Home Router Recommendation

whosmatt

Okay further careful consideration, I'm going to go with a powerful enough pfSense router that will be capable of handling VLAN's and the inter-VLAN traffic at high speed when it comes up, whether sooner or later. I'll go with Layer 2 Switches.

I'm going to build my own. I'm getting something mostly comparable to the SG-4860. I'll have great power to handle anything I'll throw at it, future proof with the AES-NI and QuickAssist as well. I wasn't opposed to official hardware when I was thinking maybe Layer 3 Switch, but I can't spend over $700 for the router, but I can build a solid equivalent for $375-400 it seems.

Supermicro Mini ITX A1SRI-2558F: $243
4GB ECC RAM: $33
120GB SSD: $30 (Already had one, probably no other use for it until now)
M350 Case with 80w Pico PSU/60w Pico Power Adapter: $69
=$375

I am missing anything else I'd need, besides a SATA cable? I have a number of those.

Sounds like a good build to me. I like the idea of running off a 12V brick as well, but be careful with that purchase. In other words, read reviews and get a good one. Those things always seem to fail on me, and can do really funny things when they do, like causing issues that are really difficult to pin down unless you have another one lying around. Actually I guess that sounds like power supplies in general.

When you're picking out a switch, it might be best to ask around, since the terminology can be confusing. There's the "managed switch" which can be L2 or L3, and generally comes will a full set of features including SSH access and a command line (which many people, myself included, who have worked with them find much better than a web interface), and then there's the "smart switches" with perhaps just a web interface, and now "easy smart" with a Java management app, which is what I have at home. Any of them can do what you want to do in a home network, but the differences in terminology for the various features can be infuriating sometimes.

Build on. Enjoy. And keep us updated.

Matt

Iceman24

Thanks, guys. It seems like the most popular one I found that's being used in similar builds is here. http://www.mini-box.com/12v-12-5A-AC-DC-Power-Adapter

That combo is actually more costly, but seems better, so minor difference in cost is fine. Revised parts list. I added more RAM, too, to be safe and have the interleave benefit, as minor as that may be.

Supermicro Mini ITX A1SRI-2558F: $243
8GB Kit of ECC RAM: $62
128GB SSD: $30 (Refurb Crucial M4, that I've had sitting around way too long)
M350 Case: $40
EDAC 12v 12.5a 150w Power Adapter with needed adapter: $44.50
Female to Female Molex to SATA Cable: ?
P4 Power Extender Cable: ?
=$419.50+

I don't offhand remember pricing on the last 2 cables, too late to lookup. Would this be all and sound good then?

whosmatt

@Iceman24:

Thanks, guys. It seems like the most popular one I found that's being used in similar builds is here. http://www.mini-box.com/12v-12-5A-AC-DC-Power-Adapter

That combo is actually more costly, but seems better, so minor difference in cost is fine. Revised parts list. I added more RAM, too, to be safe and have the interleave benefit, as minor as that may be.

Supermicro Mini ITX A1SRI-2558F: $243
8GB Kit of ECC RAM: $62
128GB SSD: $30 (Refurb Crucial M4, that I've had sitting around way too long)
M350 Case: $40
EDAC 12v 12.5a 150w Power Adapter with needed adapter: $44.50
Female to Female Molex to SATA Cable: ?
P4 Power Extender Cable: ?
=$419.50+

I don't offhand remember pricing on the last 2 cables, too late to lookup. Would this be all and sound good then?

I wouldn't worry about the extra RAM. 4GB is more than enough for a home build. But it's cheap, so no harm if you want to double up. The cables are super cheap. All sounds robust for a home build. Maybe just being paranoid but perhaps put the refurb SSD into a system where you can test it before a build.

whosmatt

@BlueKobold:

Cisco SG300-10PP-K9-NA ~$199

or

Cisco SG300-10-SRW2008-K9-NA ~$189

$375 + $189 = $564

Sorry to go off topic here, but is there a thread you could point me to that discusses the nuts and bolts of these devices? They seem compelling, but my experience with Cisco branded SOHO hardware has been abysmal over the last 10 years or so. Basically my opinion is that anything branded Cisco that isn't in at least a 1U rack config, painted green, and doesn't have fans that make it unpalatable in a home or small office environment is not worth buying. Glad to take it to another thread. Just curious.

Jailer

@Iceman24:

P4 Power Extender Cable: ?

P4 to DC Jack Cable

@Iceman24:

Female to Female Molex to SATA Cable: ?

