>+<F> with 4.3.0.
Apologies if that is what you tried.

 

[Albuquerque]

^^ I am indeed using the voltage vs speed controls in MSI Afterburner. It's not the same as undervolting the CPU where a discrete voltage offset can be universally applied at all speeds. Rather, it's adjusting the ramp rate so clocks are "allowed" at lower voltages. Thus, I can only use the voltage options supplied by the graphics card firmware (ie 800mv is the lowest allowable selection), but I can push the ramp rate down so that more of the turbo clocks fit under the 800-900mv area of the curve.

Unfortunately, I've hit a snag. Afterburner is unable to monitor the voltage of my particular 1070Q processor, similarly it appears to read the voltage curve incorrectly. I presume this is a function of firmware, as I'm also unable to use any NVIDIA installer to upgrade my video driver. This situation was encountered with prior versions of the Aero 15x line; it was solved pretty quickly last time. I knew this was a possibility via reading reviews, so I'm not unhappy (yet!)

Good news: I still have access to the clock slider, and I can get the GPU clock to settle right at 1.3GHz after running the MSI Kombuster furry donut v2 stress test on the fullscreen 1080p preset for two hours. This is also with the CPU turbo'd up to 4GHz. Calculating for the difference in CUDA core count * core clock speed, this puts my laptop about 15% behind my overclocked 980Ti desktop card. I think this is just fine

Finally had a chance to play some games: Prey, Witcher 3, the new Endless Space, even a bit of Fallout 4 because why not? The fans are certainly audible given the workload, and I did dabble a bit with the software for adjusting fan speed / ramp profile. Since I game with headphones on, I never actually hear them. Put in probably five continuous hours on Saturday night -> Sunday morning with Endless Space, which isn't super demanding, but the laptop provided a good experience the whole time.

I've got it at work with me today; gonna carry it to meetings to take notes. Will report back on battery life and noise for non-gaming things.

 

[homerdog]

I presume this is a function of firmware, as I'm also unable to use any NVIDIA installer to upgrade my video driver. This situation was encountered with prior versions of the Aero 15x line; it was solved pretty quickly last time. I knew this was a possibility via reading reviews, so I'm not unhappy (yet!)

How do you go about updating the GPU driver in that case?

 

[Albuquerque]

How do you go about updating the GPU driver in that case?

Yeah, that's the $100 question... I have a few theories: First, I may try disabling switchable graphics in the firmware, and instead direct the machine to use only the NV display chip. I also have a clean Win10 Enterprise "Windows To Go" OS install loaded on a pricey SSD-Fixed-Disk-on-USB-3 key, which I haven't yet tried pulling the NV installer into just yet. It allows me to do OS-destructive testing without actually modifying the image on the internal laptop drive.

Also, based on the Windows To Go image I created, I know Windows Update will pull down a WHQL video drive update even including the NVIDIA control panel software. At a minimum, this reasonably assures me the laptop is never critically outdated, but also never so new as the Game Ready stuff you can find directly on the NV website.

Part of me wonders if it's still possible to modify (hack) INF files to force compatibility. The root problem is the installer never finds a compatible product...

 

[CSI PC]

^^ I am indeed using the voltage vs speed controls in MSI Afterburner. It's not the same as undervolting the CPU where a discrete voltage offset can be universally applied at all speeds. Rather, it's adjusting the ramp rate so clocks are "allowed" at lower voltages. Thus, I can only use the voltage options supplied by the graphics card firmware (ie 800mv is the lowest allowable selection), but I can push the ramp rate down so that more of the turbo clocks fit under the 800-900mv area of the curve.

Unfortunately, I've hit a snag. Afterburner is unable to monitor the voltage of my particular 1070Q processor, similarly it appears to read the voltage curve incorrectly. I presume this is a function of firmware, as I'm also unable to use any NVIDIA installer to upgrade my video driver. This situation was encountered with prior versions of the Aero 15x line; it was solved pretty quickly last time. I knew this was a possibility via reading reviews, so I'm not unhappy (yet!)

