NEW Aero 14 w/ GTX1060

on Linux; for @lhl

I have Arch Linux with i3wm running on my Aero 14v6 (Skylake/1060). Partitioning was pretty easy. I will say one thing though, you NEED to use a USB 2.0 flash drive (not a newer 3.0). Otherwise the stick will not boot in UEFI mode.

Regarding Optimus/bumblebee, I set this up using instructions from the Antergos Wiki.

https://antergos.com/wiki/hardware/graphics/bumblebee-for-nvidia-optimus/

This works very well. You don't need to install the patched bumblebee as the issue was fixed upstream. You can do optirun <any program> and it will run it with the 1060, and then switch off the GPU after it is done with it.

I've had a couple of glitches with waking up from suspend (in particular i3lock doesn't activate until 5 seconds after resume) but apparently this is a driver bug and should be fixed soon. NVIDIA recognizes the issue and is working on fixing it is what I've heard.

You also need to add i8024.kbdreset=1 to the kernel line to make sure the touchpad works.

My battery life is great on Linux, about 8 hours of normal usage on a full charge. About the same between Windows and Linux.

HiDPi stuff is also easily resolved. I just set Xft.dpi: 160 in ~/.Xresources and run xrdb merge ~/.Xresources and programs scale fine. For i3 elements just add exec xrandr --dpi 160 to the end of your i3 config file.

(You can specify any DPI, I prioritize screen real estate over text size so I get an effective resolution of 1951x1097 by doing this; I set Windows scaling to 175% as well).

 

I have questions regarding the Aero 14, since I'm stuck between this and the Razer Blade 14, and I've read about the differences between the two several times, yet I haven't read anything about these,

• Does the Aero 14 have an anti-ghosting keyboard?
• If it doesn't, how many keys can be hit together at max?
• For the macro keys, how many keys can be set to one button?
• How should I clean the Aero 14's exterior? As I prefer the green one.
• Is the Aero 14 durable enough to withstand the chances of shipping damages?
• If it does come in safely, is there anything I should do once I turn it on? Like installing/uninstalling certain drivers, programs, etc.

If I were to get one, I've read that GenTechPC is the best place to get the Aero 14 cheaper (with the free shipping), but as it will be shipped for free, I do feel that it may come in with some damage.

Also there seems to be a New Year Promo ongoing with some TIM, so is the one cent Liquid Ultra worth it? Or should I just go for the IC Diamond? I didn't read about any big differences with thermals between the two, but something I did hear was about the Liquid Ultra corroding aluminum parts cause of the gallium it has.

Though the I'm not sure about the durability of the Blade 14 either, that's why I was thinking about getting the Blade at a local MS Store, but I was told about the Aero 14 and its battery life compared to the Blade 14's, which got me stuck here.

This is pretty much my first time trying to get a laptop online (more so a customized one), so I'd rather be safe than sorry about not knowing anything. If this is posted in the wrong place, please tell me, since this is my first post after all.

 

Nope, didn't find any disassemble video, even not of the older versions. The back is relatively easy to open, but the screen seems to be connected on the other side (judging from the pics I saw).

Do you want to replace the screen?

 

Yeah, I was hoping to be able to replace the screen.

I saw that a different model Gigabyte laptop (P34W) uses adhesive to hold the bezel in place, and the Aero 14 has a similar looking bezel. Apparently that would be difficult to remove.

 

I know the laptop uses nvidia optimus. However i was wondering if the Aero would be able to use Gsync with an external monitor. Anyone know if it works?

 

It isn't possible unless there is a display port, one that is directly connected to the GPU.

 

@D2 Ultima - It seems that no one is reporting alarming GPU temps with this laptop. How is that possible when it is performing similarly to other 1060s (ie. Not TDP crippled) ? What do you think they've done to keep the heat under control (besides the obvious)? Some people are however reporting CPU temps of upto 93C.

@hmscott - your opinion on the temperatures?

Sent from my SM-G935F using Tapatalk

 

Got any videos with on-screen metrics? Like FPS, GPU util and clockspeeds in realtime, CPU util and clockspeeds in realtime, and temperatures?

