• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Quick comparison, 1Gbps, 2.5Gbps, 5Gbps, 10Gbps Ethernet

TheLostSwede

News Editor
Joined
Nov 11, 2004
Messages
18,550 (2.48/day)
Location
Sweden
System Name Overlord Mk MLI
Processor AMD Ryzen 7 7800X3D
Motherboard Gigabyte X670E Aorus Master
Cooling Noctua NH-D15 SE with offsets
Memory 32GB Team T-Create Expert DDR5 6000 MHz @ CL30-34-34-68
Video Card(s) Gainward GeForce RTX 4080 Phantom GS
Storage 1TB Solidigm P44 Pro, 2 TB Corsair MP600 Pro, 2TB Kingston KC3000
Display(s) Acer XV272K LVbmiipruzx 4K@160Hz
Case Fractal Design Torrent Compact
Audio Device(s) Corsair Virtuoso SE
Power Supply be quiet! Pure Power 12 M 850 W
Mouse Logitech G502 Lightspeed
Keyboard Corsair K70 Max
Software Windows 10 Pro
Benchmark Scores https://valid.x86.fr/yfsd9w
Note that this test is limited by the fact that the target is a mechanical hard drive, even if it's a NAS drive. The NAS also has an Aquantia 10Gbps card in it.
This is obviously not a thorough test, but I wanted to see how the Realtek 2.5Gbps chip performed.
Take this for what it is, a quick test and nothing more/less.
I'm actually surprised at how decent Realtek's 2.5Gbps chip is. I guess they've stepped up their game quite a bit.

Intel 1Gbps

128914


Realtek 2.5Gbps

128915


Aquantia @ 2.5Gbps (limited to 2.5Gbps in the driver settings)

128913


Aquantia @ 5Gbps (limited to 5Gbps in the driver settings)

128912


Aquantia @ 10Gbps

128910
 
Last edited:
I just know a dedicated lan card helps if you need more speed than 100/1000 solutions lol. Also it's theorhetic speeds...
 
I just know a dedicated lan card helps if you need more speed than 100/1000 solutions lol. Also it's theorhetic speeds...

Why is it theoretic? Yes, there's some slight overheads due to how the protocols on top of Ethernet works, but on Gigabit you should be able to hit 980Mbps or more on a modern PC.
And why would a dedicated card help vs. integrated on the motherboard? It might've made a difference back in the day, but no so any more.

This is just a quick and dirty benchmark that gives people an idea what they can expect in terms of performance from a hard drive based NAS.
This is obviously not the maximum performance you can get.
 
Why is it theoretic? Yes, there's some slight overheads due to how the protocols on top of Ethernet works, but on Gigabit you should be able to hit 980Mbps or more on a modern PC.
And why would a dedicated card help vs. integrated on the motherboard? It might've made a difference back in the day, but no so any more.

This is just a quick and dirty benchmark that gives people an idea what they can expect in terms of performance from a hard drive based NAS.
This is obviously not the maximum performance you can get.

Resistance
 
I'm actually surprised at how decent Realtek's 2.5Gbps chip is. I guess they've stepped up their game quite a bit.
Intel 1Gbps
Realtek 2.5Gbps
Any chance you also measured CPU/power consumption? Realtek problem was that it used too much CPU during network traffic handling.
 
Any chance you also measured CPU/power consumption? Realtek problem was that it used too much CPU during network traffic handling.
This thread is almost four years old....
 
Note that this test is limited by the fact that the target is a mechanical hard drive, even if it's a NAS drive. The NAS also has an Aquantia 10Gbps card in it.
This is obviously not a thorough test, but I wanted to see how the Realtek 2.5Gbps chip performed.
Take this for what it is, a quick test and nothing more/less.
I'm actually surprised at how decent Realtek's 2.5Gbps chip is. I guess they've stepped up their game quite a bit.

Intel 1Gbps



Realtek 2.5Gbps



Aquantia @ 2.5Gbps (limited to 2.5Gbps in the driver settings)



Aquantia @ 5Gbps (limited to 5Gbps in the driver settings)



Aquantia @ 10Gbps
Do you recall which intel chip you were testing?
 
Helios LanTest = 17 Sep 2014 project.
Too old for my taste.
 
Did you monitor CPU usage during these tests?

After using my 13700k system daily I started noticing very high cpu usage linked the network throughput.

I then did some research which found multiple sources of information realtek nic's need higher cpu to process same through vs intel.

Comparing to my 9900k system though isnt apples to apples, however luckily I have a pcie intel gigabit card available.

Here is some data, in both cases the equivalent features are enabled on the network devices.

Realtek 2.5gbe vs Intel gigabit both running in 1000mbps full duplex mode.

2 RSS threads
Checksum offload fully enabled.
Interrupt moderation disabled.
Equal transmit and receive buffers.
Large send offload disabled.
VLAN disabled.
Power saving features disabled.
Jumbo frames disabled.
Flow control enabled.

% means percentage per logical core so this cpu e.g. can handle 2400% max. But load is only spread over core's equal to RSS threads, so in this case max 200% available. I did best to extract data used by the network itself, not the userland processes.

Realtek

100mbit 6% cpu usage
250mbit 23% cpu usage
500mbit 49% cpu usage
968mbit 112% cpu usage (requires 2 RSS threads to max out on a powerful 13700k)

Intel

100mbit 1% cpu usage
250mbit 3% cpu usage
500mbit 9% cpu usage
974mbit 21% cpu usage

The realtek can go up to max 4 RSS threads, vs only 2 on Intel, I now know why. Realtek network hardware seems very cpu heavy, even though it now has similar feature set at consumer level. Something must be done pretty badly in drivers or not optimised in hardware design?