C2G / Cables To Go 10150 15-Pin Serial ATA Female to LP4 Female Power Cable (6 Inch)

And 150 watt power brick is serious overkill. Save some cash and get a smaller one.

84w (12v/7A) AC-DC Power Adapter with Power Cord

Iceman24

@Jailer:

@Iceman24:

P4 Power Extender Cable: ?

P4 to DC Jack Cable

@Iceman24:

Female to Female Molex to SATA Cable: ?

C2G / Cables To Go 10150 15-Pin Serial ATA Female to LP4 Female Power Cable (6 Inch)

And 150 watt power brick is serious overkill. Save some cash and get a smaller one.

84w (12v/7A) AC-DC Power Adapter with Power Cord

Thanks. I thought it was overkill, too. I didn't understand why that was the go-to option. Is there some reason that it is?

If that 84w one is okay, how would the 60w be? Plenty of power support it has, plus perhaps more efficient with the power draw not being so incredibly low.

jahonix

@Iceman24:

…my dream of an advanced smart home...

That are surely devices you want separated from your LAN.

whosmatt

@Iceman24:

Thanks. I thought it was overkill, too. I didn't understand why that was the go-to option. Is there some reason that it is?

If that 84w one is okay, how would the 60w be? Plenty of power support it has, plus perhaps more efficient with the power draw not being so incredibly low.

I'm no electrical engineer, but my understanding is that commodity power supplies are more efficient when not running near their rated capacity. For example, I've had an 800W PSU draw ~1100 watts from the wall while delivering its rated output, and around 900 watts when the load was reduced just a bit. I suspect heat is a big factor. I doubt a few watts really matters here though. My instinct says to figure your system's draw and size your power supply to deliver that with a 33% margin or so.

edwardwong

PSU we used nowadays mostly "switching regulator" (at least for PC PSUs), conversion efficiency will be higher when you are getting close to the specified rating, if you are consuming power far below the spec, the conversion efficiency will be relatively low.

@whosmatt:

@Iceman24:

Thanks. I thought it was overkill, too. I didn't understand why that was the go-to option. Is there some reason that it is?

If that 84w one is okay, how would the 60w be? Plenty of power support it has, plus perhaps more efficient with the power draw not being so incredibly low.

I'm no electrical engineer, but my understanding is that commodity power supplies are more efficient when not running near their rated capacity. For example, I've had an 800W PSU draw ~1100 watts from the wall while delivering its rated output, and around 900 watts when the load was reduced just a bit. I suspect heat is a big factor. I doubt a few watts really matters here though. My instinct says to figure your system's draw and size your power supply to deliver that with a 33% margin or so.

whosmatt

@edwardwong:

PSU we used nowadays mostly "switching regulator" (at least for PC PSUs), conversion efficiency will be higher when you are getting close to the specified rating, if you are consuming power far below the spec, the conversion efficiency will be relatively low.

That's good to know. My limited (but documented) experience was with cheaper hardware. I've been told that the particular PSU I was using is not truly capable of delivering the stated output, but it did for me, at the expense of a lot of heat and very high power draw from the wall.

I suppose the takeaway from this is that quality matters, and that you can't take manufacturers ratings for granted.

Iceman24

Well, I'm only figuring a 15-20 watt usage of power. I couldn't see it ever getting about 30w, really, at max, from what I can gather. So the 60w seemed like more than enough and I didn't want to waste any power or money on something that was still overkill.

Hakker

Just a heads up it seems that the 2558 boards are running EOL possibly the 2758 also now that the xeons 15x8 are there.
I tried to buy one and after 3 weeks my delivery status went from 2-3 days to 10+ days to unknown and no other stores selling the 2558.

dreamslacker

@whosmatt:

I'm no electrical engineer, but my understanding is that commodity power supplies are more efficient when not running near their rated capacity. For example, I've had an 800W PSU draw ~1100 watts from the wall while delivering its rated output, and around 900 watts when the load was reduced just a bit. I suspect heat is a big factor. I doubt a few watts really matters here though. My instinct says to figure your system's draw and size your power supply to deliver that with a 33% margin or so.

PSUs when certified for power efficiency only need to meet certain targets at 20%, 50%, and 100% capacity loading.
A regular core i5 (or Xeon equivalent), with 1 mechanical drive typically draws about 40-50W at near idle (which is what you'd expect from a pfSense box without much going on). So in theory, you'd want a PSU rated at 250W or lower when you power a rig like this.