Because Razer owners are reporting all sorts of things and nobody is willing to provide any videos like that.

MSI Afterburner and RTSS (which comes with MSI AB) and Shadowplay to record the video is basically all that's needed; everyone has access to them. Just play and record a game for about 30 minutes so we can see temps level off. Unlocked FPS, no gsync/vsync, we get to see all the performance statistics in real-time.

If those things hold 1670MHz+ boost consistently and run cool in games that are decently hard on a system (Crysis 3, Evolve Stage 2, BF1, Dying Light: The Following, GTA V, etc) then numbers just don't lie. If further tuning of temperatures (like some liquid ultra on the CPU) keep everything in check, then hey, I'll personally pick it up as a recommendation for a decent 14" laptop with some power under the hood, as I currently don't have a laptop to recommend for that size, and the P641HK with the 1050Ti sin't strong enough for a lot of people who ask.

Also, I'd like to know an average of ambient temps for the vids. It'd be a slightly abnormal run if it's in like a 5c to 10c room or something hehe.

 

Can't find any videos like the one you want, the best we have is screenshots posted sporadically in this thread.

Sent from my SM-G935F using Tapatalk

 

If people here have the machine it'll take them under an hour to prepare it =D.

To be fair, I ask this of most anyone who tells me they have ridiculously low temps for these things. Even for P870DM3 owners.

 

Fair enough.

@bruor could you play a game like BF1 for 30 minutes and make a video of it with things like FPS, GPU clock and temps, CPU clock and temps? If not, could you please tag some other owners who nay be willing to do this?

Sent from my SM-G935F using Tapatalk

 

The only games I have are overwatch, Doom, and sc2

Sent from my Nexus 6P using Tapatalk

 

I have BF1 and could possibly do a video in a few hours recording temps and stuff through afterburner if that works for ya
I played last night for a few matches and seemed to get lower temps than earlier this week. GPU stayed below 82C and CPU below 89C. V-sync on seems to yield more FPS for me for some strange reason too.

 

That would be great! Thanks a lot.

Sent from my SM-G935F using Tapatalk

 

I asked this earlier but no response- can anyone tell me how much the AC adapter weighs?

 

This video review has the weight of the laptop and laptop + adapter starting at 02:30:

 

Got my Aero 14 today! Kaby Lake! It was sent in such a way as I couldn't have it routed to a pickup location, I had to physically be at my house to get it. I was hedging my bets and went in to work and left to try to make sure I would be home and still be able to work some hours today. I actually left about an hour before I planned just in case, and wouldn't you know it, it showed up not 5 minutes after I got home. Went through the setup before I had to go back to work. Its a nice system, looks great. Can't wait to get all my stuff loaded on it.

 

Uhh, I dunno. They just made it right? And the 1060 is a really good card. If you don't know, the laptop version is clocked 10% lower than the desktop version to help with cooling. Not much else to say, I've seen the temps myself. It can get hot but elevating the device so there's better air flow goes a long way. In comparison to the blade, it has bigger vents. Otherwise I'm pretty sure the MSI Stealth(?) with the 1060 has good temps too

 

I might find this out for you. It does have a mini display port and I've got a 144hz monitor with gsync. I'll have to get a mini display port to display port adapter though

 

7% slower, not 10%. A lot more power limited too. Hence why I would like a video with it running that shows the performance metrics. And it was announced before even Pascal was available if I remember correctly. Pascal isn't worth much if it doesn't boost and short benchmarks don't tell the story of gaming sessions, so I must ask for the whole story.

Like I said, I am prepared to eat my words... with proof.

