I couldn't have come up with comparable benchmarking tool so measured coping speeds pf of a 2 GB .zip file on Linux ( with ext4 file system, partition only in 35% full ) with and Windows 10 ( NTFS file system, partition in around 50% full ). To clarify - I have copied the file only inside the same partition between the user folder and the desktop, so no NTFS to ext4 / ext4 to NTFS copying.

It turned out that Linux took on average around 20 - 30% more time to copy the file than Windows. Is it normal? Is ext4 under Linux slower than NTFS on Windows 10 or do I lack some firmware?
My specs:
- Fedora 26 with 4.13 kernel and Windows 10,
- UEFI: legacy suport and AHCI mode set to on; Fast and Secure Boot options set to off,
- Notebook - HP Probook 430 g4:
- m.2 WDC SSD with the capacity of 250GB
- 4GB of RAM 2133 MHz DDR4 ( + 4GB of SWAP on Linux )
- Intel Core I3 7100U,

I don't doubt it.

I'm not sold on the fact that linux is better in every way and under every instance.

Every few weeks you see a test of some linux kernel and some hardware and some file system(s). They can only compare linux to linux in these tests usually. The test involves many different tests and you end up getting an idea of a winner for some use. Your test is a single test.

The file system on windows 10 is pretty advanced.


Way way way back when linux was trying to make an OS you have to realize that they didn't get all the internal hardware information from hardware makers. The close integration between microsoft and hardware makers would lead me to believe that not only are windows OS's fast but that hardware has been designed for it. The companies that make hardware almost always do it with windows in mind. Linux driver coders still can't get all the hardware information and they tend to make the drivers on their free time using what they have.

what commands did you use?
i notice that my filemanager seems to have problems in certain situations, so i'd use the command line.
it is extremely important to add a sync to your command, like this:on both systems, linux and windows (unfortunately i do not know what the windows equivalent is).
only when the prompt returns after the sync, should you stop your watch.

and show us some numbers.

The Core I3 7100U is a special processor which is designed to run on low wattage. It can run off a mobile phone as well as your notebook.
The 7100U is designed for energy efficiency rather than speed.

Intel also makes Kaby Lake processors for medium wattage (laptops) and high wattage (desktops).

It is possible that different rules are in place between MS and Linux for file transfer speeds to keep the temperature low on your processor.

Nonetheless, I have to agree that the Windows OS is faster than the Fedora OS on file transfer speed.

It's weird. I have created a 6GB Truecrypt file and copied it on ext4 and ntfs ( ext4 to ext4 and ntfs to ntfs). Here are the results:
Fedora:
through Nemo: 1:20s
through Nautilus: 1:40
through Terminal ( cp [FILE] [dest]; sync ) 2:06

Windows:
through windows explorer: 40s

How is it soo much different?
Oh and cpu usage when coping files on linux and windows is about 10 - 20% with cp / file manager rarelly exeeding 10%.

What do you mean there are medium-wattage cpus for laptops? I have thought all new Kaby Lake CPU run with 15W TDP

when windows explorer reports that the file was copied in 40 secs it doesn't means the copy was really finished. In fact this operation may still run in background, though filemanager already reported that the job done. The same for Nemo and Nautilus.
In fact only cp+sync command shown real copying time.

Just remember that when you copy a file to USB drive and then trying to disconnect it with "safe disconnection" you have to wait few seconds until copy will really finished (though it was reported as already done) and you can safely remove a drive. If you will disconnect a drive immediately after filemanager report, without "safe disconnection" (in fact it just a sync command) mostly you will get damaged file, it will be copied partially.

So these tests with file managers (explorer, nemo etc) not objective.

You might get more realistic results when copying to external drive and applying "safe disconnection". Time from copy starting to report that device can be safely removed

1 members found this post helpful.

Maybe it is like that and it takes a bit longer.

The last 20 years I have not been able to copy a directory on Windows without the copy action stopping at least once asking me a dumb question about overwriting, a file being locked for reading (!), a file being in use or a system file.

So I prefer a Linux copy operation which is 3 times as slow but continues unattended over a Windows operation which might be faster, but overall slower. If you do a copy action in Windows taking 20 minutes you better keep watching it to push buttons.

I agree with previous posters saying that Windows developers have better access to hardware internals.

And the bottom line: who cares? As long as Linux does what I want it to do and do it without constant interaction needed I am more than happy.

jlinkels

A simple search on Kaby Lake will show you that Intel makes each generation of cpus to run at many different temperatures.

If an i7-7700K is put on a laptop, the battery will run out too quickly.
A desktop pc doesn't need a battery, so it can use the i7-7700K.

But the need to create an i5 family and an i7 family is pure daylight robbery.
If Intel can make an i5 - then they can make an i7 at the same cost.
But as we know, we have to pay much more for the i7.

On top of that, to overclock the cpu (to get the most bang for your buck) you need to have a 'specially made' i5/i7 on a 'specially made' motherboard.
But the Taiwan motherboard manufacturers have already proved that any Intel desktop cpu can be overclocked - not just the expensive ones.

But Intel has put a stop to that.
Why? Cos you simply ain't paying enough money to Intel.

PS:
https://stackoverflow.com/questions/...nd-for-windows