ASUS releases first sub-100 dollar 10GbE card

[AnonymousGuy]

1 minute ago, Ryan_Vickers said:

I think you do

RAMdisk

Oh you sly bastard.

 

I'll run some tests with what I have to see if I can saturate it with proper cabling first.

[AnonymousGuy]

30 minutes ago, Ryan_Vickers said:

I think you do

RAMdisk

I only got about 500 mbyte/second.  Something is wrong.  Need to try putting the 10GbE card in my HTPC which is dramatically more powerful than my storage server.  Maybe a dual core 3Ghz chip can't handle 10Gb.

[vanished]

7 minutes ago, AnonymousGuy said:

I only got about 500 mbyte/second.  Something is wrong.  Need to try putting the 10GbE card in my HTPC which is dramatically more powerful than my storage server.  Maybe a dual core 3Ghz chip can't handle 10Gb.

That would be interesting.  I have heard that with high enough speed CPU power does actually come into play, sometimes significantly.

[AnonymousGuy]

22 minutes ago, Ryan_Vickers said:

That would be interesting.  I have heard that with high enough speed CPU power does actually come into play, sometimes significantly.

Got it.  I just maxed out Tx and Rx buffers and it seems to have fixed the problem...weird:

 

 

*sigh* 

 

Alright time to buy some cat5e cable and see if 100 feet works.

[leadeater]

1 hour ago, mr moose said:

I'm well aware of marketing BS,  I was more interested in something more like how we get GPU review benchmarks but where some dude runs a whole heap of cat 5 and compares it to whole heap of cat 6 in different lengths etc. 

So you looking more for length testing to see how consistent Cat5e works for 10GbE? That is something I might be able to do, I have the proper testing equipment for it i.e. Fluke cable analyser and enough different brands of 10GbE NICs to cross verify that too just not sure if I have any Cat5e left. I think our entire supply is Cat6a now.

 

Not sure it's worth the hassle though because environment/noise is really the only factor that is going to make any measurable difference, if one Cat5e cable works without errors at say 30m all cables will in the same environment.

 

Personally I have never seen a difference between any cables that were undamaged, the manufacturing of them doesn't have that much variance. This might be only what I have observed but I have done a large amount of cabling work over the years so if it was something that had a noticeable impact I'm sure I would have encountered it.

 

Edit:

Shitty installation of cabling I have seen though, never get an electrician to do data cabling always get a data cabler.

[Helly]

i recently bought some 10Gbit stuff (x99-e-10G WS and Asus PEB-10G/57840-2T) and actually only had cat5e (20meter). I could only use it with HDD's though. Even though i reached 500 to 600MB/s with local transfers i only managed to get 400 to 450 through the network. Not sure where the problem lies but it could be the cable (so i ordered a cat6a cable) or the fact 1 card is running on pci-e 2.0 while the card is 8x 3.0. Not sure if it should matter as 8x should be well beyond the 20Gbit (dual port card) it can use right?

 

21 minutes ago, AnonymousGuy said:

Got it.  I just maxed out Tx and Rx buffers and it seems to have fixed the problem..

you mind telling me where to find those buffers? I might wanna try that to.

 

Also not sure i like this news after just having bought a €240 card from asus

[leadeater]

2 minutes ago, Helly said:

you mind telling me where to find those buffers? I might wanna try that to.

Depends on the card but go to the Properties on the network connection/NIC then click Configure on the dialogue box that comes up, it'll be near the top.

 

 

 

The advanced tab will look different depending on the card you have.

[paradigm249]

34 minutes ago, AnonymousGuy said:

Got it.  I just maxed out Tx and Rx buffers and it seems to have fixed the problem...weird:

Spoiler

 

 

 

*sigh* 

 

Alright time to buy some cat5e cable and see if 100 feet works.

got your IP, im gonna steal your data

/s

[AnonymousGuy]

1 hour ago, Helly said:

you mind telling me where to find those buffers? I might wanna try that to.

 

Also not sure i like this news after just having bought a €240 card from asus

With Intel's X540-T2 it's Properties -> Advanced -> Performance Options -> Receive + Transmit buffers. 

 

I also disabled Interrupt Moderation (Advanced), enabled jumbo frames (Advanced), interrupt moderation rate = off (Performance Options), and receive side scaling queues to 16.  Everything else I left alone.  Flow Control = Rx+Tx enabled.