Sent from my OnePlus 1 using the force

 

Okay So I had trouble getting the FPS footage lol. I could never really get above 85% GPU usage in BF1. Additionally, whenever I'd record using MSI afterburner it would cut the FPS in half effectively limiting the real performance of the CPU/GPU. I figured it would be a waste to upload that. Side note: When not recording my GPU doesn't seem to go above 80C and my CPU stays under 90C. This usually nets 60FPS on medium-ish settings on native resolution. I will continue to tweak some settings to see if I can increase the GPU usage. I manged to give the GPU an OC earlier of about 200/300 which made it stay at 100% usage in BF1 but it crashed

Update: I've found the sweet spot for the GPU OC and have manged a score just over 10K in Fire Strike

http://imgur.com/a/vYXHH

 

So it's not good enough for me to tell you I played for several hours of Metro 2033, Fallout 4, Titianfall 2, and Wolfinstein the new order, all at 1080p default max settings and my Temps peaked at about 86C according to CPUID monitor?

 

If no one else provides video footage as hard data, then I will provide footage in about 2 weeks when my Aero arrives.
@D2 Ultima Is Witcher 3 taxing enough?

 

Should be, I think. I've never run it. But I know it's demanding so why not

 

This is a nice machine. Getting used to the keyboard and the touchpad is a little bit of a pain. Did some stress testing, CPU stress in XTU they went up to about 65 and then with Firestrike the GPU was in the 70s. Had a -.150v on the CPU but Firestrike crashed so backed it off to -.125v. Got about 9500 on Firestrike which seems pretty good for me.

 

Although I have to admit that I don't much care about what laptops you recommend to people, I was a bit curious on how my system would run so I ran a variation of your suggested test.

I used Witcher 3 set to Ultra, 1080P, Vsync off and just let it run. FPS ended up in the high 50s (below the 60fps max frame rate cap) and pegged at 99% GPU usage and w/ decent CPU usage (30-40%), which seems about right to me for the current gen of AAA games.

After a few minutes, GPU temps stayed pretty stable at 77-78C with GPU clock bouncing around 1658-1709 MHz, and didn't seem to budge any time I checked over the entire half hour. (limiting bounced between NO LIMIT and POWER - I'd love to be able to try undervolting the GPU, but sadly, looks like it's locked and BIOS modding seems like it'd be more trouble than its worth)

CPU cores were a bit more variable, temps bounced from 75-81C, average usage around 35%.

My Aero 14 is a V7 Kaby Lake from Gentech PC w/ the $0.01 Liquid Ultra repaste special they had, and my laptop was on a Allsop TriTilt stand (which actually blocks the about half the input ventilation on the bottom, so not as optimal as something like a Roost might be). Ambient temp of my apt is 23C.

Couple screens and a short phone video of the final numbers (I use bare drivers w/o Game Experience installed and 30min of streamed gameplay seems excessive anyway):
https://www.dropbox.com/sh/hy6rexxe0wpbcpy/AAACI7K8A_0jxYzgxKHSJKo1a?dl=0

 

Also I will add my GTX 970 desktop system only scored 8700. So my laptop scoring that much higher than my desktop, awesome! I now I am not optimizing on either system but still.

 

Very interesting. Glad for video, though I was hoping for a longer one. Your CPU temps for such low usage are somewhat worrying to me, however if your GPU is so pegged I can't say much. Liquid Ultra upgrade is interesting to hear for such a machine.

Mind if I get you to check Dying Light if you have it? This is good information to me. Dying Light should be generally harder on the CPU while still pegging the GPU, so it'll be a good test.

As for your stand, even if you think it's blocking some intake, lifting the machine alone grants a lot of cooling potential, so that was certainly a benefit to you. I'm sure. But this is good, I like options for laptops. Now the real question is why did the Aero 14 handle the 970M badly? The 1060 is a significantly hotter card (it's quite literally an overclocked 970M with a more condensed die size, in fact).

 

I don't have Dying Light but I play The Division, which is fairly CPU intensive - uses about 70-80% CPU in missions. I added a screenshot to the Dropbox folder so you can scope out the numbers - I have the graphics set to High where it's just about 60FPS. GPU is at 80C and stays there, freq reaches 1721MHz, hangs out a bit higher than in Witcher 3 (still power limits). My CPU gets to about 90C, which is on the high side (when I do CPU-only testing, temps are more reasonable) - It doesn't seem to throttle though, CPU core frequency sticks at 3.38GHz throughout my entire play session (also added a screenshot).

