PfSense with Gigabyte GA-J1900N-D3V
-
They released a new Bios (F4) this month for "Improve USB device compatibility". Has anyone tried it yet? I had to go the MS-DOS Bootable Flash Drive route just to get to F3, and am leery of trying for F4.
-
Finally, do you recommend to use GA-C1037UN-EU (rev2) http://es.gigabyte.com/products/page/mb/ga-c1037un-eurev_20/overview/?
-
I just got my board and I love it!! I am having difficulty with one thing though. Online gaming is giving me a tough time.
I am getting latency spikes while i play. Does anyone have experience with slowness using these boards and gaming? I was trying to tweak settings, but haven't been able to find anything that fixes it.
-
Finally, do you recommend to use GA-C1037UN-EU (rev2) http://es.gigabyte.com/products/page/mb/ga-c1037un-eurev_20/overview/?
Am currently using this board and it works fine no issues at all.
-
What is the power comsumption of this board or the Gigabyte GA-C1037UN-EU running pfSense? Has anyone done any measurements?
-
I bought this board yesterday. Bios F3 was already installed. Installation went smooth, but it can't set IP over DHCP. Latest version of pfsense 2.2.4
Do you guys know any solution, beside adding PCI card, i dont have enough space :-\ -
What is the power comsumption of this board or the Gigabyte GA-C1037UN-EU running pfSense? Has anyone done any measurements?
About 10-15W for ga-j1900n-d3v
-
Just got mine yesterday from http://www.mitxpc.com/ Model# EKGBJ1900M350 I got the 4GB RAM instead of the 2GB. Separately I got a Patriot Blaze 60GB SSD for it as well. Mine came with the F4 BIOS and I was able to boot on USB for nano flash setup as well as boot from external USB CD/DVD for full install on the SSD. Only changes I made to BIOS was turn off splash, turn off Vitalization, and disable Audio. Oh ya, using 2.2.4
I setup a test network on my bench with LAN side having my PC and 1000mb switch, and the WAN side having my NAS and a 1000mb switch. I was able to get 80MB throughput with a PC to NAS samba copy of a 3.4GB ISO. Copying it back from the NAS to PC was also 80MB. CPU got to 33% on both runs.
I did the same copy just using a switch, and no router, between the PC and NAS and I get 113MB. This tells me the bottleneck is not the PC or NAS NICs, Cables, or Switches. With the CPU being at 33% as well on the router I am thinking the Realtek NICs in the router.
I am going to setup a VPN on my bench and test this as well. May not get to that for a few days though.
My goal is to turn this into a UTM.
-
Do you guys know any solution, beside adding PCI card, i dont have enough space
If there is no space how you want to add a card then? Startech are producing PCI and PCIe expander
cases perhaps this would then a choice for you. -
@BlueKobold:
Do you guys know any solution, beside adding PCI card, i dont have enough space
If there is no space how you want to add a card then? Startech are producing PCI and PCIe expander
cases perhaps this would then a choice for you.I don't. I was thinking about some config tweaks. Many people had this issue. Im wondering if anybody was able to solve this problem.
-
Where were you measuring that CPU usage? The figure on the dashboard combines all the cores. You need to use the command line to see the individual core usage:
top -SHYou will probably find one core is at 100%.
Steve
-
Does pfsense not use all the Cores?
-
top-SH show two of the four cores at 100% idle during the transfer. The other two cores bounce around form 20% to 80% idle during the transfer. So the average would be 75% idle, or 25% usage. seeing how i am not able to nail down the numbers on the working cores I think the dashboard lower 30's% is fairly accurate. Am i missing something? Does pfSense use all cores?
-
With the CPU being at 33% as well on the router I am thinking the Realtek NICs in the router.
The realtek NICs can be it, but this is not a must be!
I don't. I was thinking about some config tweaks.
Like the most peoples are thinking. If you owns a mSATA or SSD you could try out to activate
the TRIM support and if you owns a CPU that is capable of TurboBoost mode, you could try out
also setting up or activate the PowerD (highadaptive) mode. But if you go by NanoBSD and owns
a mSATA or SSD it would be better to do a full install.Many people had this issue
.
There is no issue! Your switch is only faster then your router not more but also not less!Im wondering if anybody was able to solve this problem.
Which Problem? Let us both imagine you owns two Intel Core i7 CPUs PCs and using iPerf through the
pfSense router the you will see other numbers and if your pfSense is based on a SG-8860 from the
pfSense store I am really sure you will see once more again other numbers!The test your where doing, is comparing your pfSense router (Layer3) against your switch (Layer2)
and this would be not matching any real life scenario!Does pfsense not use all the Cores?