Good news: I still have access to the clock slider, and I can get the GPU clock to settle right at 1.3GHz after running the MSI Kombuster furry donut v2 stress test on the fullscreen 1080p preset for two hours. This is also with the CPU turbo'd up to 4GHz. Calculating for the difference in CUDA core count * core clock speed, this puts my laptop about 15% behind my overclocked 980Ti desktop card. I think this is just fine

Finally had a chance to play some games: Prey, Witcher 3, the new Endless Space, even a bit of Fallout 4 because why not? The fans are certainly audible given the workload, and I did dabble a bit with the software for adjusting fan speed / ramp profile. Since I game with headphones on, I never actually hear them. Put in probably five continuous hours on Saturday night -> Sunday morning with Endless Space, which isn't super demanding, but the laptop provided a good experience the whole time.

I've got it at work with me today; gonna carry it to meetings to take notes. Will report back on battery life and noise for non-gaming things.

Yeah it is a pain in the arse way to have to undervolt-manage those GPUs going by the frequency/voltage envelope rather than using the simpler offset, especially when testing settings with behaviour.
I wonder if some of your problems come down to compatibility between Afterburner and Max-Q; like I mentioned each version of Afterburner has additional models added for core voltage control and using done with reference design in mind (which the Max-Q may not be depending upon technical context and was musing if it makes a difference to Afterburner).
http://www.guru3d.com/files-details/msi-afterburner-beta-download.html

That said you cannot change the mv core volt but can increase or reduce the frequency range that operates at it, so if default at 800mv is 1200MHz you could increase the clock to 1400MHz (comes back to model compatibility with version of Afterburner) - yeah these are not the exact clocks for said voltage just to give the idea but consideration then comes stability.

 

[CSI PC]

^^ I am indeed using the voltage vs speed controls in MSI Afterburner. It's not the same as undervolting the CPU where a discrete voltage offset can be universally applied at all speeds. Rather, it's adjusting the ramp rate so clocks are "allowed" at lower voltages. Thus, I can only use the voltage options supplied by the graphics card firmware (ie 800mv is the lowest allowable selection), but I can push the ramp rate down so that more of the turbo clocks fit under the 800-900mv area of the curve.

Just to also say generally 0.8V is incredibly low for a dGPU and I doubt it would be possible to go lower even if an option was available with regards to actual 3D gaming loads and importantly stability; the performance envelope for 14nm/16nm generally is 0.9V-1.15V (going by a few presentations) and even just below 0.9V can be tricky without special tweaks at a firmware level in terms of stability; to put it into perspective look at the voltage profile (set to always be stable) for Polaris with 3D loads.

For Nvidia in general when it goes into protection mode against such as Furmark the GTX1060 drops to 0.87V.
Max-Q purpose is also to adjust behaviour-stability of the performance envelope with regards to the dynamic boost calculations/management beyond how it would normally operate.

 

[Albuquerque]

Just to be clear, you've not told me anything I didn't already know.

Just like my undervolting of the CPU, the goal here isn't to overclock, it's to (potentially) avoid thermal and power limits which might stand in the way of a better turbo clock bin. I'm not overclocking, I"m simply doing what I can to provide the headroom for the GPU to achieve a target for which it was already configured. Since I cannot currently adjust voltage in any meaningful way, I decided to push the clock instead. However, based on anecdotal evidence from other online resources, undervolting my GPU is likely (but not guaranteed) to yield the results I truly seek

Irrespective of this, I've solved my GeForce Experience issues, and now know the root cause of the driver upgrade issues.

GeForce Experience was somehow shipped in a broken state from the factory. I removed everything with an NVIDIA name, then reinstalled GeForce Experience to see if it would pull a driver -- and it didn't. Still, at least now it didn't crash on start or fail to connect to the NV servers, which is a step in the right direction! I let Windows Update download driver version 388.xx dated Dec 2017, and then GeForce Experience told me I had the latest driver. It also now let me "optimize" my games and futz with the other "experience things."

I went ahead and re-downloaded the newest Far Cry 5 game-ready driver from the NV website, and the installation failed with no compatible devices. A little bit of Dir *.inf /s /a | find /i "10de" | find /i "1ba1" returned a driver file which described easily there dozen different 1070 Mobile chipsets, each with different subsystem ID's. Did a quick jaunt through device manager, and I see my 1070 chipset is one digit different than one of the models described by the driver INF. I created a new descriptor for my own VEN\Dev\Subsys identifier and ignored the "signature invalid" prompt, and the installer did try to install. It then failed because Win10 rejected the driver because my modification of the INF didn't check out with the signature in the driver CAB repo, and so I got the finger.

Thus, it seems a fix can't be far off from the NV side. MSI Afterburner might be another story, but with a sustainable 1300MHz GPU clock, I'm not sure if I care.

