Hacker News Reader

Comment by hacknat | Hacker News Reader

I can personally confirm this. I bought a brand new Dell XPS 15 with Skylake (i7) in December. I installed Linux on it (kernel 4.3), and it has been a power management nightmare from day 1. I've only ever gotten 2 hours from the battery. I just sent it in because it powercycles at random now, never making it past the Dell splash screen anymore. When I run the hardware diagnostics, sans hard-drive, it dies in the middle of one of the processor tests.

Other people have been pointing out that Windows is struggling with Skylake as well, and I've heard the same.

Skylake was touted by Intel as being one of their proudest achievements in power management to date. My guess is that their changes were so drastic that the software didn't keep up.

Edit:

I do have an NVMe hard drive, which does seem to cause some issues, for reasons passing understanding.

View on HN

Perhaps this explains why Apple have been so slow to get their Skylake MacBooks to market.

I was disappointed other OEMs had beaten them to it, but it looks like they just dumped hardware on the market without suitable software. Apple obviously take responsibility for both.

Yeah, I was writing the same exact though before seeing your comment. Seems plausible. Maybe they didn't had enough cpu for the Iris Pro 15" macbook pro, but there seemed to be no reason why they hadn't updated the macbook/13" macbook pro yet.

If you visit hackintosh sites, the community is trying to make osx work properly on skylake, but there is no clear path to a solution yet either.

"..it looks like they just dumped hardware on the market without suitable software"

Intel has been doing this too with the NUC. If you go on the Intel forums, you'll find people having serious problems right now with Skylake NUCs [1].

Even the previous generation of NUCs (that were released over a year ago) still have major bugs with Linux. For example, there's a BIOS bug that reboots the machine instead of shutting down. They've known about it for at least 5 months now and it's still not fixed [2]. And it seems likely to affect all versions of Linux, not just some obscure variant.

[1] https://communities.intel.com/docs/DOC-110236 [2] https://communities.intel.com/thread/88822

My guess is that their changes were so drastic that the software didn't keep up.

It's one thing to be running hot all the time, it's an entirely different thing to actually cause damage because of that.

This reminds me of some laptops a few years ago which would overheat just sitting in the BIOS setup screen for too long, because the fans were entirely software-controlled and that hadn't been loaded yet.

I don't think power management should ever be left to software entirely if it can result in situations like this - software can make the CPU go into a lower power state, but the CPU should know when it's too hot and throttle itself without any intervention from software.

Yes, otherwise you end up with this: https://www.youtube.com/watch?v=y39D4529FM4 ;)

(I'm well aware it's not a realistic experiment, but it's a fun demonstration nonetheless.)

That video reminds me of childhood, good old Tom's Hardware :)

Those poor CPUs. The first two no-heatsink temperatures are lower than my 2600k's idle temperatures - what's changed?

The Intel CPUs throttled to an absolute minimum state of power draw in order to keep itself from frying. Those chips also came with an integrated heatsink, like modern chips, that provides just enough heat dissipation to keep the CPU from frying itself before it can throttle down.

That video was made because AMD didn't incorporate such protective features back in that era. That's why their chip absolutely fries itself within moments of removing the heatsink. 100W through an area the size of a Tic-Tac (die area dedicated to cache uses almost no power) is a recipe for catastrophic failure if left unchecked.

My old Q6600 could do this to a degree. Sometimes the fan wouldn't spin up on boot, so there was just passive cooling. i'd try playing a game/doing something and performance would start digging it's way to China.

At which point I'd take the side of my PC off, and manually spin the fan until it got the idea.

I have this exact machine with an i7 and do not get anywhere as low as 2 hours. Do you have bumblebee installed? It sounds like you might have the dedicated graphics runnign at all times, which is the case if you don't have e bumblebee + bbswitch installed.

This depends on the distribution. Fedora for example has GPU switching using PRIME working fairly well and automatically disabled the dGPU by default.

Wasn't aware of that, though the current performance of nouveau is pretty horrendous. At least for newer cards. It's a bit unfortunate but to essentially have a working card you must have the proprietary drivers installed.

Offtopic: other than power management, how has your experience been with Linux on an XPS 15? I've been holding off with mine until the next Ubuntu release, later this month.

For what it's worth, I have a Dell XPS 15 bought in January 2016, with i7-6700HQ Skylake CPU, and I think it has an NVMe drive (at least the drive ID says so). It's the basic model without a touch screen.

I run Debian-testing (aka Stretch), which ships with linux 4.4 and Wayland. I had weird issues with X11, but I had blamed that on the NVidia/dual-card setup which I did not bother investigating.

Besides that, I've had no issues at all with the laptop. I haven't even run the BIOS update yet. Only annoyance was the lack of select-to-copy in Wayland/Gnome, which apparently is fixed in the new version. :-)

There was a BIOS update that supposedly fixes power management for NVMe drives: http://www.laptopmag.com/articles/dell-xps-13-battery-life-f...

This is weird because I was reading a thread on Ubuntu forums about 4.3 problems and I've just stayed on 4.2 with 15.10 with no power management issues. Power consumption is criminal though getting 1.8 hours average at full charge and low brightness.

Reply on news.ycombinator.com