I’m using 2 Vero4K+ units. Both playing the exact same media file.
The Vero4K+ playing media over the network plays it fine.
Another Vero4K+ is playing media over an USB stick (really really really fast USB stick, capable of running Windows To Go!). This one has roughly 3 frame skips every minute. I could set my watch on it. Is it because playing over USB? Even USB 2.0 should be very capable of handling the datastream. Can you see in the logs if the HDMI is capable of doing 23.976FPS?
NTFS over USB is a non-issue, don’t waste your time changing your file system.
Definitely change your GUI to 1080p, but not 100% sure that will solve your issue.
Even if it doesn’t, you should be at 1080p GUI anyway.
I’m assuming you have a 4K TV since your GUI is set to 2160p.
Letting your TV do the upscaling could very well eliminate your frame issue.
Specs on the USB drive would be helpful too.
This again…
The fact is, high throughout will be limited by having to use FUSE. Paragon submitted an NTFS kernel driver with write support upstream, but it has not been merged then.
Until then, there will always be a significant performance penalty when using FUSE like filesystems such as NTFS.
True, but not to the point at which it will cause dropped frames so no point in trouble shooting in that direction.
I literally have six* NTFS USB hard drives all feeding into one Vero 4K+ USB 2.0 port and I watch all my UHD with Atmos without a hitch.
4x Seagate Backup Plus Hub 8TB external hard drives daisy chained with a 5TB and 4TB 2.5" external hard drive attached at the end.
It’s not even been approved in the mainline kernel yet, so it would be some time before it reaches OSMC.
This is a flawed assumption. Time and time again we’ve seen constrained throughput, particularly noticeable with applications like Transmission when the underlying filesystem is NTFS.
But this guy is asking about dropped frames on video playback, not Transmission performance.
If someone started a thread saying they’re experiencing performance issue running Transmission or other such programs and you’re reply was that NTFS might be the culprit, I’d agree.
Because it very well could be an issue with something very IO intensive, but there is no real world impact to video playback from NTFS formatted drives.
That’s not correct. It’s not constructive to tell users that this is the case when simple benchmarking proves to the contrary. You have taken my words out of context. I said that Transmission clearly exacerbates the problem, but it does not mean it is not present with other applications, and even read operations.
The first thing the user should do is try playback from another medium, such as the eMMC and then report back.
Present to me a video file that will fail to playback on my NTFS formatted hard drive but will playback on the Vero’s internal storage or an ext4 formatted hard drive.
I have tested this extensively and have yet to find a video that plays on one but not the other.
Yes benchmarks will show performance differences, that’s never been debated.
But those differences are beyond the threshold of standard video playback so moving from NTFS to ext4 should make no difference to the OP.
I will get back on this one. My problem is, that the specific movie I’ve been testing is almost 80Gb… This wouldn’t fit on the eMMC. But at the moment I don’t have a laptop with such capacity free to move files over and re-format.
In your previous posts over time, you’ve suggested USB2 as a limiting factor when this is not the case. You will always hit the limitations of FUSE NTFS before USB2.
Any high bitrate material will cause heavy CPU use when using an NTFS filesystem. This will take away time we can use pushing back to the video decoder and render pathways.
You are doing people a disservice by adamantly refusing to acknowledge the performance problems that NTFS can induce.
As I’ve said before, if NTFS works for you, then great. But it can and will cause other people problems. You continue to twist my words.
First and foremost. let me just say I love your work and I love the Vero.
I have no personal issues with you and I think you are a great guy who has gone above and beyond in your support for the product.
That being said, I have to continue to disagree with you on this issue of NTFS giving video playback issues.
There is simply no proof of that and plenty of proof that it works just fine even with the most demanding videos.
Thanks for linking to a test I did that showed problems with the ext4 file system and which I also claimed that my results must be some weird outlier and to be ignored, but existed none the less.
“I wouldn’t take my test results as anything conclusive because I find it hard to believe ex4 write performance was only 9MB/s.”
