Zotac ZBOX CI323 nano

sleeprae

Observations/notes so far:

1. The Realtek NIC drivers are not stable on this system running 2.2.6-x64. I avoided 2.3 due to the reports of Atheros being more stable in 2.2.6, but it's possible that the NICs are better behaved in 2.3 or in 2.3.1. What would happen is reX watchdog timeout and a traffic hang a few seconds after pushing max throughput through the NIC. Forcing both NICs to 1000FD + flowcontrol + master (as seen here: https://forum.pfsense.org/index.php?topic=101587.0) seems to solve it, as does forcing 100FD.
2. The throughput tops out at about 600mbps in either direction. Could also be a Realtek driver issue. The CPU load is about 30%.
3. The Dell DW1707, in NGFF m.2 form factor using an E keying and a 1630 size, is detected by pfSense. It uses the Atheros AR9565 chip. I have not yet tried connecting a client to it in hostap mode, but it took the configuration w/o complaint.
4. The system varies between 9 and 10 watts at idle, headless, with PowerD enabled. Using a 4GB stick of Kingston 1600MHz DDR3L and a 40GB Intel 320 series SSD.

kss18

I just got the ci323 nano as well with 8GB, and pfSense 2.3 installed on it without a hitch, and is very stable.

The wireless chip it uses is AC3160 (which was not specified by Zotac). This chipset is not recognized by 10.3 freeBSD yet, and therefore, neither does pfSense.

My understanding is that the AC3160 will be supported via iwm in the 11.0 release, which is due out in September 2016.

Has anyone tried porting the driver? Any modules to share? ;)

I suspect I am not the first one to run into this issue…

sleeprae

@kss18:

I just got the ci323 nano as well with 8GB, and pfSense 2.3 installed on it without a hitch, and is very stable.

Are the NICs stable when pushing >500mbps though them for extended periods? I found they were fine at low throughput, but died when pushing them hard in 2.2.6 unless the negotiation was forced to 1000FD+flowcontrol+master. It would be great if 2.3 somehow resolved this issue.

Even if the AC3160 is supported in 11, it likely won't work as an access point. If that's important to you, you can replace the card as I did and use it now.

kss18

I am not pushing my router that hard! So, I have not experienced the issue personally.

SpazShark

Forgive me if this is a basic question…...

I'm considering getting the ZBOX CI323 nano, but I'm wondering if it's overkill for my situation....

I have a 320 mbps down / 20 mbps up connection from my current router with a 10 - 14 ms ping. It is a D-Link DIR-836L router with gigabit WAN/LAN ports that I recently bought for $28.99 CAD from Staples on clearance.

I have a layer 2 switch attached to the router (the only device connected to it) and Ubiquiti Access Points for wireless, so I'm assuming all LAN communication (media streaming from Plex) is done without involving the router (layer 3).

So other than the advanced features available with PfSense, will I see a raw performance increase in my WAN connection? Is cached data the secret here? Are there other metrics I'm missing?

movax

Power consumption is highjer than PCEngines APU2.

Zotac ZBOX CI323 nano at idle - 8.9W. (two Ethernet cables plugged in, no wifi enabled, display port output enabled, and keyboard connected).

PCEngines APU2C4 - 5.7W (but this is with only one Ethernet card active right now. Other ones not plugged in. No wifi cards. Just the sata express SSD)

hazza

Is anyone experiencing issues with PPPoE dying on these boxes?
Every so often, usually overnight, the PPPoE connection will drop and the box will continually attempt redialing, with no success. It's not until I halt the box and turn it back on (restarting doesn't resolve the issue) that it dials successfully (and on the very first dial after booting).

I've got a 40 Mbps down/2 Mbps up VDSL2 connection from my ISP. It's not the ISP or other hardware - while the Zotac is having problems dialing, I can disconnect it from the modem, plug my laptop into the modem, dial successfully from the laptop, reconnect Zotac to modem and it continues failing to dial.

VDSL modem is connected to re0.

This seems to happen every 3-5 days or so, at random.

I've attached log files from May 4th - it also occured last night, but the information in the log files is identical.

The first text file shows the log while the box is failing to dial - the 25 lines displaying keep repeating over and over, ad infinitum.
The second text file shows a successful dial immediately after booting back up (IP address etc censored).

So far I have tried disabled powerd, and for unrelated reasons I have changed the VDSL modem, but it still keeps displaying this fault.

