Is it worth switching to the LibTorrent 2.0 version of qBittorrent?

[dftf-stu]

Hi all

I'm using the LibTorrent 1.2 version of qBittorrent 5.0.5.

I note that even-now, the latest 5.1.2 release still-offers both LibTorrent 1.2 and 2.0 builds separately, which would suggest there may still be outstanding-issues with the 2.0 version, otherwise the project would have moved-over to only supporting just 2.0 (like how Qt5 was dropped after the 4.6.7 release).

So, I was wondering: what outstanding issues are there still with the LibTorrent 2.0 version, and what benefits does it actually offer over 1.2? For example, are there certain torrents, 🧲 or trackers that I may not be able to connect-with by using the 1.2 version? 🤔

Thanks

[xavier2k6]

@dftf-stu Please read #19653, need any further info or have further questions then let me know.

[dftf-stu]

@xavier2k6: thanks for the link to that ticket.

I can't say it really answers my question through for three reasons:

1. The marked answer is from 26 Sep 2023, so within just-over a month will be two years old. Is the issue that person describes still an issue with the 2.0 LibTorrent as of now or not?
2. There is a comment there about adding info to the FAQ. I've just done Ctrl+F on the FAQ and cannot find any matches for "lt12" or "lt20", and only one reference to "LibTorrent", which is just where it mentions that is the library qBittorrent uses. Within the ticket itself, advantages of LibTorrent 2.0 are given-as "I2P; 256 MiB piece-sizes; support for BitTorrent v2" -- but what do those mean? Are some torrents only in v2 format, for example, and so by-using the 1.2 version I cannot use them? Adding this info into the FAQ, as was suggested in that ticket, would be really helpful.
3. When I've done a web-search and asked AI, the general answer is "LibTorrent 2.0 uses more RAM, and some users report lower disk performance, especially on external drives". Okay, so how much more RAM, and is that still a current issue? And if the external-drives are small, flash-based "thumb-drives", they usually lack a DRAM cache, so one would expect poor-performance when doing lots of small writes. But as I only save to an internal SSD, this is irrelevant for my use-case anyway.

I also note just two-weeks ago, another user has asked something similar to my question here:

Since this was posted years ago - Did anything change regarding issues with libtorrent 2.0.11?
Is libtorrent 2.0.11 better now than libtorrent 1.2.20?

It would be really good if there was a ticket or something here that summarised what the current issues or limitations are in the 2.0 LibTorrent and marked which ones have been resolved as of which qBittorrent version.

[qBittUser]

Is the issue that person describes still an issue with the 2.0 LibTorrent as of now or not?

Depending on your OS, hardware, network speed and qBittorrent settings. By default libtorrent v2 is slower, uses more resources or have other anomalies, but some users won't notice any changes. In latest qBittorrent advanced settings you can specify to use different Disk IO.

No clue why default IO hasn't been changed, especially if user who added the new option has suggested users to try using it and if it has helped others with resolving various issues.

I did create a discussion Q&A, but no replies. Maybe others prefer to wait default to be changed in libtorrent or all old issues to be fixed first, instead of a workaround. #22836

cannot find any matches for "lt12" or "lt20", and only one reference to "LibTorrent".

Well unfortunately there's a lot of suggestions to improve Wiki unresolved if look at open issue reports with documentation label, some issues still open for several years and too few volunteers.

To me it seems that quite often many new users who make a GitHub account to post issues skip reading Wiki. So most likely many don't notice that if you open official download page and click on (Additional download options), then there's written which libtorrent and other things each installer is using, long ago even libtorrent 2 was default and then mentioning RC_2_0, RC_1_2, lt12 was confusing, yet after using qBittorrent for long time, always reading news page before trying new version and reading elsewhere it became logical what different installers mean. Even opening news page has Library versions to click to see what qBittorrent is using.

Many other torrenting clients websites don't even bother providing multiple installers or don't mention anything what exact dependencies or libraries where used.