The 1060 is a Pascal chip - mine is built on TSMC's 16nm FinFET process (apparently Nvidia has recently signed to build some Pascal chips on Samsung's 14nm FinFET but I think that'll probably start w/ new products vs existing ones), while the 970M was 28nm Maxwell, so I don't think you can really say that they're literally similar at all (different architecture and process), although from what I've seen the 1060 should have higher TDP than the 970M. I suspect the actual answer is that the cooling system has been significantly improved/changed from the last generation. It's to my understanding, at the very least, a fair amount bigger (sadly, it takes up the space of the 2nd M.2 slot from the previous generation).

 

Hmm, interesting - my Arch install stick that I use everywhere is a SanDisk Extreme USB 3.0 stick and I had no problems booting it w/ UEFI (it's just a straight dd of the iso)

Thanks for the info. I ended up installing Bumblebee and power usage dropped from 20W+ idle (4.5h battery life) down to 11-12W (8.5h) which is reasonable, at least - I guess if you don't have Bumblebee installed the Nvidia GPU simply doesn't get disabled.

I have my HiDPI sorted both in X and on the console (switching to latarcyrheb-sun32 does the trick there).

I'm not sure I noticed the trackpad issue, but come to think of it, I may have only been booting in Linux w/ my external kb/mouse attached, so thanks for the pointer.

One surprising issue I have is w/ a lot of dropouts w/ my 8260 - I'll need to spend a bit of time digging into iwlwifi settings I think.

I also haven't even touched the keyboard buttons yet - if there's any setup for that, I'd be interested to hear it, I remember having to do some manually binding w/ my previous ThinkPad. I'm Arch w/ Openbox since there's a lot I ended up customizing, but there sure is a lot of yak-shaving involved.

 

Thanks. You're at 80% util at 81c being power limited (though your boost is high because you're actually not at 99% util) and your CPU is indeed very hot at just 73% util. I think the notebook can't handle simultaneous stress of the parts... it seems the cooling system must be of the shared type. When you have lower CPU util or lower GPU util the parts keep individually cool. Witcher 3's test was surprising but Division's test was disappointing. Gigabyte should have done more work, especially if you're using liquid metal to help heat transfer (it can drop temps by 15c in some cases, I've personally experienced a 12c drop from using it on my processor). This means the cooling system is pretty much at its limit. Did you have it still propped up?

I am aware it is pascal, however the specs are otherwise the same. Pascal is die-shrunk, overclocked Maxwell. There is no architectural benefit to using Pascal in terms of raw number crunching ability on the clockspeeds. The die shrink allowed Pascal to clock higher (above 1500MHz; to the 2000MHz range easily) and use less power for the performance, but its die size has shrunk and it has a lot of cores packed in for the size of its die. As a result, it is extremely thermally dense. As far as why I said a 970M and a 1060 are the same:

GTX 970M = 1280 cores, 192-bit memory bus, 6GB GDDR5 vRAM.
GTX 1060 = 1280 cores, 192-bit memory bus, 6GB GDDR5 vRAM.

The 1060 is a die-shrunk 970M.
The 1070N is a die-shrunk 980N (2048 cores each, 256-bit memory bus each, 8GB GDDR5 vRAM each).
The 1050 is a die-shrunk 750Ti/860M/960M (640 cores each, 128-bit memory bus each, 2GB/4GB vRAM each).
The 1050Ti is a die-shrunk 950 (768 cores each, 128-bit memory bus each). Only difference here is 1050Ti has a 4GB vRAM version which to my knowledge the 950 did not have.

1070, 1080, Titan X Pascal, 1060 3GB are the only full new cards for Pascal architecture.