 

 

And watch out for certain firewalls.  Kaspersky 2016 had a "bug" where the firewall couldn't handle more than about 2gbps.  They fixed it in 2017.  In 2016 I could toggle the Firewall on and off and watch my speed get cut in half.

[mr moose]

1 hour ago, leadeater said:

So you looking more for length testing to see how consistent Cat5e works for 10GbE? That is something I might be able to do, I have the proper testing equipment for it i.e. Fluke cable analyser and enough different brands of 10GbE NICs to cross verify that too just not sure if I have any Cat5e left. I think our entire supply is Cat6a now.

 

Not sure it's worth the hassle though because environment/noise is really the only factor that is going to make any measurable difference, if one Cat5e cable works without errors at say 30m all cables will in the same environment.

 

Personally I have never seen a difference between any cables that were undamaged, the manufacturing of them doesn't have that much variance. This might be only what I have observed but I have done a large amount of cabling work over the years so if it was something that had a noticeable impact I'm sure I would have encountered it.

 

Edit:

Shitty installation of cabling I have seen though, never get an electrician to do data cabling always get a data cabler.

Yeah, that's more what I'd like to see.  Someone get a 20 metre patch cable and test it at home down the hallway then at work, maybe under a fluro light or LED dimmed down light. 

[tlink]

8 hours ago, Grinners said:

How much bandwidth does 4k streaming use? 

that depends on the bitrate and compression algorithm. technically any network can stream 4k, just not at an acceptable quality.

[Grinners]

7 minutes ago, tlink said:

that depends on the bitrate and compression algorithm. technically any network can stream 4k, just not at an acceptable quality.

My point was more that 4k streaming would surely not utilize more bandwidth than a 1g/b connection! 

[vanished]

10 hours ago, AnonymousGuy said:

Got it.  I just maxed out Tx and Rx buffers and it seems to have fixed the problem...weird:

 

 

*sigh* 

 

Alright time to buy some cat5e cable and see if 100 feet works.

Damn that's fast!  

[The Benjamins]

8 hours ago, Grinners said:

My point was more that 4k streaming would surely not utilize more bandwidth than a 1g/b connection! 

BluRay bit rate is 40Mbps (h264) with UHD BD is 80Mbps (h265). so at UHD BD quality with h264 it would be 160Mbps. so in short you should be able to stream that locally.

[vanished]

19 minutes ago, The Benjamins said:

BluRay bit rate is 40Mbps (h264) with UHD BD is 80Mbps (h265). so at UHD BD quality with h264 it would be 160Mbps. so in short you should be able to stream that locally.

Yes, easily, hence my comment wondering what this article is on about talking about an upgrade to this being good for movies.  I gave an example above that was extreme but reasonable, but for fun now lets do one that's just totally out there to show how usable 1 Gb actually is:

 

Take a 240 GB movie and watch that over the course of 45 minutes 

 

That's 91 MB/s, getting close but still well within what 1 Gb ethernet can do.  I don't think any sane person can make any argument whatsoever for 10 Gb being necessary for streaming video, unless they mean uncompressed raw footage into your editor  

[The Benjamins]

2 minutes ago, Ryan_Vickers said:

Yes, easily, hence my comment wondering what this article is on about talking about an upgrade to this being good for movies.  I gave an example above that was extreme but reasonable, but for fun now lets do one that's just totally out there to show how usable 1 Gb actually is:

 

Take a 240 GB movie and watch that over the course of 45 minutes 

 

That's 91 MB/s, getting close but still well within what 1 Gb ethernet can do.  I don't think any sane person can make any argument whatsoever for 10 Gb being necessary for streaming video, unless they mean uncompressed raw footage into your editor  

the reason for 10Gbps would be for a NAS or server so you can have many people hitting it at the same time. (2.5Gbps, 5Gbps would be more what I would recommend if it was cheaper)

 

but ya most people do not need more then 1Gbps. I personally just would like my server to be faster.

[vanished]

4 minutes ago, The Benjamins said:

the reason for 10Gbps would be for a NAS or server so you can have many people hitting it at the same time. (2.5Gbps, 5Gbps would be more what I would recommend if it was cheaper)

 

but ya most people do not need more then 1Gbps. I personally just would like my server to be faster.

Yeah that would be one actual use, not streaming videos

But I already had my rant about using it for a NAS too:

15 hours ago, Ryan_Vickers said:

First, let me just say it's great to see 10 Gb hardware getting cheaper, but I have to say a few things, first about the article, and then the product itself.