Oh, my first gripe with the laptop. Can't be all roses, right? :

I carried the laptop around with me to meetings all day yesterday, and immediately noticed my battery life was estimated at 3hrs which seemed crazy low. Display brightness was 40% (which is actually quite good level IMO) and I had the machine in balanced performance mode. Upon investigation, I noticed the CPU clock would not decrease below 2.2GHz, even when functionally idle (<5 % utilization indicated.) Windows 10 Performance Monitor indicated the NV GPU was not active either.

I used the Gigabyte configuration utility to set the machine into Battery Save mode, and suddenly I had 7+ hours of battery life expectancy. The only change I could detect was the CPU would actually idle down to speeds which I would expect (just a bit over 1GHz), everything else seemed the same. However, later in the day I plugged the laptop in to show a coworker the video card performance, and the refresh rate was locked to 30Hz. I remembered my prior profile change, so I went back and re-entered balanced mode -- and still had a 30Hz refresh rate. Finally I entered Performance mode and, tada, full refresh rate again.

I'm not convinced this utility is useful; I think it's doing stupid things to the Windows CPU power profile and somehow in the GPU driver, and I don't like it. I'm strongly considering a fresh OS install with a Win10 Pro SKU so I can grab Hyper-V functionality, just to rid myself of any other oddball Gigabyte "functionality" which I may not appreciate.

Anyway, sorry for the long reply. Hope you enjoy reading!

 

[CSI PC]

Just to be clear, you've not told me anything I didn't already know.

Just like my undervolting of the CPU, the goal here isn't to overclock, it's to (potentially) avoid thermal and power limits which might stand in the way of a better turbo clock bin. I'm not overclocking, I"m simply doing what I can to provide the headroom for the GPU to achieve a target for which it was already configured. Since I cannot currently adjust voltage in any meaningful way, I decided to push the clock instead. However, based on anecdotal evidence from other online resources, undervolting my GPU is likely (but not guaranteed) to yield the results I truly seek!

Glad your investigations working out.
I personally dislike Geforce Experience.
Sure but my post was also meant for others and probably more in direction with Cyan's post that we both started to discuss this around, apologies if you feel it was all directed at you; please appreciate just expanding on above which it looks like you already know but not necessarily others.
Not many know that 14nm/16nm has a performance envelope that usually operates at 0.85V up to 1.1V and going below the lower tolerance may cause stability issues depending upon operation-frequency and firmware-driver, the manufacturer presentations with this guideline are not usually public or if they are one would need to trawl HotChips.
I have seen some specs for 16nm high performance products for Vlow spec to be below 0.8V but it is far from normal operation and was outside of dGPUs.
Maybe part of this is semantics/POV because the way I am describing the use of Afterburner is similar to how most generally undervolt Radeon; lower voltages with higher or same clock frequency for P6/P7 state to overcome thermal throttling related issues, just in this instance with Nvidia/Max-Q you are at the other end of the scale and instead of P state one is using the absolute voltage axis.
So by moving all the default clock rates lower down the voltage axis you kinda are undervolting/underclocking and reducing thermal impact/power demand unless by undervolting you mean as an absolute never going above the new predefined voltage (which you would like below 0.8V) for any state and clocks-operation and yeah I can see your point in that context.

As a reference to fixed clocks to standard voltage profile here is a game managed with the clock rate forced to extreme savings on a GTX1060 which I guess is the closest you could get to if wanting to always operate at 0.8V and with just one state for the best stable frequency you can identify; for GTX1060 this is 1500MHz, 0.85V at 62W but from defaults so not tweaked for lower clock (slightly improve TDP-TBP further but not possible without firmware-driver/direct interface) or higher clocks at 0.85V for slightly improved performance at low TDP.

Just to help for reference are you ok sharing your clock curve profile?

But just like I said no-one should expect to undervolt dropping voltage well-stable (depending upon clock-power and operation) once your below 0.9V and especially wanting to push below 0.8V ; not directed at you but others who think it is OK to try or expect such undervolting with Max-Q or even possibly dGPUs generally to such low envelope extremes as I have seen comments in various places looking to try and do that.

Just curious as it is notably different, what is the lowest core clock you can set with your Max-Q GPU?
Not sure I have seen that investigated in reviews.

