Some type of NIC aggregation might be appropriate here too.

Enough of these will saturate your FSB, even though they are PCIx cards:

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=251084127667

There exists a driver for lcdproc 0.4, here.
It requires some rewriting for 0.5x as there were some changes made to lcdproc at that time. However:

@Joris:

If you rewrite it for 0.5 (which is not difficult), it can become part of
the distro (0.4.4 is frozen).

Some differences for 0.5:

It uses a pointer that functions as a 'this' or 'self' pointer for the
driver data. See other drivers and the docs (I thought this was
documented already ?). Allocate a private data struct for driver variables so you can run
multiple instances of the driver. You should not use the driver arguments anymore. Instead use the
configfile functions. Some driver functions have been removed. Some functions have a fall-back alternative. This means that if you
don't define them, LCDd will provide the functionality. Keystuff works with names instead of letters. Therefore the getkey
function is now called get_key. Usually the keynames to use are read from
the config file.

You could try to install lcdproc 0.44 to test the driver. You would first need to compile it which might be tricky against 0.44.

Steve

No, sorry.  ;)

The Safenet safeXcell 1141 that came in the X-Core boxes (X500, X700 etc) should be supported. There are drivers in FreeBSD for it and it is recognised at boot but for some reason it does not work, or at least no-one has been able to make it work. That is a mini-PCI card and the X550e doesn't have a mini-PCI slot anyway.
The X550e, and all the other X-Core-e models, have encryption hardware on the board in the form of a Cavium Nitrox lite chip. Unfortunately there are no drivers easily available for it. It requires propriatary binary modules to work under Linux.
The X-Peak-e models had a large PCI-e VPN card with various encryption chips on it known as the 'tacoma module' that too is unsupported.  ::)

Steve

Exactly. Though the console is still available for basic tasks (and for much more complex stuff!).

Steve

without managment is working

Thanks, its a FESTOOL Oberfräse OF 1010 EBQ-Plus :)

Thanks so much for your reply Steve!

Traffic going through interfaces requires CPU usage, gotcha.

If i go VLAN, it would seem all traffic go through pfSense, as I would be firewalling various subnets., gotcha.

The trouble is there are so many variables that hard figures can be misleading.

There are a lot of variables but then there are a lot of pfSense users. Mashing a few 10,000 tables of data together would show some commonalities i would have thought. Shame I lack statistical analytic skills, would be fun and beneficial to the community.

Thanks so much again, gunna go grab some hardware!

Michael

Working perfectly with all 5 Intel NICs here too!

Thats great :) Thank you…

Emre.

Cool, thanks.  Was not sure if that meant hardware interrupts or no interrupts.

Very interested. I too am currently virtualizing my pfsense box but might be tempted to move to raspberry pi in an effort to reduce costs.

I can't see that you'd need to have more than one network interface although it would be a nice to have. My virtual machine currently has 3 virtual nics but just one physical nic.

Exactly.
Try it and see. You should be able to get your other connection working without replacing the cable by setting both ends to 100Mbps full duplex.

Steve

80W should be enough but one user found it a problem (using a 60W power brick anyway).

The Pentium-M 770 is almost certainly not worth it. There is a huge premium, even now years after release, and the 533MHz FSB Pentium-Ms are not supported directly by the est(4) driver so speedstep doesn't work.

I don't think it supports ECC ram.

Steve

You can backup the config from the old, and restore it to the new, that's works fine.

IMO Passive cooling is a rapidly growing niche, and Intel is doing pretty well with the Atom so far :) But I believe this is getting a bit off-topic. The Atom was first mentioned on this thread just as an example of why the OP shouldn't really need an i5. Many of us suggested to go to a lower-end Sandy like a Pentium or Celeron, but the Atom was never directly suggested as a solution for this, it was only mentioned to make a point.

@stephenw10:

Whilst I agree that an Atom would easily handle a 50Mbps connection with almost any number of users I have to also agree with Asterix's view that it seems pointless to use an Atom if you're building a new box. Low end Sandybridge systems can be built for almost the same cost and will likely consume a similar power level.
If at some later stage you need to implement Squid, Snort, complex traffic shaping or VPNs you could easily run out of CPU cycles on an Atom.
About the only niche left for an Atom is in an entirely passively cooled system where the maximum power dissipation of a Sandybridge CPU is too high to be practical. That particular niche is rapidly being filled by ARM powered CPUs in other markets.
My own personal view.  ;)

Steve

If I recall the details (it's been years since I've seen/looked into that), some NIC chipsets have bugs with short cables, and the drivers apply a work around to fix that. The snippet from if_sis.c seems to confirm that. It's safe to ignore as long as it's not causing problems. If it does, replace the NICs.

I don't know about 3DM2 but a number of people have struggled with LSI raid controllers.
Not that many are supported, especially under FreeBSD8.1 and hence pfSense 2.0.1.
See: http://www.freebsd.org/cgi/man.cgi?query=mfi&apropos=0&sektion=0&manpath=FreeBSD+8.1-RELEASE&arch=default&format=html

LSI do have source on their site for the 9650SE under FreeBSD 8.x though.

Steve