NNTP pipelining explained
Short answer
With NNTP pipelining, your downloader sends the next article requests before the earlier replies have arrived, so the connection doesn't sit idle between articles. It helps most on long routes, such as Australia to a US or Dutch server. It adds no retention, fixes no missing articles, and can't beat a plan's speed cap, such as XLned's 120 Mbit.
With pipelining, your Usenet client asks for several articles in a row over one connection, rather than asking for one, waiting for it and only then asking for the next. On a long route, all that waiting is what slows you down, and pipelining cuts most of it out. It affects how quickly the data arrives, but not which articles are on the server.
What goes over the connection
A classic NNTP session works one step at a time. Your client requests a single article by its Message-ID and the server sends it back, and only then does the client send its next request. So between the end of one article and the start of the next, the connection does nothing for one full round trip.
With pipelining, the client fires off a few requests back to back. The server answers them in order, so the next article is already travelling while the client is still handling the previous one. The client does have to keep track of which reply goes with which request, which is why not every client supports it and why the number queued, the depth, is kept small.
The arithmetic, no benchmarks needed
Picture one connection, and call the round trip to your server R. Without pipelining, each article costs its transfer time plus R of waiting. An NZB usually has thousands of articles, so on every connection you pay R thousands of times over.
Queue three requests at once and the waiting gap only comes up about once every three articles. On a short route R is tiny and you'll hardly notice. On a long route R is big, and the time saved can be a large part of the total. That's why people in Australia or New Zealand care about pipelining far more than someone living next to an Amsterdam data centre.
Pipelining or more connections?
Extra connections run side by side, while pipelining keeps each single connection busy. Both go after the same idle time, which is why adding connections often does the trick when a client doesn't pipeline. Where they differ is the limit: every provider caps connections, and some plans cap speed too. The connections guide explains how to find the point where adding more stops helping.
Here are the limits in our provider data, checked 2026-10-23. No client setting will get you past them.
| Provider and plan | Max connections | Speed cap |
|---|---|---|
| Frugal Usenet | 250 | None listed |
| Newshosting | 100 | None listed |
| Eweka | 50 | None listed |
| ViperNews VIPERUNL | 40 | None listed |
| UsenetBucket Ultimate | 25 | 400 Mbit |
| Pure Usenet | 20 | None listed |
| XLned XL Gold | 12 | 120 Mbit |
Pipelining counts for more on a plan with a low connection limit, such as XLned with 12, because you can't make up for latency by opening extra connections. On a plan with a speed cap, neither approach gets you beyond the cap.
SABnzbd and NZBGet settings
Most people never need to touch this. Clients that support pipelining come with a sensible default, and some show the depth as an advanced setting. If you do experiment, change one thing at a time: the connection count first, then the pipelining depth, comparing complete downloads of the same size each time. If the speed doesn't move, put the setting back. For the wider differences between the two clients, see the SABnzbd vs NZBGet comparison.
When something else is the bottleneck
- The CPU. A small board can run out of processing power for SSL and yEnc decoding before the network is maxed out. The client's status page shows a busy CPU while the speed stays flat.
- The disk. Unpacking and repairing write a lot of data, so a slow USB disk or a busy NAS holds up the queue once the download itself has finished.
- Wi-Fi. Run one test over a cable. If the speed jumps, your home network was the limit.
- Missing articles. Pipelining fetches the articles that exist more quickly, but it can't fetch ones the server no longer has.
FAQ
- Do I need to switch pipelining on?
- Usually not. Keep the client's default and tune your connections first. Only test the depth if you're on a long route and still below your line speed.
- Does SSL make pipelining useless?
- No. SSL adds encryption work for your CPU, but the round-trip delay that pipelining hides is still there.
- Will pipelining fix incomplete downloads?
- No. An incomplete job means your server is missing articles, and that calls for a provider on a different backbone or a newer NZB.