Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> Fanless laptops are where I'm now looking. The UX360CA Zenbook is fanless, low-power, and very reasonably priced

The fanless Zenbooks are fantastic. But they're also really low power with the Core-M CPUs with a TDP of 4.5 Watts [0]. That's in a power budget that could fit in a tablet.

The Airtop-D has a Core i7 CPU, which has a TDP of 65 Watts (can downtuned to 37W) [1]. You can roughly expect it to have 10x+ more computational power.

It's just a whole different category of devices.

[0] http://ark.intel.com/products/88199/Intel-Core-m7-6Y75-Proce... [1] http://ark.intel.com/products/88040/Intel-Core-i7-5775C-Proc...



You will be sorely disappointed if you expect those CPUs to have a 10x difference. First, you have a 2.75x clockspeed difference and then a double core count so 5.5 times is what you could expect if they were the same architecture but the 6Y75 is Skylake (don't be mislead by the "Core M" moniker, these the same architecture as their larger wattage brethen just with a lower TDP, they used to mark them with an Y postfix like i3-4220Y and the Y is still in the moniker) and the 5775C is Broadwell but it has eDRAM so the difference is negligible here http://www.anandtech.com/show/9483/intel-skylake-review-6700.... So in real world, you can expect to see a 3-5x speedup.


Yes, 3-5x (peak performance) sounds quite reasonable. But 10x is not out of the question under high load for extended amounts of time (like compiling big software) when you take the form factor thermals into account. A laptop like that will start throttling quite aggressively quite quickly.

The Zenbook I have isn't a powerful computer by any measure. It's a fine laptop for light use but the CPU and thermal characteristics aren't suitable for intensive computation. I'd expect any fanless desktop computer to wipe the floor with it when it comes to perf.


This is why I rent a server at Hetzner. Considering you can get an i7 3770 for 33.61 EUR a month, it's pretty economical. It also serves as a backup (2x3TB) and occassional webhost.


Ah, that explains that. Bearing in mind I'm thinking of upgrading from an AMD 350e Fusion CPU, so I know I'll be blown away by the performance of a new machine. It's just picking the right form factor.

I can't wait to see how quickly mutt and vim load. [Removes tongue from cheek]. In all seriousness I do run a VM, a situation I'm trying to address by writing the main software tool I use in Windows as a Vim plugin. The VM works for now, but is obviously a strain on a minimally resourced system.

And if I'm completely honest I'd love to be able to run my business and do all my work from a Raspberry Pi and/or netbook (think Chromebook pricing but entirely Linux friendly). I'm only editing text, why complicate things!?


Did you mean the AMD E-350? That's competition for the Atom CPUs not these ones.


Yes, absolutely. I'm not comparing it to the Core-M/iX range though, just making a point that in my situation either of those will be a vast improvement. Despite the fact that the E-350 does what it needs to just fine right now, it could be a touch quicker.


Thanks for that. It's something I've been trying to get to the bottom of, with a very limited understanding of CPU comparisons. I understand clockspeed is not something that can be directly compared, where MIPS might be more of an apples-to-apples comparison?

In that light the M7 turbos up to 3.1 Ghz, where the i7 turbos to 3.7 Ghz. That's with double the cores/threads. On paper not a 10x improvement, but in terms of MIPS it could well be. In reviews people don't seem to be phased by the new CPUs, they say they can watch HD videos and run VMs as normal, but nobody has yet given a succinct summary of the limitations. Not to mention that in these specific 2-in-1 Zenbooks I've seen the CPUs are Core-M3s, not M7s.

Aside: I didn't know the Core-M CPUs couldn't handle more than 16GB RAM, that's an interesting stat from that spec sheet.

The decision between new desktop or laptop is still to be made. A situation where something like the Zenbook could remote into a more powerful desktop sounds good but in practice would it work - I sometimes prefer to work without an internet connection, completely remotely. Sure I could sync the machines with git, rsync or similar, but the laptop still draws me in. Particularly the 2-in-1 form factor. My Kindle (keyboard!) recently went to Davy Jones' locker and I'm quite happy reading on my netbook if the settings are right. So then the tablet + desktop combination comes to mind. I don't know. Power vs portability still an issue in 2016. Hopefully not for much longer!

EDIT: Found a rudimentary comparison based on the Surface. It doesn't show even a 3x improvement: http://www.lovemysurface.net/surface-pro-4-core-m3-vs-i5-vs-...


> I understand clockspeed is not something that can be directly compared, where MIPS might be more of an apples-to-apples comparison?

There's a joke saying that MIPS is acronym for "Meaningless Information Provided by Salesmen".

It's just not a good measure of overall performance, because there's no guarantee that it translates to real world performance.

The only way to get good performance indicators is looking at benchmarks, and with an emphasis to benchmarks that resemble the workloads you expect to be running.

For ballpark estimates (of similar hardware), the power consumption is a good measure because the performance per watt figure is typically quite close for similar CPU architectures. It doesn't work for GPUs or CPUs of very different architectures, though.

For the comparisons you linked on the Surface, it will tell you how the CPU works on that particular form factor. But it's not a good estimate of the CPU power in general, because a small device like that is always going to be thermally constrained. You can't take this conclusion and apply it to another kind of laptop/desktop.


To go down the rabbit hole even further common benchmarks will run with one-off drivers to give the appearance of better performance.

I've seen vendors skip mem-zero, skew buffer sizes, overclock, disable thermal limits and all sorts of other shenanigans specifically targeted at gaming benchmarks.

Only real way to evaluate hardware is to run the exact loads(or simulacrum of them that's not disclosed to the vendor) which probably overkill for consumer hardware.


Appreciate that, thanks. Will keep up the search for the right trade-off.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: