Hacker News Reader

Comment by klingonopera | Hacker News Reader

I was thinking of trying this out, but I imagine, with VDSL(2), there aren't any significant latency bonuses when doing it. There are way to few VDSL2 PCI-E cards available, if at all.

Edit: ...or does VDSL2 run with the ethernet protocol and just needs a standard nic? The VLAN ID tag makes me think so, but the GDMT993.5 (Vectoring) tells me it could be something more...

View on HN

VDSL needs its own NIC, as it operates very differently to ethernet. Recently I discovered that my telephone line was operating on one wire, but VDSL was still operating in a degraded fashion. Suffice to say this would totally prevent ethernet from working.

ADSL could only encapsulate ATM frames (hence using PPPoA), VDSL includes a new mode called PTM that can encapsulate ethernet frames. This may be what you were reading about.

I guess? Well, suffice to say, one does need special equipment to communicate over a VDSL line, it's not just applying a different protocol over a standard NIC.

The only PCI-E VDSL2 card I've seen is the Draytek VigorNIC 132. (Which, BTW, should really also have a low-profile backplate. :/ )

I can't deny I'd be curious to play around with it on an x86 OpenWRT system, but at about 200€, it's a bit pricy...

My current setup has a latency of 8ms, of which I'd speculate that at most 2ms are due to the router+modem? Could an x86 "monster" router+modem bring that down to 0ms?

Most of that latency would be VDSL, it was around 20ms RTT extra latency on ADSL though. The encoding has extra overhIf you really want 0ms latency, you'd want to look into fibre services.

You should also look into buffer bloat, and see if that's affecting your latency.

Reply on news.ycombinator.com