OpenVPN on the BCM4912 (ZenWiFi Pro XT2 or GT-AX6000)

[RMerlin]

A quick iperf3 test run through an OpenVPN connection between my desktop and an XT12:

[ ID] Interval           Transfer     Bandwidth
[  4]   0.00-10.00  sec   288 MBytes   241 Mbits/sec                  sender
[  4]   0.00-10.00  sec   288 MBytes   241 Mbits/sec                  receiver

Mostly in line with the 200 MHz clock increase of the BCM4912. I was curious whether the newer kernel or other CPU improvements might have brought another impact.

 

[dave14305]

What kernel version?

 

[RMerlin]

What kernel version?

4.19.183.

 

[tarzan2000]

Mostly in line with the 200 MHz clock increase of the BCM4912. I was curious whether the newer kernel or other CPU improvements might have brought another impact.

Broadcom BCM4912 A0 (2.0 GHz, 4 сores ) vs Broadcom BCM4908 (1.8 GHz, 4 cores)?

 

[RMerlin]

Broadcom BCM4912 A0 (2.0 GHz, 4 сores ) vs Broadcom BCM4908 (1.8 GHz, 4 cores)?

Yes. However the BCM4912 contains other enhancements, I was curious whether any of these would also have an impact on VPN performance.

 

[tarzan2000]

whether any of these would also have an impact on VPN performance.

Waiting for WG finally

 

[RMerlin]

Waiting for WG finally

WG is not a solution here, as it will require disabling NAT acceleration, which will simply shift your bottleneck elsewhere.

 

[RMerlin]

I also tested CAKE. Still cannot keep up with my 400 Mbps connection unfortunately, so still no cake for me... CAKE hits 370 Mbps with 100% core usage.

Without CAKE (but hardware NAT still disabled), I was able to reach 400 Mbps with 75% core usage, so still some headroom left.

 

[Sylphia]

WG is not a solution here, as it will require disabling NAT acceleration, which will simply shift your bottleneck elsewhere.

WG requires disabling NAT Acceleration - is it BCM4912 specific or applies to all supported models?
Testing WG with iperf3 on my AC86U gives me:

[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-10.01  sec   591 MBytes   495 Mbits/sec  573             sender
[  5]   0.00-10.05  sec   588 MBytes   491 Mbits/sec                  receiver

But when routing traffic by WG in daily use, I got few "BLOG ERROR" in log entries.

 

[RMerlin]

WG requires disabling NAT Acceleration - is it BCM4912 specific or applies to all supported models?

All supported models.

But when routing traffic by WG in daily use, I got few "BLOG ERROR" in log entries.

Asus disabled it upstream in their own implementation, and I believe the WG script devs also confirmed that there were issues when running WG without first disabling flow cache/runner.

 

[Yota]

Yes. However the BCM4912 contains other enhancements, I was curious whether any of these would also have an impact on VPN performance.

I'm curious to know, will the power consumption of the BCM4912 be lower than that of the BCM4908? or higher? and do they have better temperature? is there power management for this CPU, such as lowering the CPU frequency to save power when it's not under load?

 

[RMerlin]

I'm curious to know, will the power consumption of the BCM4912 be lower than that of the BCM4908? or higher? and do they have better temperature? is there power management for this CPU, such as lowering the CPU frequency to save power when it's not under load?

Broadcom does not publish any public details, it`s only available to their partners.

Quad-core ARM v8 compliant 64 bit processor for residential access point applications BCM4912

The BCM4912 is a quad-core ARM v8 compliant 64-bit processor for residential access point (AP) applications.

www.broadcom.com

Because, heaven forbid someone other than a Fortune 500 paying customer might learn anything about their products. That would totally give their competitors a major advantage.

Their BCM4912 product page is even more useless than their BCM4908 page, which had a few more technical details.

Also, I suspect there might be different revisions of the BCM4912, because the internal details I have don`t fully match those of that website...

 

[Yota]

Broadcom does not publish any public details, it`s only available to their partners.

Quad-core ARM v8 compliant 64 bit processor for residential access point applications BCM4912

The BCM4912 is a quad-core ARM v8 compliant 64-bit processor for residential access point (AP) applications.

www.broadcom.com

Because, heaven forbid someone other than a Fortune 500 paying customer might learn anything about their products. That would totally give their competitors a major advantage.

Their BCM4912 product page is even more useless than their BCM4908 page, which had a few more technical details.

Also, I suspect there might be different revisions of the BCM4912, because the internal details I have don`t fully match those of that website...

Do you still believe that for this reason they continue to be closed?

Manufacturers paranoia and FCC clamping down on radio abuses have led to this.

 

[RMerlin]

Do you still believe that for this reason they continue to be closed?

Closing down source code and not even publishing actual hardware specs on the website are two different thing however.

 

[Yota]

Closing down source code and not even publishing actual hardware specs on the website are two different thing however.

Broadcom has never been friendly to the open source community. Even their wifi adapter driver didn't add support for Linux until 2012 If I remember correctly, I don't think consumers and the FCC are the reason they chose to close. Again they chose not to provide details this time because they believed their "qualified" clients had everything they needed. I believe that in the end, they will go to the grave with their own secret technology and products.

 

[N/A]

I'm curious to know, will the power consumption of the BCM4912 be lower than that of the BCM4908? or higher? and do they have better temperature? is there power management for this CPU, such as lowering the CPU frequency to save power when it's not under load?

Most likely huge improvements on power and temperature. BCM4908 is built on (fairly sure it's TSMC because the other fab with 28nm is samsung and broadcom usually use TSMC) 28nm while BCM4912 is built on (only TSMC markets as 16nm so it must be them) 16nm. The density of the process almost doubled from 15 to 29. Given that BCM4912 is mostly a refresh of BCM4908 with only 0.2GHz frequency bump, you can expect BCM4912 to consume about 45% the power of BCM4908.

 

[L&LD]

Or give more performance (than the 200MHz indicates) for the same power consumption.

 

[maxbraketorque]

WG is not a solution here, as it will require disabling NAT acceleration, which will simply shift your bottleneck elsewhere.

First I've heard of this tradeoff. One of your later posts seems to suggest that the ASUS-native version of WG will also require disabling NAT acceleration. Can you confirm?

 

[RMerlin]

First I've heard of this tradeoff. One of your later posts seems to suggest that the ASUS-native version of WG will also require disabling NAT acceleration. Can you confirm?

Yes, any WG implementation is incompatible with hardware NAT.

 

[RMerlin]

Most likely huge improvements on power and temperature. BCM4908 is built on (fairly sure it's TSMC because the other fab with 28nm is samsung and broadcom usually use TSMC) 28nm while BCM4912 is built on (only TSMC markets as 16nm so it must be them) 16nm. The density of the process almost doubled from 15 to 29. Given that BCM4912 is mostly a refresh of BCM4908 with only 0.2GHz frequency bump, you can expect BCM4912 to consume about 45% the power of BCM4908.

Not sure this is true, as my CPU idles at around 82C, pretty similar to my previous router.