It would be really good if there was a ticket or something here that summarised what the current issues or limitations are in the 2.0 LibTorrent.

That would be a very long list and you might not even notice or encounter them. Currently libtorrent v1.2, v2.0 and v2.1 series isn't perfect all have various issues and differences, some old issues in libtorrent and qbittorrent GitHub page are even prematurely marked as closed.

If you open qBittorrent issue reports, select to filter by libtorrent label, then you see 71 open issues.

If you select Waiting upstream, then there's 32 open that might be partially related to libtorrent, but not all currently have included libtorrent label.

Okay, so how much more RAM, and is that still a current issue?

You have to try yourself, there's not enough thorough easily reproducible tests, real everyday usage, hardware and settings varies between different users and some even have assumed libtorrent v2 uses less.

There's Swap, Working set, Page file or Virtual RAM, OS cache, compressed cache, not just more commonly known noticable usage of physical RAM or process usage of qBittorrent.

If using Windows OS, then it's more convenient to download Process Explorer to notice more info, some which is hidden by default in Task Manager when you wish to compare various resource usage between different libtorrent versions.

By default libtorrent v2.0 with qBit will give your OS more control how to use available RAM and in what order to read, write and keep caching things, one workaround to get more efficient is to change Disk IO to Simple, which I mentioned before.

There's different settings if using libtorrent 1.2 and it can be more predictable, efficient, clears cache more often but certain things might be slower, all depends on your torrenting usage and sometimes if various qBittorrent settings are adjusted to your preferred usage to get more what you expect and want.

Users do have asked before what's best, but unfortunately there's too many variables.

[dftf-stu]

@qBittUser: apologies, but you're talking way-too complicated when as an end-user, I'm just looking for simple, shorter answers.

Let's summarise what I'd want to know in a few points:

• If I use the LibTorrent 2.0 builds, does that still allow me to connect to older clients (i.e. seeds) and trackers who use Lt1.2?
• If I stick-with the LibTorrent 1.2 builds, are there some "v2 only" clients (i.e seeds) and trackers I won't be able to connect to?

[In short, for the above two points: if I want the best-chance of actually obtaining a torrent, particularly rare or obscure ones with few seeds, will it be best to use Lt1.2 or Lt2.0?]

• In the current 1.2 builds, how-does the "Require Encryption" BitTorrent setting compare to I2P?

• Is there a ticket on here that lists what major-issues are currently preventing LibTorrent 2.0 from becoming the shipping-default?

[qBittUser]

I'm just looking for simple, shorter answers.

Unfortunately I can't say one libtorrent is definitely always better than the other.

Torrenting is complex, most of your follow-up questions have answers if search online.

If I use the LibTorrent 2.0 builds, does that still allow me to connect to older clients (i.e. seeds) and trackers who use Lt1.2?

Yes.

If I stick-with the LibTorrent 1.2 builds, are there some "v2 only" clients (i.e seeds) and trackers I won't be able to connect to?

Yes, qBittorrent with libtorrent v1.2 doesn't properly support connecting with i2p trackers and peers that have forced their torrenting clients to support only i2p.

If you can add torrent file, then most likely you added either compatible magnet link, v1 or hybrid torrent file. If it's fully v2 only torrent file you added, then you can't add it to torrenting client that's using libtorrent v1.2.

Many public torrent sites still offer only v1 torrents and not all torrenting clients support v2 only torrents or i2p.

In the current 1.2 builds, how-does the "Require Encryption" BitTorrent setting compare to I2P?

I2P gives a lot more privacy if set up correctly and some torrenting still don't support i2p and don't even have encryption enabled by default.

Encryption means traffic gets encrypted, but your IP and with who or what you connect to can still be seen and traced back.

If you force only specific connections or only use certain few trackers, then you might not be able to connect with others.

Google Search gave this link which has good explanation about encryption:

https://askubuntu.com/questions/1257636/what-is-qbittorrent-encryption-mode

Couldn't find good link about i2p without any confusing info.

