XG-1541 Noise
-
@larrikin https://www.supermicro.com/en/products/system/Mini-ITX/SYS-E302-9D.cfm
-
@stephenw10 said in XG-1541 Noise:
The OpenVPN requirement there is really the only thing I'd expect the 6100 to fall short on. And that is because OpenVPN is (currently) single threaded. The XG-1541 will also not pass 1Gbps OpenVPN over a single connection for that reason. It will do multiple tunnels that sum to 1Gbps though.
I really want the grunt of a decent CPU and the best one is in the 1541, but the noise is the prohibiting factor. I am also not sure that fq_codel won't work well at gigabit speeds with the 6100. It's CPU intensive. In fact, I think the 6100 will fall well short with the different VPN's I'll have running at 256 bit, plus IDS/IPS and running at gigabit speeds.
Is there any chance that Netgate could replace the fans with nocta ultra-quiet fans and a fan controller and sell me the 1541 under those conditions? That would achieve my goal of keeping it quiet, and I get the best box you sell on the market.
-
@q54e3w said in XG-1541 Noise:
@larrikin https://www.supermicro.com/en/products/system/Mini-ITX/SYS-E302-9D.cfm
That looks like a really decent box. Unfortunately I'd be stuck running pfsense CE on it rather than pfsense+. Now if Netgate did a box like that with decent RAM and an SSD, I would be very happy.
-
@Larrikin I understand plus will be available for white label boxes for homely users sooner or later.
-
@q54e3w said in XG-1541 Noise:
@Larrikin I understand plus will be available for white label boxes for homely users sooner or later.
But we haven't been publicly informed as to how they plan to allow that. Until its been publicly announced, I wouldn't buy any hardware based on it being able to run pfsense+ until we hear from Netgate on how they plan to allow pfsense+ to run on third party hardware.
-
@stephenw10 said in XG-1541 Noise:
The OpenVPN requirement there is really the only thing I'd expect the 6100 to fall short on. And that is because OpenVPN is (currently) single threaded. The XG-1541 will also not pass 1Gbps OpenVPN over a single connection for that reason. It will do multiple tunnels that sum to 1Gbps though.
Steve
Another question - what's the difference in architecture for cooling between the 1537 and the 1541? Are the designed exactly the same way? I've seen the 1537 opened up on a youtube video and the only fans that exist are one chassis fan and one power supply fan. The CPU doesn't have a fan on it, its passively cooled.
Is that the same as the 1541?
-
Neither are passively cooled. The CPU heatsink is ducted to the rear chassis fan(s).
The XG-1541 has two chassis fans to pull that air through. There is one there on the XG-1537.Steve
-
@stephenw10 said in XG-1541 Noise:
Neither are passively cooled. The CPU heatsink is ducted to the rear chassis fan(s).
The XG-1541 has two chassis fans to pull that air through. There is one there on the XG-1537.Steve
OK got it. I think I have one last question between the 6100 and the 1541.
One of the things I HAVE to do in Australia is use buffer bloat (fq_codel) which I believe is CPU intensive. Believe it or not, I will have to do it on my 1 gig symmetrical connection, otherwise the National Broadband Network just drop traffic.
In my experience, low CPU performant boxes can't handle gigabit bufferbloat as it wasn't really designed for gigabit symmetrical. It was really designed for lower speed, high latent connections. Would the 6100 handle buffer bloat at symmetrical gigabit as per this setup: https://www.pimdegreef.nl/bufferbloat-solution-for-pfsense/
You've got me thinking that maybe the 6100 is OK for what I want. The thing is, I will have the potential for multi WAN links at a gigabit speed and that buffer bloat solution needs to be applied to each one, so its a question of whether I need the 1541 for its CPU or whether the 6100 will handle it just fine.
BTW, do you know what the DB noise is once the 1541 is booted up and not under heavy load?
Thanks again for your help. I am getting close to buying - just want to know which way to go.
-
@stephenw10 said in XG-1541 Noise:
Neither are passively cooled. The CPU heatsink is ducted to the rear chassis fan(s).
The XG-1541 has two chassis fans to pull that air through.So to clarify, there are only two fans that I'd need to replace on the Netgate 1541 (the chassis fans) with Noctua fans (ultra quiet) if I wanted a more quiet Netgate. Is that something you can do or will I have to do that myself?
-
Hmm, I don't have any absolute data for applying codelq shaping. When I have tested here on the past I never saw any huge CPU use. It doesn't really do much for me here though so I don't use it.
As far as I know it's not buffer-bloat specifically that the NBN has a problem with with but rather an absolute limit. You can probably just use a local limiter to do that. It does just drop packets locally instead however. I could imagine codelq with a fixed bandwidth might give you a better experience. I'll try to get some numbers.I don't have a dB value for the XG-1541, nor do I have one I can easily test.
It is not a desktop device though, it would be loud if you tried to use it that way.We can't swap out fans and we can't recommend you do either. Whilst Noctua fans are (probably) aerodynamically better they mostly make less noise by just turning slower and moving less air. Which is fine in many applications. However these rack mount devices have been tested with the stock fans and controller producing known airflow. Reducing that means running them in an untested condition.
Steve
-
@stephenw10 said in XG-1541 Noise:
Hmm, I don't have any absolute data for applying codelq shaping. When I have tested here on the past I never saw any huge CPU use. It doesn't really do much for me here though so I don't use it.
As far as I know it's not buffer-bloat specifically that the NBN has a problem with with but rather an absolute limit.i’ve found with heaps of different testing bufferbloat is indeed the best way to combat the problem. it works 100% whereas other solutions failed.
I don't have a dB value for the XG-1541, nor do I have one I can easily test.
It is not a desktop device though, it would be loud if you tried to use it that way.i’ve heard the 1537 and that is fine noise wise for me. I’m not looking for silence as i already have three desktop computers running in that room. I just don’t want something screaming at me all day, but a lower noise is totally fine.
We can't swap out fans and we can't recommend you do either. Whilst Noctua fans are (probably) aerodynamically better they mostly make less noise by just turning slower and moving less air. Which is fine in many applications. However these rack mount devices have been tested with the stock fans and controller producing known airflow. Reducing that means running them in an untested condition.
Steve
That’s fair enough. i’m after a box that netgate can fully support as a turn key solution so i don’t want to mess with it and then get old i’ve wrecked it :)
-
hey mate - so just to confirm (as this will be my deciding factor) do you know how loud the 1541 is relative to the 1537? I've heard the 1537 and I'm fine with that noise. I have three desktops in my room and I sit between 40db and 50db of noise. I just don't want the 1541 screaming at me :)
Alternatively, if the 6100 can handle fq_codel at gigabit speeds, then I would seriously consider that.
-
@larrikin Not to bump an old forum, but I am eyeing a 1541 as well. Did you ever get one? If so how loud is it?