 

[Deleted member 2197]

^^ Unfortunately, I've hit a snag. Afterburner is unable to monitor the voltage of my particular 1070Q processor, similarly it appears to read the voltage curve incorrectly. I presume this is a function of firmware, as I'm also unable to use any NVIDIA installer to upgrade my video driver. This situation was encountered with prior versions of the Aero 15x line; it was solved pretty quickly last time. I knew this was a possibility via reading reviews, so I'm not unhappy (yet!)

Have you tried the new Afterburner 4.5.0 released on 4/25/2018?

 

[Cyan]

In general the Nvidia voltage is locked on laptops as you probably realised.
But have you tried the Core Clock Curve option and raise the frequency earlier relative to voltage axis in MSI Afterburner while keeping power limit at 0 and current-ish version (4.3.0 onwards but not sure which version had 1050 voltage compatibility added)?
It is meant to be accessible by <Ctrl>+<F> with 4.3.0.
Apologies if that is what you tried.

thanks for the suggestion, I haven't tried that yet. In fact I dont use MSI Afterburner on this laptop, but Palit Thunder Master. I am not in a hurry to try that given how efficient the 1050Ti is at dealing with power consumption and heat -lack of extra money aside, it was one of the reasons I got this laptop model, but mostly because it is light (1.8Kg). Plus I use a cooler -turned off most of the time but it helps the air flow anyways. Having a blast playing games like Crysis 2 at 1080p 60 fps all the time on the 1050Ti, and the game is also playable at 4k, but I haven't connected the laptop to my 4k monitor yet.

 

[Albuquerque]

Ha! You guys are all over this (in a good way )

Sure but my post was also meant for others and probably more in direction with Cyan's post that we both started to discuss this around

This makes sense, no worries at all

I personally dislike Geforce Experience.

Yeah, I mostly share your opinion on GeForce on the desktop. For laptops, it's marginally useful because it allows some specific capabilities for managing per-game framerate profiles to extend battery life and keep the fans at a lower speed. It's also kinda-sorta useful for keeping your driver up to date, since it notifies you directly when an update is available versus me coming in here, reading about it, and then wandering to the website to grab it.

Just to help for reference are you ok sharing your clock curve profile?

Unfortunately, MSI Afterburner (even the newest 4.5.0 as suggested by Pharma above) cannot view voltage, and apparently also cannot manipulate the voltage. I also know the clock curve is wrong, because it starts at 1400MHz / 0.800mv and ramps from there. This is likely accurate for a desktop card, but it's utterly wrong for this particular Max-Q chipset. As such, I know it's not working and therefore not worth sharing the bad data.

Also, as of about 9am this morning, NVIDIA published the 4/25 driver set which now includes the new PCI + Ven + SubSystem ID specific to the chipset in my laptop. GeForce Experience notified me of the driver, and it installed without issue. So, I'm now back in the regular production support loop for new drivers which makes me happy

If I never get access to voltage monitoring or control, then it's a bummer but not the end of the world. The laptop is doing very well for me, I think my only complaint now is some of the Gigabyte software inconsistencies / glitches. Things like not keeping the battery charge profile (Please stop charging my battery to 100%) and whatever weirdness happens with the performance profiles forcing the CPU to never enter LFM in the Windows balanced power profile.

 

[Albuquerque]

Was tinkering over the weekend, and I figured out two things...

First: As you might expect, the Gigabyte device management software (fan speed, screen calibration, battery charge limit, performance mode stuff) makes really stupid changes to the default Windows power management schemes. While trying to understand why battery consumption is so lopsided, I noticed core parking is fully disabled in all power profiles, even power savings mode. Full transparency: I'm not a fan of core parking for anything to do with gaming, but I expect (and even want) the behavior in power savings mode (ie I'm doing something mundane in an Office app.)

I looked through Powercfg, and then went into the registry to dig further. There's some really bizarre stuff, like limiting maximum CPU frequency to 50% in any battery mode except Full Performance, but then making threshold changes so it almost never enters MFM / LFM in any profile except power saver. They also completely disable USB selective suspend and any NVMe power savings states, they also disable the Realtek USB3 card reader from entering selective suspend. I suppose all of this could be chalked up to making sure performance is always on tap, but why can't I have a truly power savings mode, and then a truly balanced mode, and then a truly performance mode? Windows offers these schemes for good and simply GUI-switched reasons, why must they make it painful?