Is there a ticket on here that lists what major-issues are currently preventing LibTorrent 2.0 from becoming the shipping-default?

No there's not, unless I missed it.

Maybe waiting for libtorrent issues related to cache to be resolved: https://github.com/arvidn/libtorrent/issues?q=is%3Aissue%20state%3Aopen%20cache

If want less issues with libtorrent v2 build, then I did give link to my discussion and you can try: Open qBit>Tools>Options>Advanced, then find Disk IO and change it to Simple.

Wether it's enough to resolve most old issues to make v2 default again I have no clue. Maybe have to wait for libtorrent v2.1 to be released instead of qBittorrent being offered only with v2.0 to everyone.

particularly rare or obscure ones with few seeds, will it be best to use Lt1.2 or Lt2.0?

Then use libtorrent v2 and try luck with VPN, port forwarding, add more working trackers and I2P trackers and enable support for i2p.

There's no very easy to follow guide to enable i2p and too few clients support it, afterwards you need patience and luck. Long ago I tried getting it enabled, but couldn't figure it out.

There's always a chance another torrent shares exactly the same file that might help to get another transfer completed.

[dftf-stu]

@qBittUser: thanks for the reply.

I assumed "I2P" was something built-into BitTorrent v2, but yes, seems it's an additional thing that has to be setup.

So for everyday users, the stuff you get "right-out-the-box" with v2 are the SHA-256 hashes, so less-chance of getting the wrong data than SHA-1; and up-to 256 MiB piece-sizes, the benefit-of-which is unclear to me, outside of possible reduced CPU and SSD use by limiting the number of overall operations. Beyond that though, I'm not really sure why receiving a 1GiB file in say 4 x 256 MiB chunks would be better-than 2048 x 512KB chunks? 🤷🏻‍♂️

Anyway, for-now seems the vast majority of users just go for the LibTorrent 1.2 builds still:

And for two-other Torrent apps I use:

• "Flud" (an Android client) doesn't support BitTorrent v2 -- I tried the "v2 only" test .torrent and 🧲 links here, and both fail
• The "Transmission" team aim to add v2 by their 5.0 release; but 4.0 isn't even at stable yet, so 5.0 will be around 3-5 years away

So seems for a good number of years yet, may as well just stick-with the Lib1.2 builds, as seems very-few are using Lib2.0 yet.

[Obegg]

Tested latest qBit with latest lt20, even as suggested in this discussion to change disk I/O to Simple pread/pwrite, I found issues when downloading at high speeds: the HDD is at 100% usage, even if I increase the disk queue size to max the HDD is simply too slow to write. For now I'll stick to lt12.

[Obegg]

So... I was trying to reproduce to issue with the high disk usage, because when I wrote my first reply I just noticed there is high download speed and it would jump back to 0 mbps, then when I opened the task manager I saw 100% disk usage, then I opened the stats in qbit and saw the total buffer size climbing to 2GiB (max), that's when I wrote my reply and switched back to lt12.
Now when I try to reproduce the issue (with lt20) with the high disk usage I can't seem to succeed, maybe the issue can only be reproduced when the torrent had seed and peers?

here's my current tests from just now:

Installed latest qBit and latest lt20
Task manager and qBit stats open at all times
New testing - same settings both tests, same torrent both tests (only seeds), all that's changed is disk I/O

Disk I/O: Default
300 Mbps = No Issues
500 Mbps = No Issues
HDD reached 100% usage for a few seconds but had good write and read speeds and then went down from 100% usage.

Disk I/O: Simple + Restart qBit
80 Mbps = No Issues
300 Mbps = No Issues
HDD is reaching 30%, 40%

maybe I should use a torrent with both seeds and peers?

[qBittUser]

I quickly read through these 2:

• Add "Simple pread/pwrite" disk IO type #21300

• Libtorrent 2.x memory-mapped files and RAM usage arvidn/libtorrent#6667

Not much mention about writing speed, mostly about reading speed and efficiency or memory usage.

