Asrock J3455B ITX
Board = 1 PCIe 2.0 x16 (x2 mode)
INTEL ETHERNET i340-T4
PCIe Card = 1 PCIe 2.0 x4

So it should be able to stitch this card directly into the PCIe x16 slot for sure, without any hassle or
problems and it should be also be able to place it into that case as well too! So there is no need for a riser cable.

I need information what is the best way to fit the Intel card on the motherboard. I read somewhere in this forum that it involves cutting either the motherboard or the nic.

If the card is greater then the PCIe slot, it could be a work around, but in your case the slot is greater then
the PCIe card and this might be then no problem, insert the card into the slot and all will be fine for you.
(Take a look on the attached picture)

What is the best way to do it? The motherboard is brand new while the nic is used.

Board into the case and card into the slot, that´s the way to go with.

I am also open to suggestion on using a riser card if it fits on Corsair 250D but I have no idea what to buy.

Why do you think there will be a riser card or cable needed? Perhaps this card comes with a low profile bracket
or it is not matching to the slots of the case then for sure you will bee able to went to amazon.com and get the
right working tolls to walk around any problems. Here are shown some of them right to buy or go with.

maxresdefault.jpg
maxresdefault.jpg_thumb

The guide you were following misses one point and that is correct filtering.

Usually you filter on each interface individually.
There's an advanced option at System Tunables where you can set pfSense to filter on the bridge instead.
net.link.bridge.pfil_member Set to 0 to disable filtering on the incoming and outgoing member interfaces. | default (1)
net.link.bridge.pfil_bridge Set to 1 to enable filtering on the bridge interface | default (0)

Better than using an interface group and way easier to understand when doing changes in 7 months or so.

Frank,
thank you very much for your very detailed explanation!
I checked it and pins and springs are OK.
There are only 2 Pins for the cooler and you can move it so that the springs are become more tention.
So it seems that it works as it should work and that there is some lash for the cooler.
Temp is about 50° in my supermicro 505 rackmount case without active cooling in a full network rack.

I took a better look at the standard shield from the case. and yes, it works if you open every inlay and leave it closed for displayport/hdmi.

Hi

thanks for your input.

I have now ordered following stuff:

APU2C4 Router/Firewall in red chassis 60GB disk USB serial cable AC adapter

Shipment is on its way. Looking forward to start learning, I expect to be a board member with a lot of questions to you guys!

@BlueKobold:

And the IPMI ports let us say from the Supermicro Intel Atom C2000, C3000 or Xeon
D-15xx boards are not nice too?

If you don't need IPMI, it's just a waste of electricity. In a low power board the OOB management computer is a significant fraction of the overall power consumption.

@kklouzal:

Fiber is a completely different story since FTTP (Fiber To The Prem) brings a standard fiber line into your home in which you can simply plug into an existing fiber channel card, get the PPPoE instructions from your ISP and away you go.  8)

FTTN (Fiber To The Node) typically uses copper for the 'Last Mile' going from the distribution box. That last mile is either Coaxial or RJ11. In the case of RJ11 you can still use a PPPoE setup but you're already using RJ11 and that's going to be horrible in terms of speed and latency especially the further you get from the DSLAM.  >:(

In a FTTN situation where Coax is used for the last mile you typically have an extremely short distance of copper (in terms of the entire connection length). The consumer has no control of anything past the drop towards their ISP. Basically you have a direct fiber connection except for this 'Last Mile'. Why not directly convert back to fiber, or as close to it as possible once it passes the drop back into our home/office?

The Chelsio line of ASIC provides EXCELLENT latency, sub millisecond down to around 25 microseconds. The optimal solution would be to bridge the Ethernet PHY to the Ethernet PHY of a 'modem on chip' ASIC or something a little deeper and more integrated if you're not using a modem on chip solution. I guarantee you would receive at minimum an order of magnitude latency reduction as well as maxing out the line rate supplied from the your purchased plan.  8) :-*

https://www.chelsio.com/terminator-6-asic/

In the crude design above we have our Chelsio T6 ASIC interfacing with the 'Modem On Chip ASIC' via one of its two provided ethernet interfaces. The secondary free interface is not wasted and we provide a SFP+ port to be made available for the user to do with as they see fit. The T6 ASIC then interfaces with the host machine and presents itself as two available NIC, one dedicated to WAN which has been coupled with the modem on chip. Throw a huge heatsync over the entire thing with a dual 40mm fans and away we go!

