PC Engines apu2 experiences

stephenw10

Yes. It will show the drive wear level as spare blocks are used for example. It can't show if the controller is about to die.

Steve

sikita

Hi guys, are you running pfSense with /var & /tmp in RAM DISKS enabled on APU2?

fireodo

@sikita said in PC Engines apu2 experiences:

Hi guys, are you running pfSense with /var & /tmp in RAM DISKS enabled on APU2?

Hi, no RamDisk here - as stated in my signature its all on a 16GB SSD.
/var/run 4% von 3.4MiB - ufs im RAM

Qinn

@sikita The whole filesystem zfs is running on 16GB mSATA SSD with ZFS only /var/run
4% of 3.4MiB - ufs in RAM

kinch

i retested. same result.
iperf client pc(lan) -> pfsense APU2E4 -> pc2 (opt) iperf server

With 2.4.4 p3 I have 750-850Mbit for 1 connection
After the update to 2.4.5 P1 I have 400-500Mbit.
Nothing changed, only the update was installed.
For me it means, there has changed something in the performance between 2.4.4 P3 and 2.4.5 (P1). But i dont know what.

BR

dugeem

@kinch
Try disabling ICMP Redirects to enable the faster tryforward routing path (under System / Advanced / System Tunables set net.inet.ip.redirect & net.inet6.ip6.redirect to 0)

sikita

@Qinn said in PC Engines apu2 experiences:

@sikita The whole filesystem zfs is running on 16GB mSATA SSD with ZFS only /var/run
4% of 3.4MiB - ufs in RAM

Ok, thanks for a hint

Stella_161

Hi,

First id like to say thanks for everyone advice.

Id like to ask for a little help from you.

Im having issues with slow LAN interface compared to the WAN.

I have a 100Mbit PPPoe HFC connection. When running a speed test from the apu4c4 or directly plugged in to the HFC box, i get 98Mbit but if i run it from my PC that's behind the apu4c4 its only at 71Mbit no matter what i do.

Ive tried:

Swapping Ethernet cables
Plugging directly in to LAN Port 2 and 3
https://teklager.se/en/knowledge-base/apu2-1-gigabit-throughput-pfsense/
disabling ICMP Redirects to enable the faster tryforward routing path (under System / Advanced / System Tunables set net.inet.ip.redirect & net.inet6.ip6.redirect to 0)
Upgraded to 2.4.5-RELEASE-p1 (amd64) from 2.4.4
Upgraded the BIOS to: Version: v4.11.0.4
Factory reset with no packages enabled

All with no change to my LAN speed. I'm all out of ideas here, can anyone offer any more advice on what i can try next.

Thanks

sToRmInG

How is your CPU utiluzation when you run your test?
Can you try and update the BIOS to 4.12.0.2 and see if this helps?

Stewart

@Stella_161 This is specific to your box, not a general issue with the devices themselves so you would get a much better response opening a general support thread for this. When you do, you'll need to give more detail, specifically what packages are installed and how you're performing the test. Try a mix of speedtest.net and fast.com and see what you get. Make sure to do them both with an ad-blocker running on your PC. I've seen many instances where advertisements can eat up a CPU on a computer and give artificially slow speed results.

I have many of these boxes that are doing up to 500Mbps so the hardware is more than capable. That being said, @sToRmInG is right in that the first place to look would be the CPU %. Something may be chewing up CPU cycles on either the APU2 or the laptop you are using. Obvious suspects could be Snort / Suricata, Squid / SquidGuard, and pfBlockerNG.

Stella_161

@Stewart @sToRmInG

Thanks Guys, Ive opened a new thread for this.

Ive tried upgrading the BIOS with no change. After the BIOS upgrade i did a factory reset, no packages have only setup PPPoe and default LAN with ad blocker enabled on PC. Still only 71Mbit~.

CPU usage is minimal when running from PC: Speedtest.net and Fast.com with the same result

PID USERNAME PRI NICE SIZE RES STATE C TIME WCPU COMMAND
11 root 155 ki31 0K 64K RUN 3 42:59 100.00% [idle{idle: cpu3}]
11 root 155 ki31 0K 64K CPU1 1 42:57 100.00% [idle{idle: cpu1}]
11 root 155 ki31 0K 64K CPU2 2 42:56 98.49% [idle{idle: cpu2}]
11 root 155 ki31 0K 64K CPU0 0 42:42 88.96% [idle{idle: cpu0}]
12 root -92 - 0K 496K CPU0 0 0:31 14.70% [intr{irq256: igb0:que 0}]
12 root -92 - 0K 496K WAIT 2 0:08 5.66% [intr{irq259: igb1:que 0}]
12 root -92 - 0K 496K WAIT 3 0:07 1.66% [intr{irq260: igb1:que 1}]

CPU when running Speed Test CLI on pfsense:

PID USERNAME PRI NICE SIZE RES STATE C TIME WCPU COMMAND
11 root 155 ki31 0K 64K CPU1 1 47:13 86.77% [idle{idle: cpu1}]
11 root 155 ki31 0K 64K RUN 3 47:12 82.57% [idle{idle: cpu3}]
11 root 155 ki31 0K 64K CPU2 2 47:09 79.88% [idle{idle: cpu2}]
11 root 155 ki31 0K 64K RUN 0 46:46 71.29% [idle{idle: cpu0}]
12 root -92 - 0K 496K WAIT 0 0:50 33.25% [intr{irq256: igb0:que 0}]
52611 root 74 0 135M 71416K CPU3 3 0:02 7.96% /usr/local/bin/python3.7 /usr/local/bin/speedtest{python3.7}

Cheers

Qinn

@Stella_161 Just to be sure, did you halt the system and then pulled the power plug, wait for a minute and then re-plug the power, this should be done after every bios update.

kevindd992002

If I use a USB2.0 cable to the internal header of the APU2C4, how do I route it so that I can access it externally from the official black case provided by PCEngines?

dotdash

@kevindd992002 said in PC Engines apu2 experiences:

If I use a USB2.0 cable to the internal header of the APU2C4, how do I route it so that I can access it externally from the official black case provided by PCEngines?