[Obegg]

[Obegg]

Those screenshots I just posted? Those are with Memory mapped files, but guess what... even with Simple pread/pwrite (yes, I restart qbit after changing the setting) I have the same issue, high disk usage, high buffer size.
Damn, I really wanted to use lt20, I even gave it another chance today just to make sure I really did test it properly.
Back to lt12 we go...

[qBittUser]

You should share what other advanced settings you had used with your tests, maybe some defaults or your custom settings need changing.

Screenshots don't show how much Task Manager or Process Explorer shows your qBittorrent used and uses Working Set memory.

One good libtorrent and drive efficiency test with Windows OS would be with SSD or RAM cached drive and then transferring from another fast computer while keeping open Process Explorer or something else to make sure which libtorrent and settings are faster (finishes torrent download) and which spends or wastes more drive read and writes. Yes, it might seem bit unrealistic, but that's more stable way to verify differences.

[Obegg]

Wait... Correct me if I'm wrong here, but here's my interpretation of what libtorrent 2.0.11 actually does compared to libtorrent 1.2.20:

libtorrent 2.0.11:
If you set Disk IO Type to Memory mapped files - Whenever you download new pieces from seeds/peers it will write that new piece to disk but also read that new piece from the disk (which causes the high disk usage issue everyone complains about) and send it to your RAM (control the RAM usage by changing the value of Physical memory (RAM) usage limit), so whenever a different peer needs a piece from you it won't create another read from your disk, instead it will read from your RAM.
This kind of behavior results in better upload speed (because it's directly from your RAM) and lower wear to your disk (hard drive) and also makes rechecking a lot faster (because it's directly from your RAM).
Question: By writing to disk and then reading from disk (in order to send to your RAM) - How can we compare the level of wear done to the disk? Like.. Is it better to write to disk and read from disk one time or does write to disk one time and random reads whenever a peer needs a piece? Which one wears out the disk more? I guess it depends on the usage, and the amount of RAM, right?

libtorrent 1.2.20:
Every time a peer requests a piece from you it will create a read job from the disk which will increase wear to your disk but also harm your upload speed.

[soredake]

Good news arvidn/libtorrent@fb4a525

[master255]

With the help of AI and the best torrent programmers, we’ve come to the conclusion that V1 is better than V2.

But the heavy train of V2 development can’t be stopped by logic. They understand this, but they continue developing V2 anyway. Without explaining anything to anyone.

Maybe I can understand them. It’s something they love to do. Even if it’s not the right thing to do.

In any case, as of today, with the help of AI, anyone can apply all the improvements to V1 and get a super torrent library. Which is lightning fast!

[qBittUser]

V1 is better than V2.

I might be dumb and blind, but currently seems to me that only you with your own AI (chatGPT) discussion example came to that conclusion and that was more about using v2 vs v1 torrent files not much info about anything else.

When I asked different simple separate 5 questions then chatGPT clearly wrote me that using v2 torrent files is better. arvidn/libtorrent#8366 (comment)

So it's misleading to write "we've" or "best programmers" if you're the only one and others just haven't bothered to spend time to confirm or deny everything. So just write "I and chatGPT".

they continue developing V2 anyway.

Then there's a chance it will become noticably better, get more features and be more efficient in some areas.

Just tweak qBit advanced options, disable i2p and make a qBit build without v2 torrent file and v2 protocol support if you don't like them, then less time to spend modifying older libtorrent version.

Although I think there has been mentioned by others about the inefficient 16 kb Disk IO blocks and higher memory usage, worse read/write speed, so maybe some issues related to them will probably stay with libtorrent v2.

Unfortunately there's at least one user who currently wishes libtorrent v2.1.0 to be used by default for everyone, as if doesn't care, is unaware or has forgotten the fiasco when libtorrent v2 was set by default without it being previously properly tested.