 

Uh, no.  There is no shortage of NVMe SSDs that can perform faster than this connection by quite a bit, so no it's not "just what you need", it's "about 2/3^rds of what you need".  Furthermore, who the hell has a NAS built out of NVMe SSDs?

 

Uh, yes, and you can already do that easily with 1 Gb ethernet.  A 50 GB bluray file streamed over the course of 1.5 hours needs less than 10 MB/s, under 1/10 of what's already doable.

 

Can they seriously not come up with one valid use for it?  (This is where I transition to my problem with the product itself.)

 

Well luckily @TheRandomness did  

 

It's for those people in Sweden/Denmark/etc. that get 5 gigabit internet access.  

 

For everyone else, it's just not worth it.  If you have a NAS, sure, this will be faster than what you've got, but are you really gonna pay $99 per machine, plus whatever it takes for all new CAT 6A cables, plus 1 or more 10 Gb switches just to less than double your speed (assuming you've got HDDs in your NAS)?  If so, good on you, but I suspect the answer is no, and that's the enthusiast end of the market.  For the other 99% of people, they simply don't transfer files often enough or with enough urgency to make it worth it.

[The Benjamins]

14 minutes ago, Ryan_Vickers said:

Yeah that would be one actual use, not streaming videos

But I already had my rant about using it for a NAS too:

Ya I kinda wish 2.5Gbps and 5Gbps was more of a thing that we could use as a stepping stone.

[Juanitology]

19 hours ago, SansVarnic said:

Please adjust your post for the dark ones here.

This is what I see;

  Reveal hidden contents

Thanks.  

not sure how to fix it

[Juanitology]

10 hours ago, Grinners said:

My point was more that 4k streaming would surely not utilize more bandwidth than a 1g/b connection! 

Not consistently, no. But when you're starting the stream and buffering, it can mean the difference between instant playback and waiting 10-15 seconds for the stream to buffer. I know I know 1st world problems..

[paddy-stone]

20 hours ago, Ryan_Vickers said:

Uh, yes, and you can already do that easily with 1 Gb ethernet.  A 50 GB bluray file streamed over the course of 1.5 hours needs less than 10 MB/s, under 1/10 of what's already doable.

 

 

Agree, already have done this many times with multiple streams concurrently too. It's a great way of testing network or NAS throughput IMO, real world testing I mean of course.

 

I love that higher bandwidth NICs are now starting to be more affordable though, not sure about it requiring 4 PCI-e lanes though, 2 should suffice shouldn't it? can't be arsed to do the math right now, lol.

[vanished]

Just now, paddy-stone said:

Agree, already have done this many times with multiple streams concurrently too. It's a great way of testing network or NAS throughput IMO, real world testing I mean of course.

 

I love that higher bandwidth NICs are now starting to be more affordable though, not sure about it requiring 4 PCI-e lanes though, 2 should suffice shouldn't it? can't be arsed to do the math right now, lol.

4 lanes of PCIe 2.0 should be 2 GB/s, so yes, more than enough, but perhaps they didn't want to cut it so close with just 2

[paddy-stone]

Just now, Ryan_Vickers said:

4 lanes of PCIe 2.0 should be 2 GB/s, so yes, more than enough, but perhaps they didn't want to cut it so close with just 2

Ahh yeah, didn't think about that... thanks for clearing that up for me.

[paddy-stone]

2 hours ago, Terodius said:

Not consistently, no. But when you're starting the stream and buffering, it can mean the difference between instant playback and waiting 10-15 seconds for the stream to buffer. I know I know 1st world problems..

If it's buffering for that long it's device problems IMO (NAS/PC/tablet). I haven't had ANY noticeable buffering <2 seconds when doing multiple 4K streams even when using wired devices, including android 4k boxes. Even wifi should be able to deliver 4k streams fairly well, with a little buffering, I would think. There ya go, will have to test this when I get some time

 

[vanished]

1 hour ago, paddy-stone said:

If it's buffering for that long it's device problems IMO (NAS/PC/tablet). I haven't had ANY noticeable buffering <2 seconds when doing multiple 4K streams even when using wired devices, including android 4k boxes. Even wifi should be able to deliver 4k streams fairly well, with a little buffering, I would think. There ya go, will have to test this when I get some time

 

Yeah, if you consider a 50 Mbit stream, which is very high quality, a gigabit link could buffer up 20 seconds of video in 1 second of actual time (best case scenario) so I can't see that being an issue.