You’re the one who’s doing people a disservice when you give misleading information about performance like you did in that thread right after the post of mine you just linked.
That is just not true and I really don’t understand why you keep perpetuating this false claim.
I have so many untouched UHD videos on my NTFS hard drives that play without any issue it’s ridiculous.
(The only issue I ever see is audio drop out on certain known Disney movies, which ext4 has the same issue.)
Some people love Linux and hate Windows, maybe that’s your thing, I don’t know.
I don’t have a horse in this file system race and I have Linux machines and Windows machines (both have their pros and cons).
It would be really nice if we could help this guy fix his strange issue with out getting him side tracked.
You suggest he changes his file system (that can be a big to do). I suggest that’s not it and he should look into the other trouble shooting methods (GUI resolution, etc.).
Anyone who really cares to know the truth about NTFS v ext4 in real world video playback performance can read that thread and see that I’m not just some special guy with some special Vero that just magically likes NTFS.
Just two posts below the post you linked:
And I won’t mention how I lost 113GB of data because of ext4s delayed file allocation failed on me once and how I’ve never lost a single byte of data yet with NTFS.
We’ll see what ends up being the solution for the OPs problem, but I seriously doubt it has anything to do with the file system.
Peace!
PS: I’m open to testing any file you think wont playback over NTFS via USB2.0.
There’s been a lot of eager reception since Paragon recently announced plans to mainline their NTFS code in to the Linux kernel. Unfortunately their initial patch was just a single commit without any atomicity, so it isn’t suitable for inclusion and will likely need several revisions before it is eligible. Whether Paragon put in the effort to do this remains to be seen.
NTFS mounts could be performed by the kernel if we knew for a certainty that a user only needed an RO mount. The problem stems from the fact that writing doesn’t work with a kernel module, so we have to run everything through userspace.
You’ve clearly had some issues with ext4 and as such I can appreciate your reluctance to use it.
Some of the speeds you’ve reported in your previous posts don’t seem right, and it should still be the case that NTFS will hit CPU bottlenecks before it starts to saturate the maximum speeds of USB2.
NTFS is a robust filesystem. It’s used on almost all Windows PCs. But its Linux implementation leaves much to be desired. I am also unsure of how robust its fsck implementation is – but provided you have a Windows system at hand, that is a non-issue.
I haven’t personally tested, but a 400Mbps jellyfish clip would probably show a significant difference on ext4 and NTFS. The reason being that when the CPU is being taxed significantly, we’re losing time to serve the decoder ISR. If we start missing our timeslot, we aren’t pushing back frames to the VPU and render pathway in time, which will result in stutters.
I’d honestly suggest that the user tries playing back a file from eMMC or another drive formatted as ext4 or exFAT. Yes – migrating filesystems is indeed a big ask, which is why this should be verified by the user first. But I would not suggest this unless I genuinely believed that this could be an issue here. I don’t want to waste anyone’s time (mine or OP’s).
They each have their strengths.
We will launch a product next year which will be heavily dependent on Windows Server infrastructure – because we cannot stabilise the workload reliably under Linux (thanks to NVIDIA). I’m not an ext4 fanboy or a hater of NTFS, but it should be appreciated that its performance is degraded on Linux systems. Even if you disagree with the belief that there are 4K files that will not play well with an NTFS filesystem on the Vero 4K +, I think we could probably find common ground that an ext4 filesystem will outperform NTFS in terms of throughput.
In the longer term, two things will happen:
We will release more powerful Vero devices, so CPU usage will be more powerful and more tolerant of the load that FUSE based filesystems cause
NTFS (with write) support will likely be merged in the kernel one day, and this will mean that we no longer need to use FUSE at all.
We don’t have to agree on everything. There is no offence taken by your comments – we just take different views on the matter.
Strange, I’ve been using EXT4 for 10+ years, and have never lost data, except when I had a drive physically fail. And then I only lost about 50GB of the 4TB on the drive.