• Switch the first version in downloads page to be lt21/lt20 #24396

If look at qBit and libtorrent issue reports, then it's clearly visible that libtorrent v2 still needs work and qBittorrent developers and contributors are aware of libtorrent v2 related issues, but no one so far hasn't bothered to make a consolidated list.

with the help of AI, anyone can apply all the improvements to V1

You've created working libtorrent fork?

Will you document changes you've made and make it public for others to try?

Will you be planning to create PR's to backport already merged Libtorrent v2.0/v2.1 improvements into libtorrent RC_1_2 branch?

If backporting and using AI is really easy for you, then I hope this one important thing can be solved first, so others could eventually get official libtorrent v1.3 series that's compatible with newer Boost version:

• WIP - boost-1.87.0 api changes for RC_1_2 into RC_1_3 arvidn/libtorrent#8027

Also good if someday someone solves the 256mb piece size crash issue that happens with libtorrent v1.2:

• Crash when checking torrent with piece size of 256.0 MiB #21011

[master255]

When I asked different simple separate 5 questions then chatGPT clearly wrote me that using v2 torrent files is better.

You've already lose in your own thread. If you want to prove something and have arguments to back it up, post them there. Stop following me everywhere!

[qBittUser]

Title of the discussion here is about to know about reasons to pick or to avoid v2 libtorrent not about v2 torrent files or v2 protocol.

That's why I wrote that in your discussion (link you shared here) seems to be more about related to usage of v2 files.

You've already lose in your own thread.

You probably mean "lost" not "lose".

I already explained there. My purpose there wasn't to win at anything.

I just:

1. Showed that chatGPT AI should be teached or avoided if it is sometimes misleading or lying.
2. To understand if you where using right terminology.
3. See your discussion purpose and real intentions.
4. If you can be bothered to backup almost all your related and unrelated claims.
5. To see if anyone else would someday care to try to explain and confirm various things about v2 torrent protocol and v2 torrent files.

With the help of AI and the best torrent programmers

Since very few bothered to write anything back there, then it's inappropriate to write "programmers" if it's only you "a programmer".

post them there

Then please try not to write typos, promote AI or claim new things here if you can't be bothered to backup.

have arguments to back it up

You wrote here previously yourself:

"with the help of AI, anyone can apply all the improvements to V1"

I'll try my luck again:

Can you give examples for beginners or at least specify which AI is good to achieve that?

Will you be planning to create public PR's to backport libtorrent v2 improvements into v1?

Do You have public GitHub repository that others could see what you mean about "all the improvements"?

Is there a page where others can help you and everyone else out to test and improve things?

Can your programming skills and somekind of AI help to fix those 2 things I mentioned in my previous comment?

Sharing again:

• WIP - boost-1.87.0 api changes for RC_1_2 into RC_1_3 arvidn/libtorrent#8027
• Crash when checking torrent with piece size of 256.0 MiB #21011

Afterwards would be nice to get i2p and qBit work with libtorrent v1.x+ as good as with libtorrent v2.x+, then there's more reasons for more users to not prefer libtorrent v2+.

Do you think it's not worth it to try or it's too difficult?

Which is lightning fast!

Can you create a separate discussion somewhere with little bit detailed info and examples how and what exactly you used and tested so far?

(I already wrote that I and others are clearly aware that libtorrent v2 still needs more testing and improvements. So no need to argue about it. Just would be nice to have something new that others could be able to reproduce to confirm your findings and see later on what and how much official v1 and v2 libtorrent still needs improvement compared to your own or with help with your AI applied improvements.)

[qBittUser]

There's a plan to have qBit v5.3+ use libtorrent v2.1+ as default and eventually discontinue (drop) support for older libtorrent:

• Support of libtorrent < 2.1 EOL #24689

Previously someone else also suggested it, but no evidence it's good for everyone:

• Switch the first version in downloads page to be lt21/lt20 #24396

So I wonder if someone is already planning to thoroughly test things and if there's issues, then publicly provide unofficial qBittorrent with libtorrent v1 support?