When I loaded from the USB header, I had the top of the case off. I think you'd have to modify the case if you wanted to put the top back on.

kevindd992002

@dotdash said in PC Engines apu2 experiences:

@kevindd992002 said in PC Engines apu2 experiences:

If I use a USB2.0 cable to the internal header of the APU2C4, how do I route it so that I can access it externally from the official black case provided by PCEngines?

When I loaded from the USB header, I had the top of the case off. I think you'd have to modify the case if you wanted to put the top back on.

Yeah, drill through it or something. I would love to just use the USB2.0 header with external USB ports than wait for the devs to fix that long-standing USB3 issue.

kevindd992002

@veldkornet said in PC Engines apu2 experiences:

@kinch said in PC Engines apu2 experiences:

@DaddyGo

I've examined it and found the following:

iperf3 LAB
No matter if APU2/3/4 I get a throughput with single connection of 750-850Mbit/s (LAN 2 OPT)

With 2 or more connections I get about 945Mbit/s.

A Speedtest.net gives the value 934Mbit/s UP/DOWN

I have tested the following BIOS versions:
V4.9.0.2
V4.11.0.2
V4.12.0.1

All the same.

But what I noticed, tested on APU2 and APU3, that pfsense version 2.4.5 does not reach the same throughput as 2.4.4-p3.
With Single Connection pfsense 2.4.5 reaches "only" about 400-500Mbit/s. With pfsense 2.4.5 I can only reach 940Mbps with 8 connections.

I also noted something weird with 2.4.5.

My iPerf speeds were also much lower than normal. What I did to fix it was:

disabling hardware offload (check the check boxes)

save, no reboot (iPerf results the same)

re-enabling hardware offload (clear the check boxes)

save, no reboot (iPerf results back up to the expected levels)

The only thing I did between having good iPerf results and slow ones was reboot the device. It's as if after a reboot, it forgot that hardware offload was enabled, and I just needed to set it again.

@dugeem

Do you have any comments on this? As we discussed before, here are the tweaks I have:

/boot/loader.conf.local contains:

hw.igb.rx_process_limit=-1

Also, it looks like the guys here updated their article but they are still insisting to enable all three HW offloads as they see that it gives better speeds.

And what is the recommended setting for hardware checksum offload? Checked (disabled) or unchecked (enabled)?

I'm asking again because my ISP connection is at 400/400 now and I'm thinking of upgrading to 800/800 soon. I want to know if my APU2C4 can handle these speeds with the tweaks that we have.

@Veldkornet

Which of the HW offloads did you enable (unchecked) to get significantly better multi-connection speeds?

@VAMike

What is your comment on the HW offload settings? I see that you had a comment against the post of Veldkornet where he enabled those settings to get better speeds. If he saw better speeds with those enabled, is that an inaccurate test or something because it only helps the client and not end-to-end? Sorry if I misunderstood.

Qinn

@kevindd992002 Why only the receive (rx) and not the transmit (tx) in the loader.conf.local file and why did you not uncheck?

Hardware TCP Segmentation Offloading
Hardware Large Receive Offloading

kevindd992002

@qinn said in PC Engines apu2 experiences:

@kevindd992002 Why only the receive (rx) and not the transmit (tx) in the loader.conf.local file and why did you not uncheck?

Hardware TCP Segmentation Offloading
Hardware Large Receive Offloading

Answers to your questions:

https://forum.netgate.com/post/908003

Qinn

@kevindd992002 You are refring to my own reply, but these settings are from a older date and referred to pre pfSense version 2.4.5-p1

You should now have these and then reboot

pfSense-2.4.5-p1.png

https://forum.netgate.com/topic/95148/pc-engines-apu2-experiences/384

kevindd992002

@qinn said in PC Engines apu2 experiences:

@kevindd992002 You are refring to my own reply, but these settings are from a older date and referred to pre pfSense version 2.4.5-p1

You should now have these and then reboot

https://forum.netgate.com/topic/95148/pc-engines-apu2-experiences/384

I'm still confused. I wasn't referring to your reply. I was referring to @dugeem 's reply when we discussed this.

Again, read https://forum.netgate.com/post/908003

Let me quote some parts of his reply:

"Firstly TSO & LRO should always be disabled on routers. Netgate recommend this (hence pfSense defaults) as do others. BSDRP have even tested this and found routing performance drop negligible from enabling TSO & LRO (see link below)."

"hw.igb.tx_process_limit already defaults to -1 so no need to change this."

So those are the answers to your questions. Also, like I mentioned in my reply today:

"Also, it looks like the guys here updated their article but they are still insisting to enable all three HW offloads as they see that it gives better speeds."

So, to summarize, dugeem explained very well that TSO and LRO needs to be disabled (so checked) and that the teklager article is wrong. With their updated article, they are still insisting to keep those three offloads enabled.

Everything clear now?