I exported the proper power profile defaults out of my big desktop gaming rig, and reloaded them into the laptop. One reboot later, and holy hell is everything behaving much better. Balanced mode on battery now allows the CPU to wanders into the full gamut of LFM / MFM modes (1.1GHz++) while still retaining full turbo speed on demand. High performance mode sees the CPU bottom out at HFM (2.2GHz) and moves in and out of turbo like you'd expect, rather than pegging itself at 4GHz and running the fans at 50% at all times. Power saving mode also now sees core parking return, and selective suspend works in the non-high performance modes.

But my battery life still wasn't right somehow...

Second: This laptop comes with a 1080p 144Hz IPS display, and it's buttery smooth. Upon continued investigation, I discovered refreshing at 144Hz makes the Intel iGPU kick into a higher speed. Short version is: according to Windows power management, reducing the refresh from 144Hz to 60Hz doubled my balanced profile battery life. Using my aforementioned newly corrected balanced profile settings and 144Hz refresh, I'd get right at five hours of life at half brightness (plenty bright for an office setting) and the WiFi enabled and me hacking on the keys. Flipping to a 60Hz refresh, I used it for almost six hours, put it to sleep, woke it up again, and used it again for almost four hours before it needed to charge. I probably got 9 1/2 hours out of it, and at least a tiny bit of that was consumed in suspend.

That's more like it I know Intel produced at least one version of their driver where a transition to battery power profile would allow automatic changes to refresh rate. Doesn't appear to be an option in the 630 video controller. Might look into making a little Windows task scheduler widget that fires on power source change, invokes a WMI call to detect which profile I've entered, and make a call to swap refresh rate accordingly.

Finally, did you guys know about this?
There's a combination of Intel GPU and Windows 10 bug which makes switchabe graphics-enabled laptops hitch and jerk when the GPUs switch back and forth on mundane desktop tasks.

Intel appears to have their part solved: https://communities.intel.com/thread/119093
Microsoft now has a fix in one of the preview builds, not sure when it will hit mainstream: https://answers.microsoft.com/en-us...-problem/93e7004a-62b1-4211-8e37-4c136608865e

I don't seem to have the problem, but it appears to be pretty pervasive.

 

[Cyan]

Was tinkering over the weekend, and I figured out two things...

First: As you might expect, the Gigabyte device management software (fan speed, screen calibration, battery charge limit, performance mode stuff) makes really stupid changes to the default Windows power management schemes. While trying to understand why battery consumption is so lopsided, I noticed core parking is fully disabled in all power profiles, even power savings mode. Full transparency: I'm not a fan of core parking for anything to do with gaming, but I expect (and even want) the behavior in power savings mode (ie I'm doing something mundane in an Office app.)

I looked through Powercfg, and then went into the registry to dig further. There's some really bizarre stuff, like limiting maximum CPU frequency to 50% in any battery mode except Full Performance, but then making threshold changes so it almost never enters MFM / LFM in any profile except power saver. They also completely disable USB selective suspend and any NVMe power savings states, they also disable the Realtek USB3 card reader from entering selective suspend. I suppose all of this could be chalked up to making sure performance is always on tap, but why can't I have a truly power savings mode, and then a truly balanced mode, and then a truly performance mode? Windows offers these schemes for good and simply GUI-switched reasons, why must they make it painful?

I exported the proper power profile defaults out of my big desktop gaming rig, and reloaded them into the laptop. One reboot later, and holy hell is everything behaving much better. Balanced mode on battery now allows the CPU to wanders into the full gamut of LFM / MFM modes (1.1GHz++) while still retaining full turbo speed on demand. High performance mode sees the CPU bottom out at HFM (2.2GHz) and moves in and out of turbo like you'd expect, rather than pegging itself at 4GHz and running the fans at 50% at all times. Power saving mode also now sees core parking return, and selective suspend works in the non-high performance modes.

But my battery life still wasn't right somehow...

Second: This laptop comes with a 1080p 144Hz IPS display, and it's buttery smooth. Upon continued investigation, I discovered refreshing at 144Hz makes the Intel iGPU kick into a higher speed. Short version is: according to Windows power management, reducing the refresh from 144Hz to 60Hz doubled my balanced profile battery life. Using my aforementioned newly corrected balanced profile settings and 144Hz refresh, I'd get right at five hours of life at half brightness (plenty bright for an office setting) and the WiFi enabled and me hacking on the keys. Flipping to a 60Hz refresh, I used it for almost six hours, put it to sleep, woke it up again, and used it again for almost four hours before it needed to charge. I probably got 9 1/2 hours out of it, and at least a tiny bit of that was consumed in suspend.

