SABnzbd Forums

Hi,

Transferring from NZBget where I was receiving ~80MB/s pretty consistently.

I clicked the wrench and ran that test.

Used cache 0 B (0 articles)
System load 0.13 | 0.26 | 0.32 | V=118M R=70M
System Performance (Pystone) 167084 Intel(R) Celeron(R) J4125 CPU @ 2.00GHz SSSE3
Download folder speed 203.9 MB/s /Temp Download/incomplete
Complete folder speed 58.6 MB/s /media/Pending/Complete
Internet Bandwidth 15.2 MB/s 121.6 Mbps

I am 1gb internet and on the speed test running docker on the same machine (syno DS920) I am getting ~930mbps. I can't post a screenshot.

Any advice?

Thanks

Do an actual download with the refence download:

In SABnzbd's upper right corner, click on the wrench symbol ("Status and interface options"), then click on first tab Status, click Test download 1GB. After downloading, in history, at the download in the drop down, what was the resulting speed? Example "Download Downloaded in 55 seconds at an average of 22.9 MB/s"
Same for the 10GB test download.

Thanks for your help here are the results:

1GB:
Downloaded in 32 seconds at an average of 32.3 MB/s

10GB:
Downloaded in 5 mins 35 seconds at an average of 31.5 MB/s

2 other notes:
1. I did change the permanent location to the SSD for testing and which didn't affect the results. I do understand that the slower drives will be a bottleneck at some point but at least my understanding is it shouldn't affect the download and rather the transfer.
2. I saw that the test downloads actually come from my providers. So I changed priorities and still get the same speeds. I have played with the connection values as well, which has not changed anything.

Just thought of something after reading some more posts on here.

Does sab process the downloads (unpackage) in the temp download folder then move to completed the same as Nzbget? Or does it move to complete and then do the processing? This won't affect my current issue as it's the download speed that seems to be the bottleneck at the moment. But as that increases I want to make sure I'm using the SSD to fully process the files prior to moving to the larger slower drives.

Does sab process the downloads (unpackage) in the temp download folder then move to completed the same as Nzbget? Or does it move to complete and then do the processing?

None of the above. You can see it in your sabnzbd.log:
... with
Incomplete on / temp on /home/sander/.sabnzbd/Downloads/incompete/
Complete on /media/sander/eentering/

So it happens in the unrar step

okay makes sense thanks.,

Any thoughts on the test results I provided? This is strange.

Upgraded to 3.8 dev and still have the same results.

After a download, go to Wrench -Status, and see if there is a line "Download speed limited by"

If so ... what does it say?

Download speed limited by Disk speed (6x)

However.... when I changed to the SSD I think it said the same thing. I'll QC right now but that would put both over 150MB/s

Re-Ran under the SSD for both not just temp:

1GB test::
DownloadDownloaded in 30 seconds at an average of 35.3 MB/s
Age: 197d


Download speed limited by Disk speed (6x)
System Performance (Pystone) 167000 Intel(R) Celeron(R) J4125 CPU @ 2.00GHz SSSE3
Download folder speed 213.7 MB/s /Temp Download/incomplete
Complete folder speed 194.6 MB/s /Temp Download
Internet Bandwidth 35.5 MB/s 284 Mbps

> Download speed limited by Disk speed (6x)

So ... that is what SAB experiences and tells. The "6x" means "6 times" (not "times 6"). So not too much ... just a few times.

And your disks speeds look fast enough.

So no clear reason nor answer from me.

Appreciate your assistance. CPU over 100,000 is what I saw somewhere is that correct?

I'll post back if I find anything. Appreciate your assistance Sander.

I'm impressed by that 167000 on a Celeron!

My Celeron:
System Performance (Pystone) 93552 Intel(R) Celeron(R) J4105 CPU @ 1.50GHz SSSE3

But my system runs Python 3.10. If you run Python 3.11: that's much faster, so higher pystone

Interesting. Yep I'm on 3.11. It must be that as .50GHz difference between our processors probably doesn't do that much

Is running this via docker taking too much overhead? Should I try and install the syno package? I'd prefer to run it on docker if it's the same

Without docker is faster.