Realtek Develops New Onboard LAN Chip to Win Back Gamer Crowd

[silentbogo]

But the second thing... Wi-Fi drivers on Linux for Intel were not the smoothest ride for me also. But again... not as horrid as Realtek one lol... albeit nothing beats Atheros driver mess in wifi department

Don't forget to throw a tomato at RaLink[MediaTek]. I'm still furious that my Asus PCE N53 "upgrade" only works in Windows and [barely] Ubuntu 12....

 

[Jstn7477]

If only they would work on their laptop/tablet wireless chips that everyone and their mom seem to use. I've had to put an Intel 7260 in my latest laptop because the Realcrap 8723AE would drop my Microsoft Sculpt mouse a trillion times a day and completely shut off on battery for no reason, and the 8723BS SDIO in my Onda Windows tablet has terrible throughput and drops all the time even with Bluetooth off. I've resorted to using a USB dongle on my tablet (still Realtek though) because of course you can't even poke 5GHz networks with plebtek to eliminate Bluetooth interference since they are chill with using one antenna for both as well...

 

[Tomorrow]

Intel also ain't saint, my 82579V is plagued with various issues...

But... this... is this a rebrand? They just smelled the late trend

The pics look horrid... the soldering flux isn't washed off, almost looks like done by hand...

https://communities.intel.com/thread/54594
https://communities.intel.com/thread/43498

...and intel has still not fixed it. FYI the threads are ~two years old!

 

[FordGT90Concept]

My problem with Realtek is the generic drivers they offer for their chips often don't work very well. On the other hand, Intel's drivers typically work no matter what OEM integrated the chip. If Realtek stops farming out driver development to OEMs and provides a catch-all driver, I think acceptance of their products in DIY would improve. Additionally, why can't we get NICs and audio controller chips that are all-in-one (interface, logic, and RAM)? You know, all of the performance advantages of a dedicated card but embedded in a motherboard solution.

I'm glad things are improving but the tiny incremental steps are not satisfactory in my opinion.

 

[Ferrum Master]

https://communities.intel.com/thread/54594
https://communities.intel.com/thread/43498

...and intel has still not fixed it. FYI the threads are ~two years old!

I know, I can trigger the DPC by using any proset driver still, the only one that works is that comes with WIN10 inside... it is fine. Latency mon doesn't show anything bad, otherwise ndis.sys put up the stutter fest around 6000us

 

[Easy Rhino]

Huh? Do any "gamers" out there choose their motherboard based on what network interface it has?

 

[Disparia]

Huh? Do any "gamers" out there choose their motherboard based on what network interface it has?

It will be a deciding factor when all items of a higher priority are equal.

Today's Realtek has certainly improved a great deal from 10-15 years ago, enough to where I don't mind using one of their NICs if the rest of the board meets my needs. However when it comes down to it or another, they're probably going to lose out. Case in point: There were plenty of AM3 boards for me to choose from and I happen to get one with a pair of Broadcom NICs

 

[Aquinus]

The RTL8167 on my motherboard is solid enough to support getting bonded to the Intel adapter. Also the RTL8111 in my gateway seems to work fine. In all seriousness, I have both the RTL8111 and a dual port Intel 1000 PT adapter in my gateway and you really wouldn't notice a difference between the two, to be honest. Granted Linux is a little different where most common Ethernet adapters seem to work pretty well ootb.

How about some affordable 10Gbe adapters? I would be more interested in that.

 

[ZeDestructor]

I know, I can trigger the DPC by using any proset driver still, the only one that works is that comes with WIN10 inside... it is fine. Latency mon doesn't show anything bad, otherwise ndis.sys put up the stutter fest around 6000us

Interesting.. I've seen ~2100µs max (500-600µs average) on my laptop (Dell M4600) with an 82579LM controller (also an Intel 7260AC) for ndis.sys. nvddlkm on the other hand will happily hit 14k, though that may be related to it compensating for the overclocked GPU. Shiniest Intel 20.1 drivers on Windows 10 here, with all the dynamic power bits enabled. I'd test my desktop too, but I'm a bit far from it right now, so...

How about some affordable 10Gbe adapters? I would be more interested in that.

If you're willing to invest in fibre/TwinAx, SFP+ adapters are rather affordable nowadays. (only around $75 per port). The bigger cost is the switch, really

 

[Ferrum Master]

. Shiniest Intel 20.1 drivers on Windows 10 here, with all the dynamic power bits enabled. I'd test my desktop too, but I'm a bit far from it right now, so...

I did a retest... with 20.1 DPC spikes around 6000µs, using the windows shipped driver everything is ok...(it is actually higher number than the 20.1). There are reports that 19.1 work fine too... Quite disturbing to be honest...

 

[Frick]

In my last move I left 100+ old realtek NIC's to goodwill. I don't think they were happy with me.