Luckily my board has spare x1 slot for card, although its annoying as I prefer to use onboard. Wouldnt care if the difference was maybe only an extra 20% or so, but these numbers are astounding.
 
I have extensively tested different Realtek and intel chips, and have very different results. (It is dutch forum, but you can use translate button if screenshot is not clear)
I think there is your problem:
Interrupt moderation disabled.
Large send offload disabled.
Power saving features disabled.
I have all 3 enabled, especially interrupt moderation is the most important.

You did not specify what chips exactly you use, there are many different ones. Latest Realtek chips do not use CPU excessively.
Older intel chips (i210) have special setting “DMA coalescing” that needs to be enabled. It has amazing effect on CPU and power consumption.
@chrcoluk
 
I have extensively tested different Realtek and intel chips, and have very different results. (It is dutch forum, but you can use translate button if screenshot is not clear)
I think there is your problem:
Interrupt moderation disabled.
Large send offload disabled.
Power saving features disabled.
I have all 3 enabled, especially interrupt moderation is the most important.

You did not specify what chips exactly you use, there are many different ones. Latest Realtek chips do not use CPU excessively.
Older intel chips (i210) have special setting “DMA coalescing” that needs to be enabled. It has amazing effect on CPU and power consumption.
@chrcoluk
I did state the same settings on the intel nic, that also had those disabled.

For reference enabling interrupt moderation doesnt bring it down to the intel numbers, not even close.

Chip wasnt specified because I dont know the exact chip.

Checked hwinfo for you, it is a RTL8125 Gaming chip.
 
Last edited:
@chrcoluk Even with Intel chip, the cpu should not go higher than 6% at full load. Interrupt moderation is a feature that is specifically designed to reduce CPU usage and improve throughput at the same time. It is weird to not enable it.
The same goes for hardware offload - network chips can do all the processing much more efficient than CPU, so it does not make sense to not let them.
As I mentioned, i tested with a Realtek 2,5G chip, and did not observe such high numbers, they were slightly higher than Intels, but not much. It could be driver or settings or many other factors.
 
The same goes for hardware offload
This. With hardware offload off you are shooting youself in the foot. You are literally telling the OS not to offload work to the NIC. That’s said as Andru eluded too some chips differ greatly in not only hardware accel but drivers. I’m not sure why anyone would think anything “gaming” would be highly performant in anything network related.
 
This. With hardware offload off you are shooting youself in the foot. You are literally telling the OS not to offload work to the NIC. That’s said as Andru eluded too some chips differ greatly in not only hardware accel but drivers
I think I will give up here.

The point has been bypassed whether deliberate or by mistake.

So I will try to repeat the point once more.

Like for like settings the realtek is a cpu gobbler, the same offloading options are disabled on the intel.

I of course did try them enabled (with all the downsides that come with it, batch processing of packets, latency etc.) and the cpu load is still ridiculous.

Whilst realtek cards seem to be much more stable these days, its not hard to find people reporting high cpu usage, and even server vendors, as well as enterprise software vendors like truenas warning people away from them, so realtek are still not where are intel are. But they probably dont need to be, as being cheaper sells units, and many people will be ignorant of the differences, as fundamentally they do still work, just less efficiently.

I am on the latest driver available for download as well, however if I do find a resolution, I dont mind updating the community. Likewise if anyone has a suggestion I havent tried yet I will of course try it, but as i said I did try all the offloading enabled.
 
Like for like settings the realtek is a cpu gobbler, the same offloading options are disabled on the intel
Yes this is known. Intel nics are much better for network sensitive tasks this even reaches into the server space.

The one thing I will give Realtek and maybe it’s because of there lack of features is that the drivers are generally detected without issue on windows systems
 
Yes this is known. Intel nics are much better for network sensitive tasks this even reaches into the server space.

The one thing I will give Realtek and maybe it’s because of there lack of features is that the drivers are generally detected without issue on windows systems
Yep in terms of stability it has been fine, and although the default driver was old, it beats not having no network out of the box. I may end up just living with it anyway, isnt the end of the world.
 
It was true in the past with older Realtek chipsets/drivers. However their recent chip revisions have significantly improved - that is just an observation I am trying to convey. I was also surprised, tbh, how well Realtek chip performed compared to Intel.
Intel i210, i211 and i219 were extremely efficient and rock solid. With i225, i226, unfortunately, it changed - they are known to lose connection and consume more power.
If you really want to use onboard nic, why don't you try the best settings/drivers for it, because it can be efficient.
 
Yes this is known. Intel nics are much better for network sensitive tasks this even reaches into the server space.
That is, if the stupid thing can maintain a link. The Intel i225 and i226v Ethernet chipsets have been plagued by issues in which they drop connections for no good reason.
 
That is, if the stupid thing can maintain a link. The Intel i225 and i226v Ethernet chipsets have been plagued by issues in which they drop connections for no good reason.
Yeah iv seen your posts and thread about being in that hate train. This is off topic here though. I’d still take one over Realtek and those intel chipsets aren’t used in servers so….
 
Last edited:
Yeah iv seen your posts and thread about being in that hate train. This is off topic here though. I’d still take one over Realtek and those intel chipsets aren’t used in servers so….
I don't know about taking a defective one over realtek but I have one of the revised ones on my file server as well as a dual port X540-T2. Both do fine.
 
I’d still take one over Realtek and those intel chipsets aren’t used in servers so….
Which Intel NICs are used in servers?

Which now begs the question... If Intel can make a good server-grade Ethernet controller, why can't they make a good consumer-grade Ethernet controller?
 
Back
Top