This is even a bit more or less changing at the moment, in earlier days pfSense was only using one CPU
core at the WAN interface but more CPU core for the rest of the entire system. But the developers got
even many more skills and then this thing will be during a change at the time.But on the other hand it will not change anything for you. If you go by hardware fiddled together by
your own and then the pfSense is not serving the same numbers as your switch this is not pointed to
pfSense, then more at the switch you compared to! If you are using a Intel Xeon E3-1286v3 @3,7GHz
and Intel 10 GBit/s server NICs you will archive total other numbers for sure and then it is not relevant
how many cores was in the game but more from which CPU and on which frequency it was running on. -
Exactly.
In pfSense <2.2X the pf process was giant locked and only ever used a single core so machines with fewer but faster cores were preferred if raw throughput was the aim. Since 2.2 the new multithread capable pf in FreeBSD 10 means this is less of an issue but it still won't spread the load evenly across all the cores. You can't just use the dashboard CPU meter when you start hitting limits.
That said you aren't seeing any cores at 100% so I would also start to suspect the Realtek NICs.Steve
-
Well my pico psu died so i had to wait until they shipped me another. I have the system backup since Friday night. I put windows on it and ran a series of test over a two day period. (memtest, prime95, etc) Everything seems fine. I got pfSense back on it today.
I was doing some surfing and read about the checksum offloading. I have that disabled now and my performance went up and CPU utilization went down. I can now copy files across the router at 89MB, both ways, with only 26% CPU utilization (dash board meter). That's still slower than without the router by about 24MB but better than the shaky 80MB i was getting before disabling the checksum offloading. I will try some other stuff/tweaks but if it does not get any faster than that, and not loose stability, I can live with it.
I have plans to order another and will set it up next to the one I just finished. I will then do VPN between the two and see what I get for performance there. I hope to make another post with findings in the next few weeks.
BTW Thinking of getting the other router with Intel NIC's so I can compare the difference there as well. Will still use the J1900 2.4GHz though. This seems like a really low power, robust little CPU.
Thanks!
-
So for grins I grabbed a PC from my sons room. Its a AMD 6 core 4.0Ghz CPU with a built in Realtek. I pulled a TPLink card from My NAS i am not using at the moment (It is using the on board Qualcom) and stuck it in the PC to get dual ethernet. I also stuck a WD SATA 3 10,000 RPM drive i have laying around in the PC so i could do a fresh install of pfSense without using the SSD in the PC. I ran the same copy test and got 113MB through the router, same speed i get not using a router and just going through a switch. This PC shows two realtek 8168/8111 cards, just like my mini ITX machine that I am getting just under 90MB from. I used same switch and cables in this test.
The PC did not even max out a single core, and never touched the other 5. The mini ITX bumps 100% on one core while a second core also get used. The other 2 cores see no action.
My conclusion so far is pfSense does not do a good job with multi cores. How sad in this day and age of muti core CPU's that BSD has such a problem. A quad core 2.0 with 2.4 boost is not able to max a 1Gb connection.
With Windows 7 on this same Mini ITX I was able to get over 112MB on the transfer.
So again, its not my hardware, but a limitation of the OS to not use all my hardware.
-
I dropped Centos 6.7 with KVM on the mini ITX. Installed pfSense guest using the virtio Intel n1000 drivers. As others reported I get double the CPU usage and half the bandwidth. I guess if i want true 1Gb bandwidth I will have to go with a system using something like an i3 Dual Core 3.4Ghz.
For now i may just turn the mini ITX into a Session Border Controller so I have something new to play with :D
-
i ordererd this Gigabyte GA-J1900N-D3V 2 Weeks ago which came already with F3 Bios.
Everything worked (Sophos, IPFire, VYos etc) . Couldn´t install on my 128gb Crucial SSD Pfsense. During formatting process it showed me Read Error etc. so i thought my SSD got problems but everything else worked.
Yesterday i finally got pfsense on it setting Bios F4 to Storage UEFI First and the rest to Legacy. CSM to Legacy and Uefi and after that PFsense would install flawlessly without a hitch.
ah, i used the 64bit Version of PFsense.
-
hopefully this update this board with:
1)intel nics
- n3150 or n3700 braswell cpus @ 14nm as they use less power and have aes-ni