Again, I think a card of this particular type design, for at least a first wave of exploring the real user demand for something like this, would be extremely attractive.

This doesn't actually change anything compared to a bridge mode cable modem. In fact, it pretty much is a cable modem. That is just the issue here: all of the proposed "solutions" are just normal modems crammed on a PCI(e) card with a ethernet interface to the host. It doesn't change a thing, except physical form factor.

Just to be safe, download a config.xml from the WebUI. If the device crashes completely or the FS becomes corrupted, it's an easy way to get back to work. Something else: when reinstalling pfSense, some hardware requires a specific image to get going. You probably already used the right image: pfSense-memstick-serial-VERSION. If you didn't get a console after boot, you need to check the BIOS as well. pfSense just uses whatever the hardware has, but if it isn't configured properly it will just die as soon as it switches to the right console.

If you have spare time, I'd suggest (after backup) trying to install or at least live-boot pfSense, just to make sure your hardware isn't about to die. The APU's usually last a pretty long time.

@Gektor:

Hi, i need to add patch to support MSI interrupt in Atheros WiFi drivers, but i don't know how a can do it, please help me.
Patch is here:
https://patchwork.kernel.org/patch/9999249/
pfSense 2.4.1

This is a Linux kernel patch, not a FreeBSD kernel patch. This will require a C/C++ software engineer with FreeBSD kernel experience to port and make functional. Nobody in pfSense will fix this for you (and definitely not for free). That patch for Linux was done by Qualcomm, maybe you should ask them to also make a patch for FreeBSD.

If you want to have more help with this, ask FreeBSD, not pfSense. Once it is in FreeBSD, it will also be in pfSense.

@wzochert:

DeLorean, I've seen you mention that you've used the QC5000M in other threads.  Have you upgraded one to 2.4 yet?  It seems that would be a pretty simple and cheap way to solve the AES-NI problem with 2.5 when it comes out.

I have build several firewalls with the Asrock QC5000M in a Supermicro 1U chassis, and these work also pretty fine without any problems,
very silent, low power draw (arround 17Watt idle @ 220V) and…......AES support :-)

Grtz
DeLorean

While in Legacy mode i got

[ 760 ] generic_netmap_dtor            Restored native NA 0

and finally i got some alerts on suricata  :)

This is indeed a driver and/or firmware issue with this chip and revision. If updating does not work your next best option might be a dual-port mini-PCIe NIC…

I tried downgrading to 2.4 and 2.3.4 with no success.

The switch is a Cisco SG350XG-24F, so it has only 10Gb ports. When then interface finally goes UP, it synchronizes as 10Gb with no problem.

I cannot try immediately your suggestions about mbuf and num.queue, as the device has to go back to production (I will use RJ45 links instead at the moment), but if I can I will of course try this !

from the Connecting to the serial console
The default serial console settings in pfSense 2.2 and later are 115200/8/N/1, meaning:

Speed: 115200 Data Bits: 8 Parity Bits: None Stop Bits: 1

Must be set into the Putty or TerraTerm or any other terminal program too! In pfSense is it the default setting.

I read through that a couple of times, and my understanding is that, while it does not prioritize VoIP traffic, fq_codel manages the queue more efficiently, so that the small VoIP packets are worked into available spaces, which means they're not waiting in the queue for other 'stuff' to finish.

That's way over simplified, but is it an accurate way to describe it?

@belt9:

for VoIP, I would strongly recommend you read the linked thread in its entirety, then ask whatever questions you may have either here or in the traffic shaping forum. - playing with fq_codel in 2.4

In short, VoIP traffic suffers badly from high latency as I'm sure you know.

pfSense just implemented (as a result on upstream FreeBSD developments) fq_codel.