20160504_pfsense_hazza_zotac-pppoe-failure.txt
20160504_pfsense_hazza_zotac-pppoe-success.txt

rajl

So I bought one of these a while ago, but waited until 2.3 came out before installing and putting this into my network to replace on old Netgear WNDR3700 (trusty little bugger that thing was). I upgraded to the latest BIOS prior to installing PFSense, which may or may not affect stability compared to others. I also don't push this box as hard as others. I have a connection that is 10/75. I don't have any stability problems, throughput is great and my OpenVPN connection for clients runs at line speed. CPU never seems to go above 25% and a lot of times powerd throttles the CPU down to 900Mhz from 1600 Mhz, so it burns less power. For prosumer use, it's really hard to beat this little machine.

interfasys

Regarding the realtek cards dying. when using netmap, you'll clearly see it in the graphs and logs when it happens. Throughput drops significantly and you get this in dmesg

625.301882 [2925] netmap_transmit           re0 full hwcur 187 hwtail 34 qlen 152 len 1146 m 0xfffff800570d4700
625.320610 [2925] netmap_transmit           re0 full hwcur 187 hwtail 34 qlen 152 len 70 m 0xfffff8006a0a1000
625.339337 [2925] netmap_transmit           re0 full hwcur 187 hwtail 34 qlen 152 len 1146 m 0xfffff8006c390c00
625.358056 [2925] netmap_transmit           re0 full hwcur 187 hwtail 34 qlen 152 len 1408 m 0xfffff8006a0d8800
625.376787 [2925] netmap_transmit           re0 full hwcur 187 hwtail 34 qlen 152 len 1146 m 0xfffff8006a129400
625.395498 [2925] netmap_transmit           re0 full hwcur 187 hwtail 34 qlen 152 len 1294 m 0xfffff8006a0d8800
625.421220 [2925] netmap_transmit           re0 full hwcur 34 hwtail 186 qlen 103 len 70 m 0xfffff80006f09200
625.439940 [2925] netmap_transmit           re0 full hwcur 34 hwtail 186 qlen 103 len 184 m 0xfffff8006a1e7d00
625.458646 [2925] netmap_transmit           re0 full hwcur 34 hwtail 186 qlen 103 len 1410 m 0xfffff80006fe0700
699.889875 [2925] netmap_transmit           re0 full hwcur 84 hwtail 193 qlen 146 len 1408 m 0xfffff8006c391b00
700.912739 [2925] netmap_transmit           re0 full hwcur 231 hwtail 163 qlen 67 len 622 m 0xfffff8006c0fc000
700.931456 [2925] netmap_transmit           re0 full hwcur 231 hwtail 163 qlen 67 len 1408 m 0xfffff800570ca600
700.950206 [2925] netmap_transmit           re0 full hwcur 231 hwtail 163 qlen 67 len 622 m 0xfffff8006c38fd00
700.968927 [2925] netmap_transmit           re0 full hwcur 231 hwtail 163 qlen 67 len 1408 m 0xfffff800570ca600
700.987657 [2925] netmap_transmit           re0 full hwcur 231 hwtail 163 qlen 67 len 622 m 0xfffff80006f01600
701.006385 [2925] netmap_transmit           re0 full hwcur 231 hwtail 163 qlen 67 len 1408 m 0xfffff800570ca600
701.025103 [2925] netmap_transmit           re0 full hwcur 231 hwtail 163 qlen 67 len 622 m 0xfffff8000680a300

Depending on your config, that's all that happens, but I've had some issues with some kernels and I think it's C-state related, so play with your PowerD settings or even disable deep sleep if your card stops responding.

To recover, use the attached keyboard (since remote access won't work any more)

# ifconfig re0 down
# ifconfig re0 up

If you look at the stats, you should see a lot of Rx missed frames

# sysctl dev.re.0.stats=1
dev.re.0.stats: -1 -> -1

# dmesg

...
re0 statistics:
Tx frames : 20412388
Rx frames : 13724308
Tx errors : 204
Rx errors : 0
Rx missed frames : 2545
Rx frame alignment errs : 0
Tx single collisions : 0
Tx multiple collisions : 0
Rx unicast frames : 13652465
Rx broadcast frames : 66333
Rx multicast frames : 5510
Tx aborts : 0
Tx underruns : 0

hazza

Thanks for the info interfasys!

I turned off hardware checksum offloading a few days ago, and haven't seen a dropout since then…famous last words, perhaps!

duren

It's too bad none of the new ones have dual ethernet other than the MC beast with the GTX.

http://www.tomshardware.com/news/zotac-zbox-mini-pc-gtx980-vr,31889.html

abc123

I just bought this box specifically to replace a EdgeRouter (ERPoE-5) which barely did anything…

So I have downloaded the AMD64 (64-bit) with USB Memstick and VGA as my options the SHA256 checksum was a match and I used PowerISO to create a bootable USB drive.

