Linksys Begins Shipping The WRT1900AC

[phoronix]

Phoronix: Linksys Begins Shipping The WRT1900AC

Belkin announced today they are starting to ship the WRT1900AC, their WRT54G open-source router successor that now supports 802.11ac and other modern features. Belkin is working on open-source OpenWRT support for this new router...

http://www.phoronix.com/vr.php?view=MTY2MDQ

[doom_Oo7]

128MB of flash memory

More than enough for systemd

[torsionbar28]

More than enough for systemd

Lol! But seriously, I wonder why they thought it needed 1.2 Ghz, 128 MB flash, and 256 MB ram? DD-WRT Mega, the largest one, runs marvelously on a WRT54GS with just ~200 Mhz, 8 MB of flash and 32 MB of RAM.

[zanny]

Lol! But seriously, I wonder why they thought it needed 1.2 Ghz, 128 MB flash, and 256 MB ram? DD-WRT Mega, the largest one, runs marvelously on a WRT54GS with just ~200 Mhz, 8 MB of flash and 32 MB of RAM.

The point is to make it overpowered so new use cases can be explored. It is a computer, after all. Also, most classic gigabit switch routers can't actually handle multi source traffic totaling a gigabit because their SoCs choke. So this thing probably is one of the few routers to be able to handle its Wifi and ethernet being saturated.

[Morpheus]

I'm still a bit pissed that Belkin pretends to use "WRT54G" fame to promote this router. The last revisions of the old boy were crippled with vxWorks and packed with too small memory to get real firmware loaded in. At least when I bought one in 2006 (a v5 one, bad luck), it was priced at 60? (I'm french).

The new kid is NOT open-source-friendly yet, nor it is cheap (an almost x5 high). Their PR manipulations sucks.

[torsionbar28]

I'm still a bit pissed that Belkin pretends to use "WRT54G" fame to promote this router. The last revisions of the old boy were crippled with vxWorks and packed with too small memory to get real firmware loaded in. At least when I bought one in 2006 (a v5 one, bad luck), it was priced at 60€ (I'm french).

At least here in the US, Linksys brought the original Linux based WRT54G back, with original large memory size, due to popular demand. They called it WRT54GL, I guess the "L" is for Linux. It is still sold brand new here (newegg.com) for ~$50:

http://www.newegg.com/Product/Product.aspx?Item=N82E16833124190

[droidhacker]

I'd like to point people to another option, its specifications aren't quite as impressive, but still *really* good, and a price that is at least affordable....

TP-Link Archer C7 v2

The v1 had a really crippling limitation that makes it basically useless for open source. Not TP-Link's fault though, rather qualcomm's.
Specifically, the difference is between a QCA9880-AR1A and a QCA9880-BR4A. The AR1A chip was put into production before anyone knew what the AC protocols would look like, so it lacks interoperability with a many AC devices, in particular, it hates Broadcom devices. Because it was such junk, Qualcomm never even bothered adding support for it into ath10k (and they aren't going to... ever). The BR4A, however, is totally supported by ath10k.

So, the Archer C7 v2;
The SoC is a QCA9558-AT4A, which is a 720 MHz MIPS74Kc.
... which isn't as awesome as a dual-A9 by any means, but isn't bad.

The 2.4 GHz 802.11N is on the SoC.

Second wifi on QCA9880-BR4A, 3-channel 802.11AC at 5 GHz.

Switch is an AR8327N-BL1A
This is some kind of magic smart-switch with hardware NAT. The hardware NAT functions aren't (yet) supported by OpenWRT, but if anyone is interested in it, I'm planning to dive into the code for it, could use some help.

NAT, believe it or not, is THE biggest drain on a CPU for this kind of device. The reason is that it has to modify the packets going through as fast as they flow. The SoC on this device is good for about 500 Mbps on software NAT (iptables).

16 MB Flash
128 MB RAM
2xUSB2.0

OpenWRT... **DONE**. Well, there is one trivial patch against barrier breaker needed for it waiting to be committed, but its pretty much flawless.

Best of all, its TP-Link.
Yeah yeah, you're all going to say something stupid like "chinese junk".
Nonsense. Chips on it are qualcomm and winbond.

I ordered mine from newegg.
I ordered a v2.
They sent me a v1.
I RMA'd it, specifically complaining about it being the wrong hardware version.
They sent me another v1.... after waiting 2 WEEKS and several screw ups regarding the advanced-RMA process.
I sent an email to TP-Link describing the problem with newegg, asking for advice on how to obtain a v2.
They replied with an RMA# and said just send them the v1 and they'll send out a v2, just that the v2's were out of stock, expected by the end of the week.
Noticing that their address is only a few minutes from where I live, I asked them if I could just do the exchange in person, they said no problem! We will email you when the new stock arrives and you can come by!

#1: Newegg screwed up. TP-Link was NOT obligated to deal with this.
#2: They're willing to hand-swap without messing around with shipping.
#3: Its practically immediate, no messing around with the internal 7-10 business days to process nonsense. Bad luck that they didn't have any on hand at TP-Link, but not a big deal.

These chinese companies are VERY VERY eager to prove themselves. Total pleasure to deal with them.

[zxy_thf]

The point is to make it overpowered so new use cases can be explored. It is a computer, after all. Also, most classic gigabit switch routers can't actually handle multi source traffic totaling a gigabit because their SoCs choke. So this thing probably is one of the few routers to be able to handle its Wifi and ethernet being saturated.

I'd like to see an atom-based router so I can install whatever I want.
An atom processor is not something too expensive for a ~300$ router.

[droidhacker]

I'd like to see an atom-based router so I can install whatever I want.
An atom processor is not something too expensive for a ~300$ router.

... can't see a reason why you would need x86 for the ability to "install whatever you want".

[droidhacker]

Switch is an AR8327N-BL1A
This is some kind of magic smart-switch with hardware NAT. The hardware NAT functions aren't (yet) supported by OpenWRT, but if anyone is interested in it, I'm planning to dive into the code for it, could use some help.

NAT, believe it or not, is THE biggest drain on a CPU for this kind of device. The reason is that it has to modify the packets going through as fast as they flow. The SoC on this device is good for about 500 Mbps on software NAT (iptables).

Just to reply to myself....

https://lafibre.info/images/doc/201106_spec_AR8327.pdf

... switch programming guide.