Why do you care? Because fq_codel is very easy to implement over your entire network, it does not require much tuning (usually none at all) and it will do an exceptional job of giving traffic such as VoIP very low latency connections even when your WAN is maxed out.
It's currently a CLI-only implementation, but very easy to do (done in <5 minutes).

Thanks for the information. I will not go for this product then, had no idea this was their propose.

I saw an update online from the microsoft for that type of chipset.  From 2010!

I think I've decided to move the C2758 to Site B and use the i3-6100 at Site A.  Will report back my results when this is all up and running.

Can you post the following from console command prompt:
usbconfig (find ugen for the modem then do a dump_device_desc for the VID and PID)
ls /dev

at!entercnd="A710" OK at!udusbcomp? !UDUSBCOMP: 8 OK at!udusbcomp=? 0  - reserved                                    NOT SUPPORTED 1  - DM  AT                                      SUPPORTED 2  - reserved                                    NOT SUPPORTED 3  - reserved                                    NOT SUPPORTED 4  - reserved                                    NOT SUPPORTED 5  - reserved                                    NOT SUPPORTED 6  - DM  NMEA  AT    QMI                        SUPPORTED 7  - DM  NMEA  AT    RMNET1 RMNET2 RMNET3        SUPPORTED 8  - DM  NMEA  AT    MBIM                        SUPPORTED 9  - MBIM                                        SUPPORTED 10 - NMEA MBIM                                    SUPPORTED 11 - DM  MBIM                                    SUPPORTED 12 - DM  NMEA  MBIM                              SUPPORTED 13 - Config1: comp6    Config2: comp8            NOT SUPPORTED 14 - Config1: comp6    Config2: comp9            SUPPORTED 15 - Config1: comp6    Config2: comp10            NOT SUPPORTED 16 - Config1: comp6    Config2: comp11            NOT SUPPORTED 17 - Config1: comp6    Config2: comp12            NOT SUPPORTED 18 - Config1: comp7    Config2: comp8            NOT SUPPORTED 19 - Config1: comp7    Config2: comp9            SUPPORTED 20 - Config1: comp7    Config2: comp10            NOT SUPPORTED 21 - Config1: comp7    Config2: comp11            NOT SUPPORTED 22 - Config1: comp7    Config2: comp12            NOT SUPPORTED OK

For an AM3+ system?

I'm not really aware of too many boards with onboard graphics, but tossing away one of the x16 slots on my ASUS M5A99FX PRO R2.0
that I'm using for a FreeNAS build (I already had it and it supports ECC) doesn't hurt so much, especially since I had an extra Quadro 600 card lying around that doesn't consume the slot immediately adjacent to it like many GPUs do.  My board has 4 x16 (physically) slots; 2 of them actually have 16 lanes and the other 2 have 4 lanes.  I'm using one of the x4 slots for the GPU; that leaves 2 x16 and 1 x4 for other devices.  One slot holds a x4 Intel NIC, one of the x16 slots has an x8 SAS HBA in it, and there's still one free, even with the GPU in there.

I used to have that same Asrock board as well, it's a fine board. I ran 3 GPUs on it and it didn't break a sweat. There's no reason NICs in those slots would put any undue strain on it.  If you need to dedicate the multi-lane PCIe slots to NICs, I'd look at picking up an old PCI graphics card.  There's also at least one PCIe x1 GPU out there IIRC.  You don't need a lot of bandwidth for the console on pfSense.

And finally, you could also look at consolidating NICs by picking up a server pull quad NIC or two. That way you could have 8 physical ports in just 2 PCIe x4 slots.

Hope this helps.

EDIT:  Just re-read and realized you have 10Gbps cards in there.  So disregard the last bit.  If you really want a quality AM3+ board the M5A99FX PRO R2.0 is great.  2x16 and 2x4 all in x16 physical configuration.  Plus a x1 and a legacy PCI slot for good measure.  Enough for your 10Gbps cards and a GPU.