Only option changed in BIOs was to Boot Mode - Legacy,
8GB Ram Crucial
180GB Intel 520 SSD

Running Installer on a 64GB PYN USB 3.0 Drive
It boots up fine, installer runs:
[Enter] for multi user mode

I get a timeout here shown in IMG_3354.JPG

_to install
[Accept all Settings]
[Easy/Default]
Failure of command Here
Shown in IMG_3357.JPG
Details shown in IMG_3358.JPG

SOLVED: Took 180GB SSD and reformatted it to DOS, then re-inserted and ran installer. Worked perfectly as previously stated on pfSense 2.3

IMG_3354.JPG_thumb

IMG_3357.JPG_thumb

IMG_3358.JPG_thumb _

abc123

@movax:

Power consumption is highjer than PCEngines APU2.

Zotac ZBOX CI323 nano at idle - 8.9W. (two Ethernet cables plugged in, no wifi enabled, display port output enabled, and keyboard connected).

PCEngines APU2C4 - 5.7W (but this is with only one Ethernet card active right now. Other ones not plugged in. No wifi cards. Just the sata express SSD)

APU2C4 is a 1.0GHz quad core…not a good comparison

duren

@duren:

It's too bad none of the new ones have dual ethernet other than the MC beast with the GTX.

http://www.tomshardware.com/news/zotac-zbox-mini-pc-gtx980-vr,31889.html

Looks like the new 545 has dual ethernet, albeit still realtek :(

CI545 Nano Intel Core i5-6300U, dual core 2.4 GHz, up to 3.0 GHz

https://www.zotac.com/ca/product/mini_pcs/zbox-c-series/all

opticalc

can we assume that these realtek network throughput issues go away when running pfsense in esxi? as esxi is now handling the actual hardware, and pfsense sees a pair of e1000 NICs?

duren

@XFN:

Observations/notes so far:

1. The Realtek NIC drivers are not stable on this system running 2.2.6-x64. I avoided 2.3 due to the reports of Atheros being more stable in 2.2.6, but it's possible that the NICs are better behaved in 2.3 or in 2.3.1. What would happen is reX watchdog timeout and a traffic hang a few seconds after pushing max throughput through the NIC. Forcing both NICs to 1000FD + flowcontrol + master (as seen here: https://forum.pfsense.org/index.php?topic=101587.0) seems to solve it, as does forcing 100FD.
2. The throughput tops out at about 600mbps in either direction. Could also be a Realtek driver issue. The CPU load is about 30%.
3. The Dell DW1707, in NGFF m.2 form factor using an E keying and a 1630 size, is detected by pfSense. It uses the Atheros AR9565 chip. I have not yet tried connecting a client to it in hostap mode, but it took the configuration w/o complaint.
4. The system varies between 9 and 10 watts at idle, headless, with PowerD enabled. Using a 4GB stick of Kingston 1600MHz DDR3L and a 40GB Intel 320 series SSD.

Does the Dell card work?

messerchmidt

@abc123:

I just bought this box specifically to replace a EdgeRouter (ERPoE-5) which barely did anything…

So I have downloaded the AMD64 (64-bit) with USB Memstick and VGA as my options the SHA256 checksum was a match and I used PowerISO to create a bootable USB drive.

Only option changed in BIOs was to Boot Mode - Legacy,
8GB Ram Crucial
180GB Intel 520 SSD

Running Installer on a 64GB PYN USB 3.0 Drive
It boots up fine, installer runs:
[Enter] for multi user mode

I get a timeout here shown in IMG_3354.JPG

_to install
[Accept all Settings]
[Easy/Default]
Failure of command Here
Shown in IMG_3357.JPG
Details shown in IMG_3358.JPG

SOLVED: Took 180GB SSD and reformatted it to DOS, then re-inserted and ran installer. Worked perfectly as previously stated on pfSense 2.3

format it in a pc and then load pf sense and everything is fixed? getting the same errors._

duren

@messerchmidt:

format it in a pc and then load pf sense and everything is fixed? getting the same errors.

I used Rufus and never had any issues. https://rufus.akeo.ie/

speeddragon

This machine just arrived to install pfsense, after 2 years ago my alix2d13 died on me :(

I was wondering, it worth going to dual channel ? I was thinking in just put a dimm of 8GB instead of 2x4GB.

Thanks

kss18

Yes, the Dell DW1707 wireless card works (paid $5.00 on ebay). I have since got a second identical unit that is ready to be put into production.

I have also got this Zbox with pfsense 2.3.2 talking to a Tomato Shibby router via OpenVPN with shared key – both bridged and routed -- using the cookbook recipes. Thanks for the overall good documentation!

Hard to beat the specs on Zotac CI323 nano for the price, even after adding 8GB memory, new wi-fi card, and 120GB SSD. My only wish would be that it used Intel instead of Realtek eth.