How about some affordable 10Gbe adapters? I would be more interested in that.

Yes, this.

 

[Aquinus]

If you're willing to invest in fibre/TwinAx, SFP+ adapters are rather affordable nowadays. (only around $75 per port). The bigger cost is the switch, really

I was thinking more 10Gbps on copper. A drop in replacement for networks already utilizing CAT-6 cabling.

 

[FordGT90Concept]

There's little agreement in the industry for 10GbE. CAT6 I think can only handle 2GbE over about 50 feet; CAT6A can handle 2GbE over the normal 300 feet. CAT8 will be 10GbE and CAT8 cables will be ridiculous compared to CAT6 (each pair will probably have a CAT6-like divider and have its own jacket with a divider separating all four pairs, also jacketed). CAT8 will probably be the limit for standard copper cables. I don't know if radio guide can even exceed the bitrate of CAT8 but if it can, we might be moving back to coaxial.

The NICs aren't the problem with >CAT6, it's the cable. CAT6 FTP is already considerably more than CAT5 and the price of CAT6 isn't going to come down much/any more. CAT8 will be prohibitively expensive (at least $0.50/foot).


I think what we need is a fiber push to consumers.

 

[Aquinus]

CAT6 I think can only handle 2GbE over about 50 feet

10Gbit at 55 meters for CAT-6, 10Gbit at 100 meters for CAT-6a, depending on the installation.

Class E channels assessed and mitigated according to the guidelines in this Technical Report are expected to support 10GBASE-T to distances from 55 m to 100 m using unscreened Category 6 components.

https://webstore.iec.ch/preview/info_isoiec24750{ed1.0}en.pdf

Category 6 cables can carry 10GBASE-T for shorter distances when qualified according to the guidelines in ISO TR 24750 or TIA-155-A.

Above doc references ISO TR 24750. I think it's supported.

I think what we need is a fiber push to consumers.

I won't disagree with that but, things like PoE are nice and you're not getting that with optical. I can see there being reasons to hold on to copper but, not for a connection to the interwebs.

 

[Frick]

There's little agreement in the industry for 10GbE. CAT6 I think can only handle 2GbE over about 50 feet; CAT6A can handle 2GbE over the normal 300 feet. CAT8 will be 10GbE and CAT8 cables will be ridiculous compared to CAT6 (each pair will probably have a CAT6-like divider and have its own jacket with a divider separating all four pairs, also jacketed). CAT8 will probably be the limit for standard copper cables. I don't know if radio guide can even exceed the bitrate of CAT8 but if it can, we might be moving back to coaxial.

The NICs aren't the problem with >CAT6, it's the cable. CAT6 FTP is already considerably more than CAT5 and the price of CAT6 isn't going to come down much/any more. CAT8 will be prohibitively expensive (at least $0.50/foot).

Cat6 = 10GbE 55m, Cat6A = 10GbE 100m, 500mhz as opposed to 250Mhz. And for prices, 100m cat6 is about €15 more than 100m cat5e, I wouldn't call that considerably more. Did a quick lookup now, and at least in Sweden I would definitely say it's pointless to buy cat5e instead of cat6, unless you're wiring up an entire house and you for some bizarre reason simply cannot afford the extra few hundred euros cat6 cables would cost you.

 

[FordGT90Concept]

I'm running all CAT6 here, some FTP and some UTP.

It looks like 10 GbE cards start well north of $100. 4-port unmanaged 10 GbE switches are $400'ish (figure $300 if it were embedded in a consumer router--just for the switch component). The price is coming down but not very fast.

 

[Uplink10]

I won't disagree with that but, things like PoE are nice and you're not getting that with optical. I can see there being reasons to hold on to copper but, not for a connection to the interwebs.

PoE is very nice but most people I know do not have the devices that support PoE or are not using it but then they do not have the latest hardware and tech.

I think that we already have a big problem in households and that problem is with voltage in outlets, a lot of devices have transformers and they transform it to below 24 V for most devices and in outlets the voltage is 110 V in USA and 230 V in Europe. And now we are also using LED and energy efficient bulbs which require a lot lower voltage and that means that a lot of the devices are using a lower voltage and we would not need so powerful transformers if there were outlets with voltage a lot lower.

Now with such lower voltage it would be easier to have fiber connection and more low voltage outlets for devices like switches or you can just get something like fiber/copper hybrid.


10 GbE could be used instead of load balancing with more than 1 GbE ports. For example you have a NAS with one 10 GbE port which is connected to a 10 GbE switch but clients are connected with only 1 GbE connection.

 

[ZeDestructor]

I won't disagree with that but, things like PoE are nice and you're not getting that with optical. I can see there being reasons to hold on to copper but, not for a connection to the interwebs.

Why not just have copper power lines along the fibre strand? It's already being done in some places, like undersea cables.

 

[FordGT90Concept]

I think Thunderbolt supports that.