Maxwell was extremely cool because while being on 28nm, it had very large die sizes which spread out the heat well. It was easy to get heat transferred and thus easy to cool. Pascal no longer has this benefit, and is clocked significantly higher. It's hotter than Maxwell, Kepler, even Fermi. In fact, the reference cooler for the GTX 580 (GF110, full-cored big die Fermi) is used on the GTX 1080 (GP104, full-cored midranged Pascal) and cools the 1080 worse than it does the 580, and is a 250W-rated vapor chamber cooler (note that the 1080 is only 180W rated). The Titan X pascal isn't even full GP102, and it has an even beefier vapor chamber cooler that barely works (it's so bad that watercooling the Titan X Pascal even without attempting overclocking bumps its performance up from averaging 25-30% over a 1080 to 40-45% over a 1080).

Explain your drop-outs? Is it huge ping spikes/inconsistencies every few minutes, or does it drop connections entirely?

 

While I don't disagree with the overall characterization of Pascal as mainly a die-shrink (at least in terms of general architecture and how the CUDA cores, geometry front end, ROPs are all balanced) there are still a fair number of differences. I think the one that's maybe most relevant as far as thermals go is the updates to GPU Boost - from what I've read there's much more fine grained voltage/clock adjustment and I suspect there's a lot of fine tuning in the BIOS of the mobile SKUs, and could explain a lot why there's a lot less throttling w/ than the previous generation.

For those following along, to clarify, I've only seen this issue when I'm in Arch Linux - the 8260 has been rock solid in Windows. I'm getting recurrent disconnects - it may be that it's simply not happy w/ some configuration in my AC router, there are some config options/troubleshooting I've found discussion/documentation on, but I've been busy/lazy so I'll just be working my way slowly through all my various Linux issues issues (setting up my original Thinkpad for example took weeks, and ended w/ a 10-page setup doc. Luckily there's not as much stuff to worry about this time around that I haven't solved previously, but I've been pacing myself. Today's Linux accomplishment was fixing Optimus.)

 

Its indeed not only a die shrink. The chroma compression algorithm is new, and the basic architecture is far more refined. But it is true it is built upon Maxwell.

Still the 1060 is 2x as fast as the 960. The 1070 is 2x as fast as the 980M and the 1080 is 2x as fast as the 980GTX. These impressive results are not only because of the die shrink.

 

Of course there's differences, mainly in the back end. Things the architecture supports, like better async compute, etc. The voltage variance band is indeed much smaller, but this does not directly have to do with GPU boost and a custom video BIOS on Maxwell can force voltage to a constant; it isn't really an architecture "feature" but rather nVidia doing silly tricks to get "efficiency" (that introduces instability, no less). Pascal is indeed designed to far more heavily rely on boost, though. I will give it that. The throttling was always based on a TDP limit, for the most part. Maxwell was more aggressive with its TDP limits, but custom vBIOSes removed that. The throttling I speak about if I say a last gen model throttled is always going to be thermal, though. Power throttle I will always specifically point out if I mean it.

I see, I won't be able to help then. Must likely be a driver conflict or something to that effect if the card proves it is capable of operation in another OS.

 

High 70's, low 80's in terms of degrees Celsius is pretty normal for this processor and inside a very thin Unit like this. The max temperature on this processor is 100 degrees Celsius, so unless you're hit high 90's and are approaching this point, these temps seem normal.

 

Is this with the newest nvidia graphics driver? Just curious because I expected the kabylake version to perform better because of the improved cpu.

 

Yeah I had to load the latest manually, GE told me the driver that was loaded stock was from 10/16 and said ti was up to date. I was just running the Fire Strike, I didn't really do any tweaks except for CPU voltage, I was able to get a few more points with overclocking the GPU. But I didn't have the power mode in performance or anything.

 

That is very nice!
While I am very much happy with the skylake version, I feel that I should have waited couple weeks more. I am seeing good results with new kaby laptops. Does your laptop run at lower watt per hour while idle on desktop or just browsing/doc/pdf? Thanks!

 

I am not sure and I am not really in to setting up a way to measure this. But I do know when I was just sitting on the desktop the battery said I had 12 hours left!

 

I have seen that much time left on my skylake unit.

Sent from my Nexus 6P using Tapatalk

 

That is true, my laptop also shows around 12 hours. I used hwinfo or batteryinfo. Anyways, I think kabylake's advantage is being able to run faster while using the same power.
Thanks for the information!