When your power draw is below 20% of the PSU rating, that's where you find the greatest impact - mostly because the base power draw to run the components factor in significantly at that kind of loading.

Your best bet, if it's that important to reduce the power losses, is actually to run a DC-DC PSU like a PicoPSU and a high efficiency 12V adapter.

edwardwong

I don't think so, DC-DC PicoPSU is also a kind of switching regulator, just like what I mentioned before, the efficiency will be somewhat lower when your output is far below from the designated load.

@dreamslacker:

@whosmatt:

I'm no electrical engineer, but my understanding is that commodity power supplies are more efficient when not running near their rated capacity. For example, I've had an 800W PSU draw ~1100 watts from the wall while delivering its rated output, and around 900 watts when the load was reduced just a bit. I suspect heat is a big factor. I doubt a few watts really matters here though. My instinct says to figure your system's draw and size your power supply to deliver that with a 33% margin or so.

PSUs when certified for power efficiency only need to meet certain targets at 20%, 50%, and 100% capacity loading.
A regular core i5 (or Xeon equivalent), with 1 mechanical drive typically draws about 40-50W at near idle (which is what you'd expect from a pfSense box without much going on). So in theory, you'd want a PSU rated at 250W or lower when you power a rig like this.

When your power draw is below 20% of the PSU rating, that's where you find the greatest impact - mostly because the base power draw to run the components factor in significantly at that kind of loading.

Your best bet, if it's that important to reduce the power losses, is actually to run a DC-DC PSU like a PicoPSU and a high efficiency 12V adapter.

Iceman24

Not that it makes much of a difference, but I'm going to go with the Kingston Technology ValueRAM 4GB 1333MHz DDR3L PC3-10600 ECC CL9 SR x8 1.35V SODIMM Notebook Memory KVR13LSE9S8/4 instead of the CL11 1600MHz variety. Seems CL9 1333MHz would be faster anyways, plus it's cheaper.

Will do the EDAC 60w as well, since I won't go near 60w at full load.

Supermicro Mini ITX A1SRI-2558F: $243 (http://www.superbiiz.com/detail.php?name=MB-A1RI25)
8GB Kit of ECC RAM: $48 (https://www.amazon.com/Kingston-Technology-ValueRAM-PC3-10600-KVR13LSE9S8/dp/B00CLBJOQW)
128GB Crucial M4 SSD: $30
M350 Case: $40
EDAC 12v 5a 60w Power Adapter: $15 (http://www.mini-box.com/60w-12v-5A-AC-DC-Power-Adapter)
Female to Female Molex to SATA Cable: $6 (https://www.amazon.com/dp/B001CSIPA6/ref=cm_sw_su_dp)
P4 to DC Jack Cable: $5 (http://www.mini-box.com/P4-DC-Jack-Cable)
=$387

Ordering time probably not that far off. :-)

Not that it makes much of a difference, but I'm going to go with the Kingston Technology ValueRAM 4GB 1333MHz DDR3L PC3-10600 ECC CL9 SR x8 1.35V SODIMM Notebook Memory KVR13LSE9S8/4 instead of the CL11 1600MHz variety. Seems CL9 1333MHz would be faster anyways, plus it's cheaper.

Before you are realizing that, please read that thread here on Reddit about RAM and GBit/s WAN speed
on pfSense, please read the comments from the person named "gonzopancho" carefully about this point.
Gigabit routing hardware

Iceman24

@BlueKobold:

Before you are realizing that, please read that thread here on Reddit about RAM and GBit/s WAN speed
on pfSense, please read the comments from the person named "gonzopancho" carefully about this point.
Gigabit routing hardware

Interesting. Thank you. 1600 Mhz it is. :-)

dreamslacker

@edwardwong:

I don't think so, DC-DC PicoPSU is also a kind of switching regulator, just like what I mentioned before, the efficiency will be somewhat lower when your output is far below from the designated load.

It's a lot better than if you use a conventional supply. It's a bit difficult to explain the intricacies of it all (especially once you involve double-forward regulation and such) but it's built to actually handle low loads and you're more likely to be in the 'efficient' region by virtue of the fact that the PicoPSU is rated at much lower power ratings.

For example, if you had a system that needs 30W and you use a 300W PSU, you are at <20% loading. A PicoPSU would start at 60W where your loading is 50% and even if you go up to the 120W model, you are still at 25%.

As mentioned, most PSUs are certified for their 80+ ratings only at 20%, 50%, and 100% load. Drop below 20% load and there are no guarantees on the efficiency anymore.