Resource icon

Is my Realtek ethernet really that bad?

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,680
jgreco submitted a new resource:

Is my Realtek ethernet really that bad? - Yes, it probably is.

One of the longest running issues faced here in the forums may be people who come to these forums, wanting to recycle an old PC into a NAS. In general, this is a bad idea because typical PC hardware isn't optimized towards the task. However, the issue that most commonly bites people is that a NAS is a network oriented device, and having a low end ethernet chipset such as a Realtek or Atheros is likely to lead to inconsistent performance and unpleasant experiences.

As a result, we usually...

Read more about this resource...
 

Ericloewe

Server Wrangler
Moderator
Joined
Feb 15, 2014
Messages
20,194
I just want to say two things:
  1. I cannot claim to have dug up the rl.c comment, it was brought to my attention by someone else.
  2. To anyone optimistic enough to claim that Realtek can do better: Just look at the unmitigated disasters that are Realtek's 2.5GbE NICs and USB NICs. They make the 8111-alphabet-soup family look competent in comparison.
 

jgreco

Resident Grinch
Joined
May 29, 2011
Messages
18,680
I cannot claim to have dug up the rl.c comment, it was brought to my attention by someone else.

Which, ironically, may have been me. I've actually been aware of that comment for many, many years, dating back to the days that Bill Paul wrote the driver. There was a time, way in the distant past, when one of my companies specialized in outsourced Usenet news service that involved placing gear on-site at the local ISP. Being a renegade startup, not funded by venture capital, we used one of a fairly particular set of inexpensive Gigabyte and Tyan AMD Athlon boards for many of the server nodes, and the Gigabytes almost always had 10/100 Realteks on them. We ran a lot of these until they eventually flamed(!) out, I have several burnt-up Athlon CPU's on display in my office.

If you ever wonder why I seem like a Supermicro fanboi, it's really simple. Switching from Tyan/Gigabyte to Supermicro was a major shift in reliability. It went from systems that burned out in three to five years of 24/7 use, to some low end Supermicros that survived fourteen(?) years in revenue service. Plus the Intel ethernets that just gave you full gigabit all the time anytime you wanted. I eventually gave up on the Antec and AICIPC chassis as well, the Supermicro stuff was a better design and not much pricier.

So I have this very strong association with Realtek being junk. And to a lesser extent, Gigabyte (had a lot of boards fall prey to capacitor plague). I talk about the difference between desktop-grade and gamer-grade systems vs. server-grade boards because I've lived the experience.

I am not really a Supermicro fanboi. However, I strongly appreciate the server-compatible aspects of their ecosystem, and much prefer their lego style model of letting you pick from a wide selection of server-optimized parts to make your own ideal config.
 

Constantin

Vampire Pig
Joined
May 19, 2017
Messages
1,829
I am not really a Supermicro fanboi. However, I strongly appreciate the server-compatible aspects of their ecosystem, and much prefer their lego style model of letting you pick from a wide selection of server-optimized parts to make your own ideal config.
Ditto. The universe of options available, long term support, and server-grade component focus is precisely why I went for SM. Ditto customer support, which was excellent and only rivaled by iXsystems. By comparison, what "customer support" I've seen reported for AsRock was variable at best and apparently dependent on whom you got put in touch with. I am so glad that my multiple instances of C2750 unhappiness were all mitigated by iXsytems, including the C2000 issues.
 

das1996

Dabbler
Joined
May 26, 2020
Messages
25
Running a simple homelab on repurposed hardware

asus x570-e strix motherboard
5800x cpu
64gb ram

Proxmox 7.4

The board has 2 nics, an i211 and a rtl8125

Rtl8125 configured as the physical nic to a vmbr0.

Vmbr0 associated with the virtio (vnet0) nic for the truenas 13 (core) vm instance w/ 4 vcores assigned.

Running iperf3 tests back and forth over the lan (client with i211 nic) and truenas vm reveals exceptionally high cpu usage on the proxmox host. Exceptionally as in 200-250% cpu usage . Power wise it spikes upwards of 50 watts relative to idle.

In addition to onboard nics, the box now has a i340-t4 nic. Cpu usage and power results below.

I didn't record hard data, but updating the rtl driver (latest offered on realtek site) on proxmox did improve cpu usage. As does updating host to 6.2 kernel (from 5.15). However, all of these I think are bandaids. Use the proper nic and it will be a none issue in the first place.

from perspective of truenas; passthrough refers to igb passthrough, not rtl.

Code:
Receiving

passthrough -     37-40%
virtio rtl  -     222%
virtio igb  -   72-75%

power

passthrough    92-94w
virtio rtl      142w
virtio igb      94-95w

------------------------------
Sending

passthrough -     20%
virtio rtl -     56%
virtio igb -     60%

power

passthrough    92w
virtio rtl    95w
virtio igb    94w
 
Top