That's more like it I know Intel produced at least one version of their driver where a transition to battery power profile would allow automatic changes to refresh rate. Doesn't appear to be an option in the 630 video controller. Might look into making a little Windows task scheduler widget that fires on power source change, invokes a WMI call to detect which profile I've entered, and make a call to swap refresh rate accordingly.

Finally, did you guys know about this?
There's a combination of Intel GPU and Windows 10 bug which makes switchabe graphics-enabled laptops hitch and jerk when the GPUs switch back and forth on mundane desktop tasks.

Intel appears to have their part solved: https://communities.intel.com/thread/119093
Microsoft now has a fix in one of the preview builds, not sure when it will hit mainstream: https://answers.microsoft.com/en-us...-problem/93e7004a-62b1-4211-8e37-4c136608865e

I don't seem to have the problem, but it appears to be pretty pervasive.

hmmm, I experienced that but several times --my Intel GPU is also a 630. The worst thing about it is that so imetimes the system become unresponsive and restarts by itself without a warning to save your content --lost part of my progression in an Unity project 'cos of that. But most of the time the screen just jerks and it returns to normal, but a square of like 100x100 pixels shows totally corrupted graphics, so it's better to restart, and you lose your MSI True Color (best MSI program, imho) settings.

 

[Albuquerque]

Still had some flaky behavior over the past few days, specifically with power management functions. My corporate Microsoft rep found me a Win10 Pro upgrade key, so I completely wiped and reloaded the box last night with the new April 2018 OS media. I sourced all the versions of the drivers I wanted to use, got it all installed without any of the Gigabyte flair, and also got Hyper-V turned on.

A few things have suddenly started working... I can open the lid and it wakes from sleep, the Powercfg /Energy report no longer shows any errors (selective suspend not working) or warnings (timer resolution value too small, high CPU usage from certain services), and opening task manager on battery power doesn't cause a huge pile of context switches for whatever reason. It always booted quickly (maybe 15 seconds from power button to logon screen) but the speed has increased dramatically with the reload; it's less than ten seconds from button to logon now.

Also now the "silent" mode of GeForce Experience works; it will framerate-throttle a game as to avoid the fans kicking on high speed, while on battery. It's a bizarre sensation to play Crysis 2 without the fans being audible. It's throttled to roughly 30fps, with little but visible peaks and dips. Still with high fidelity settings, so it's entirely fine for playing on battery. Not sure why this hadn't worked before...

I hacked around on the configuration of MSI Afterburner a bit. I tried a change to the raw configuration file for the 1070MQ chip, and it's now reporting some level of voltage awareness. I'm not sure it's good data though, it shows between 650 and 720mv under high load, with lower voltage as clocks increase. Pretty sure this just isn't a thing which is going to work...

 

[Deleted member 13524]

At first I thought your title had a typo, but Gigabyte really does have a brand named "Aorus" and another named "Aero".

 

[Albuquerque]

So, my laptop has gone to shit in the last few days, and it's a bad NV driver.

The combo of the full Windows reload and a set of known-good drivers cleaned up all my bad power management complaints; battery life while running regular Office apps or the Edge browser was regularly seeing 9+ hours, gaming performance was excellent, all around I was very happy. And then I got greedy and downloaded the newest NV driver back on the 9th. That was a patently awful idea.

Laptop is now physically warm on the bottom, the fans never stop moving, battery life went from 9 hours to 90 minutes, boot time extended significantly (added easily five more seconds on a boot time that started at below 10 seconds), and now I'm getting the occasional DXGKRNL BSoD after resume and occasionally after a cold boot (really a hybrid boot, which mirrors my findings of the resume function doing the same.)

Thought maybe it was a bad interaction with an older Intel UGH 630 driver, so I downloaded a slightly newer version of the Intel iGPU setup to no avail. Finally went to rip out the current NV driver, and discovered I can no longer download the April 25th / 397.31 version of the NV driver -- it's been pulled from all their site. I have a copy on a USB key at home, but I'm travelling and it's gonna be a few days. As a quick fix, if I go into the BIOS and turn off the dGPU function, the battery life immediately goes back to "normal" and so too does the boot performance and the resume / boot stability.

Thanks NV, you dicks. </rant>