That is clear, but we are here in speed ranges where this doesn’t really make an impact to the results as the write speeds of the disk are above the interface speed and we are writing continuously a single file.
True, you are limited by USB speeds, but it can still show a difference precisely because you aren’t saturating the disk, depending on how the interrupts handle the “hey, the USB bus is saturated, stop sending” messages and the size of the cache on the disk (and probably a bunch of other factors).
Basically, the disk could write a sector, and then spin past the next sector before it gets the data for that sector from the host (the Vero), and so would have to rotate around again before it can write the sector.
In the old days, disks had sectors on a single track out of order (“interleaved”) to avoid this, and picking the right interleave to avoid the “go around again” depended on things that were all external to the disk, so the best interleave for a particular drive would be different on computers that were very different.
Now, it’s unlikely that this would happen on every sector on a modern hard drive, but it could happen every 5 or 10, etc., and that’s enough to really affect the benchmark. And, such a problem could be exacerbated by NTFS being a user-mode driver.
Well will swap the partitions and do the test again to see if it makes a difference (I doubt)
1 Like
@nabsltd, thanks for elaborating my point!
I thought I was going to have to reply all that information, but you saved me the trouble.
That’s why when I did my test, it was the same drive, completely blank after a format.
Now from there the file system can handle the data differently on how it wants to write to the disk, but you wouldn’t want to interfere with that as that would defeat the point of comparing file systems.
So you use the same drive, with one partition, completely empty, and let the file system do what it wants from there.
In my experience, I have found that the ext4 slightly outperforms NTFS (no surprises there) but it is marginal, and definitely not the cause of such issues like not being able to playback UHD remuxes and what not.
Outside of running things in the background that can be disk intensive, like a torrent client, there really shouldn’t be any real world differences between the file systems when it comes to playback and definitely not on file transferring as the USB2.0 interface would bottle neck transfer rates before the file system.
I appreciate that we all have hashed this out in the open.
I have learned a lot more about the Vero, and *nix OS’ too, thanks to this experience, and I hope it helps others as well.
In the end, the long standing advice should remain intact with maybe an exception about exFAT, which is to stick with ext4 if you plan on leaving the HDD attached to the Vero and all workflow will be done remotely, but use exFAT or NTFS if you also need to attach that drive to something other than a *nix OS.
I would suggest to avoid exFAT personally, at least at this time, and just use ext4 or NTFS depending on your use case.
Personally, since I prefer to attach my USB HDDs directly to my Windows laptop to transfer files at much higher speeds than the Vero 4K+ can handle over LAN, I have found a middle ground of using ext4 and having Linux File Systems for Windows by Paragon installed on my Windows laptop.
For those not wishing to invest $20 in software like that, just go ext4 and transfer files to you DAS via network at slower speeds or go NTFS and get full speed access to the HDD.
Either way, playback should be a non-issue.
Among the major highlights for Linux 5.4 is the initial Microsoft exFAT file-system support, integration of the LOCKDOWN LSM, DM-Clone as a new means of remotely replicating block devices, case-insensitive F2FS support, support for several new AMD Radeon GPU targets, initial support for Intel Tigerlake with Gen12/Xe Graphics (still very much a work-in-progress), beginning to see various consumer Arm laptops working off the mainline kernel, a kernel fix around UMIP to help various Windows games in Wine, and a lot of other new hardware support.
- Long-awaited support for Microsoft exFAT file-system with a preliminary driver being merged following Microsoft recently publishing the exFAT specification and no longer objecting to said support being in the kernel.
Next update to the kernel will have better support for ExtFat giving our Windows users a glimmer of hope
Along with some additonal fixes to various FS systems
-
NVMe-of P2P support for supported peer-to-peer DMA with AMD Zen systems and select Intel chipsets.
-
Improved FSCRYPT support for this file-system encryption framework used by the likes of EXT4 and F2FS.
Source: Linux 5.4 Features Are Huge From exFAT To New GPUs To Enabling Lots Of New Hardware - Phoronix