 

I forgot to mention I played a lot Dying Light on it too, in 1080p and 1440p. Obviously ran better in 1080p but it otherwise ran as well as my 970 desktop. Ran smooth as butter

But, I tried dolphin emulator on it and was disappointed with how some games ran. Super Mario Sunshine ran really well but I couldn't run Metroid Prime at 1080p and max settings like I can on my desktop. I got it running pretty well but with lower settings. This is where I noticed the difference between my laptop and desktop. My desktop has an i7 4790 clocked at 4ghz and it seems like that made the difference. Or maybe there's some dolphin setting I'm missing. Now that I think about it, I'll double check to see if it was using the gpu

 

This. My main laptop for a long time was my Alienware m15x which under load always ran at 100C and about 50C at idle. I even ran it 24/7 for w years to seed torrents and it still works fine after 7 years . These temps are great in comparison

 

I'm no expert, and I might be wrong, but here is my impression: transistor and pipe sizes inside the processor have gotten much smaller than 7 years ago. So much smaller, in fact, that now they're reaching the limits of physics; this means that they probably cannot tolerate temperatures as high as was tolerable 7 years ago. Smaller transistors are less stable than larger ones, and heat further decreases stability/lifespan.

 

Since we have some Linux users here I thought I would ask. I am wanting to run Linux native on the laptop to do some CUDA/OpenACC development for school. I don't want to partition the internal hard drive but figured an external msata SSD in a USB3 enclosure might give me the speed I need. Its just going to be programming anyway.

I found some guides on setting the Linux system up to enable the Nvidia GPU and not worry about the intel. Since it is a secondary environment for me I don't care about Optimus (bumblebee) just getting access to the GPU.

So I was thinking the easiest way would be to disconnect the M.2 drive and then perform the install on the USB SSD. Then put it all back together and when I wanted to boot into Linux just connect the USB SSD and change the boot order to USB (or just change it to always try the USB first) so I can just boot into Linux. When done, power down, remove the USB SSD, and boot back into Windows. Does this sound right?

 

Nvidia sent out its latest video card drivers.
Any chance it stops the stuttering with g/sync and dedicated graphics switching without the need of running emptyproject11 ?

 

I have the 1/24 drivers and I renamed the emptyproject so it wouldnt run. I am not sure I can really detect the stutter. Maybe I am just not doing enough for it to affect me.

 

Considering the Aero 14 from XoticPC, so a few questions:

SSD
What's the default SSD in the aero 14 v7?

Is it worth while choosing 500GB 960 Evo or the 512 960 Pro over this (which one)?

Or would 1TB Samsung 850 EVO be better than the 512 GB NVme options?

Wifi
Choose the built in wifi chip (again can someone confirm it) or choose from these:

Bigfoot Networks Killer™ Dual Band Wireless-AC N1525 with BT4.0/4.1
OR
Intel® Dual Band 8260-AC (2x2)

Cooling
Is the IC Diamond Thermal Compound preferred over stock thermal compound?

General Questions:
Does optimus work in terms of battery life?
How much throttling occurs?
How loud are the fans at full load? (witcher 3/crysis 3)

 

Is there that noticeable a difference between a SATA3 SSD and an NVME? I know benchmarks are probably nuts, but when it comes to observable speed difference. Unless you're doing video editing and need to seek massive files, then I could see the desire for it.

Sent from my Nexus 6P using Tapatalk

 

The two scenarios I've found that reliably trigger the stutter (about a half second where you can't move the mouse) is clicking on the battery meter in the tray or opening the Display settings (right click from the desktop) - in neither case does anything show up if you're monitoring the usage in Nvidia GPU Activity.

I've been running w/o emptyproject11 (well, actually it complained to me on startup when I just renamed it, so I replaced it with a simple noop application) and besdies those two scenarios have yet to find anything else that causes stutters like that. Day to day, it doesn't really impact my life too much, although it'd be nice if *someone* went ahead and fixed that.