RT-AX58U VPN Speeds?

[HarryMuscle]

Anyone know what kind of VPN speeds to expect on a RT-AX58U router? Does it have any hardware acceleration for VPN?

Thanks,
Harry

 

[RMerlin]

No hardware acceleration, no. Expect about 100 Mbps max for OpenVPN throughput in general.

 

[HarryMuscle]

No hardware acceleration, no. Expect about 100 Mbps max for OpenVPN throughput in general.

That's not bad. I was worried it would be on par with the RT-AC68U which only gets around 30 Mbps.

Thanks,
Harry

 

[RMerlin]

That's not bad. I was worried it would be on par with the RT-AC68U which only gets around 30 Mbps.

Thanks,
Harry

The RT-AC68U can hit about 50 Mbps, with its 800 MHz CPU. The RT-AX58U is 1.5 GHz, so the performance progression is fairly linear.

 

[Odkrys]

RT-AX58U seems to be a very suitable router for WireGuard like raspberry pi. no aes instruction but quad core 1.5ghz cpu. So I think it can easily achieve ~700Mbps.
For example, pi4 achieved 800~850Mbps. But pi4 has Cortex-A72 quad-core 1.5Ghz, so it will be a bit lower than that.

 

[RMerlin]

For example, pi4 achieved 800~850Mbps. But pi4 has Cortex-A72 quad-core 1.5Ghz, so it will be a bit lower than that.

Cortex A72 probably has AES acceleration operands however. The RT-AX58U Cortex ARMv7 doesn't.

type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
aes-256-cbc      22704.70k    26863.03k    28240.30k    28661.42k    28691.11k    28639.23k

 

[Odkrys]

Cortex A72 probably has AES acceleration operands however. The RT-AX58U Cortex ARMv7 doesn't.

type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
aes-256-cbc      22704.70k    26863.03k    28240.30k    28661.42k    28691.11k    28639.23k

The throughput is WireGuard's. It doesn't related with aes .
And raspberry pi series has not purchased an AES license for price.

 

[Yota]

Cortex A72 probably has AES acceleration operands however. The RT-AX58U Cortex ARMv7 doesn't.

Maybe it has AES?
BCM6750 Documentation:

• Comprehensive wireless network security support that includes WPA,WPA2, and AES encryption/decryption

 

[RMerlin]

Maybe it has AES?
BCM6750 Documentation:

That's wireless support, not CPU acceleration.

 

[RMerlin]

The throughput is WireGuard's. It doesn't related with aes .
And raspberry pi series has not purchased an AES license for price.

I see it uses Chacha20. Sounds like an odd choice, because it means Wireguard will lack hardware acceleration on any platform.

 

[Odkrys]

I see it uses Chacha20. Sounds like an odd choice, because it means Wireguard will lack hardware acceleration on any platform.

That's right. WireGuard maybe fine up to 10gbps. But beyond that, aes accelerated aes-gcm performs best. (I know that IPsec can handle 40gbps)
I think one day, WireGuard might introduce aes-gcm.

 

[RMerlin]

That's right. WireGuard maybe fine up to 10gbps. But beyond that, aes accelerated aes-gcm performs best. (I know that IPsec can handle 40gbps)
I think one day, WireGuard might introduce aes-gcm.

Even with a Cortex-A53, hardware accelerated AES is faster than Chacha20:

type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes  16384 bytes
chacha20         53445.72k   127876.59k   291298.82k   337318.57k   354241.19k   355226.97k
aes-256-gcm      59369.00k   182342.27k   383542.27k   540085.59k   608064.85k   608403.46k

So, chacha20 is mostly of benefit for non-accelerated devices.

 

[Odkrys]

Even with a Cortex-A53, hardware accelerated AES is faster than Chacha20:

But we know the number doesn't mean real throughput.
On RT-AC86U, IPsec = ~300Mbps / WG = ~450Mbps.

Surprisingly, finding IPsec benchmarks on ARM64 devices is very difficult.
Perhaps the result you did is the only one.
Maybe it's faster than chacha20 if WireGuard supports aes-gcm?

AES acceleration seems to be more effective on intel devices than arm.

Edit. And chacha20 did not behind aes-gcm in openvpn.
https://www.snbforums.com/threads/openvpn-performance-of-the-rt-ac86u.41217/page-13#post-457660
As all standards follow Intel's aes-ni, so aes acceleration seems to be more complicated than I think.

 

[Makaveli]

RT-AX58U seems to be a very suitable router for WireGuard like raspberry pi. no aes instruction but quad core 1.5ghz cpu. So I think it can easily achieve ~700Mbps.
For example, pi4 achieved 800~850Mbps. But pi4 has Cortex-A72 quad-core 1.5Ghz, so it will be a bit lower than that.

The 58U is a 3 core design not quad, which has no AES

Cortex A72 details.
https://www.anandtech.com/show/9184/arm-reveals-cortex-a72-architecture-details

A72 does support AES
http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.100097_0002_04_en/way1395175472464.html

more info on the Cortex Cores.
https://www.mccoycomponents.com/blog/view/arm-cpu-architecture-system-armv8-armv7-a-series-of-cpu

 

[Odkrys]

The 58U is a 3 core design not quad.

oh. yes you right. Confused with ax56u.

their line up is really mess..
AX58U has triple-cord cpu, 2 stream 2.4ghz and 4 stream 5ghz.
AX56U has quad-core cpu, 2 stream 2.4ghz and 2 stream 5ghz.

better cpu but worse wifi..
I don't know which one is good really.

 

[Odkrys]

A72 does support AES

AES instruction is not free.
Probably soc maker Broadcom did not buy a license from ARM to lower the price.
Raspberry Pi3 and 4 do not support aes acceleration.

 

[RMerlin]

better cpu but worse wifi..
I don't know which one is good really.

The BCM6755 is a higher-end SoC as it has a fourth CPU core and both wifi built-in. Asus went with the cheaper tri-core and single band SoC for the RT-AX58U, which allowed them to add a more capable 5 GHz radio than what the RT-AX56U has.

That fourth core will never make any measurable difference in real life performance, but the external 5 GHz radio will. So globally, the RT-AX56U will be a lesser performer than the RT-AX58U, and be also slightly cheaper (due to everything being built in a single chip).

 

[tiko]

The BCM6755 is a higher-end SoC as it has a fourth CPU core and both wifi built-in. Asus went with the cheaper tri-core and single band SoC for the RT-AX58U, which allowed them to add a more capable 5 GHz radio than what the RT-AX56U has.

That fourth core will never make any measurable difference in real life performance, but the external 5 GHz radio will. So globally, the RT-AX56U will be a lesser performer than the RT-AX58U, and be also slightly cheaper (due to everything being built in a single chip).

Merlin do you think the AX58U will be the replacement of the AC68U in terms of support? Because what you just have explained might allow for the AX58U to be like a very nice router for average user, maybe making it massive, or am I wrong? Thanks!

 

[Deleted member 22229]

I think you are correct. Stellar performance at a great price.

 

[Gar]

Makes sense.