Talk:10 Gigabit Ethernet/Archive 1

For any "nontrivial distance"?
This article says

Unlike earlier Ethernet systems, 10-gigabit Ethernet (for any nontrivial distance) is so far based entirely on the use of optical fibre connections

although there is 10GBASE-CX4 -- Copper interface using InfiniBand CX4 cables and InfiniBand 4x connectors for short-reach (15 m maximum) applications mentioned... copper implies copper, not fiber.... so is there a mistake? --Abdull 12:12, 17 Mar 2005 (UTC)


 * The first 10GbE IEEE spec (802.3ae) did not support any copper (other than MAC to PHY interconnects). CX4 was added later in IEEE P802.3ak.  You can also run XAUI for about 20 inches on copper or printed circuit board.  I am not not sure what the "nontrivial distance" comment is referring too because CX4 can be 25m, so I assume that comment was referring to XAUI and is no longer true in light of CX4 and will definitely need deleting when 10GBASE-T is ratified.  Patrick0101 20:33, 19 May 2005 (UTC)


 * Non-trivial means greater than the reach of CX4. I think this paragraph should be changed to "Unlike earlier Ethernet systems, 10 gigabit Ethernet was initially based mostly on the use of optical fiber connections.  More recent developments introduced 10GBASE-CX4 for links up to 25m over Infiniband cable; and work is progressing on 10GBASE-T to allow links up to 100m over UTP."  Thoughts? Andybryant 19:46, 16 June 2006 (UTC)

Clock/Bit/Baud Rate?
I searched everywhere for the clock rate of the 10GbE. I've just found in a datasheet of one transceiver vendor: 10.3125 Gbps, but I don't know how and where to put it in the text... Akira - Cleber Akira Nakandakare 12:15, 23 May 2005 (UTC)


 * This is the bit rate for the LAN PHY varient of 10GbE. The 10GBase-xW varients run at the SDH/SONET rate of 9.953 Gbps.  I've made this change to the page.  The only thing I'm not sure about is if LX4 is LAN PHY or not. Andybryant 19:46, 16 June 2006 (UTC)

The baud rate for each lane of CX4 is 3.125 GHz implying that with four lanes it's possible to transmit 12.5 Gbps though I know a large portion is lost by the 8b10b Clause 48 PCS conversion meaning ideal spead would be 10Gbps (8/10 * 12.5).

10GBASE-T Approved!
PDF Warning - Check it out: http://grouper.ieee.org/groups/802/minutes/jul2006/Minutes%20-%20Monday%20July%2017%202006.pdf

10GBASE-T has been approved! They can start officially making the devices! Yay! --UNHchabo 05:00, 28 July 2006 (UTC)


 * Note: According to that article, the standard will be published on 15 Aug 2006. --UNHchabo 05:01, 28 July 2006 (UTC)

unjustified revert
Some-one (no user account) has just reverted my correction of 'faster' to 'as fast' with no explanation of why. Is this vandalism? T23c 13:33, 27 October 2006 (UTC)
 * Probably not, it may just be ignorance. Try reverting back, with a pointer to the talk page where you explain why the use of 'as fast' is better. WIth luck, if the anonymous editor disagrees with you, you can then debate the issue.WLD 13:40, 27 October 2006 (UTC)
 * It is my mistake actually. I put 'as fast than' by mistake, and he must have not seen the history. I have changed it to 'as fast as' now. T23c 14:56, 27 October 2006 (UTC)

Implementation
Is the "implementation" section really necessary? 10gig devices have been on the market for a few years now, plenty of manufacturers make them. Seems like a ripe place for spam. --UNHchabo 07:40, 20 February 2007 (UTC)
 * 'tis gone now -- KelleyCook 21:39, 20 February 2007 (UTC)

CX4 form factor availability
CX4 modules are now available in XPAK (from molex) and XFP (from emcore) form factors. Ottawa matt 00:14, 24 April 2007 (UTC)

WARP -- delete?
The WARP section looks like it was originally created as an advertisement for the company, by the same author who created the article about the company. It isn't quite so bad after rewriting, but does it need to be here at all? The funny shield is interesting, but the claim that this "enables" 100 meter links is clearly invalid, since others have demonstrated 100 meter 10GbaseT links without it.

I'd suggest the best answer might be simply to delete it. If there's anything notable to be said, it can then be brought up. Trying to take what started as an ad and making it "better" isn't necessarily a good way to arrive at a quality result. Paul Koning 01:13, 23 May 2007 (UTC)

Real-world Uses and applications?
It would be nice to have a section of this article discussing the uses and applications of 10 gigabit Ethernet.

As far as I know, no disk storage technology is currently capable of sustaining transfer at such speeds, and the top-end disk bus standards are not anywhere near the speed of 10 gigabit. SATA 3.0 has a burst rate of 3 gigabit, but once again no storage technology can even sustain that except via cached reads and writes.

The only thing that could possibly saturate at 10 gigabit link would be a solid state drive, and those are still extremely uncommon due to the immense cost.

Most people I talk with say that a 1 gigabit backbone is still reasonable for most applications, and from my own tests I have found that file servers often have a hard time saturating even a 1 gigabit link. What with the continued popularity of 100 megabit switches with 1 gig uplinks, that seems to be the case.

So who is using 10 gig-E, and for what purpose? I can see it being installed as merely a "huge central pipe" which is fed from a farm of servers, but it seems unlikely to be used to directly link one individual server to a switch, and it also seems unlikely for any one server to be capable of sustained speeds in excess of even 5 gigabit.

Some edumacation for those of us not operating at the elite cusp of the 10 gig-E technological revolution would be helpful.

DMahalko (talk) 03:44, 5 June 2008 (UTC)

I am not an expert in the application space but I will make a couple of observations:

1) You do not need to use the full 10G to find 10G useful. If you need more than 1G then you arguably have a need for 10G Ethernet.

2) There is a trend towards consolidation and virtualization in the data centre. If you have network attached storage that is used by work group doing engineering or graphics applications you can imagine how a 1G pipe to the disk server is not going to be adequate. —Preceding unsigned comment added by 77.97.110.196 (talk) 14:45, 28 June 2008 (UTC)


 * Clusters. ~ Jafet Speaker of many words 19:16, 16 August 2008 (UTC)


 * Storage chassis. For example http://www.coraid.com/products/sr2461  —Preceding unsigned comment added by Quanstro (talk • contribs) 13:57, 26 October 2008 (UTC)

There are many applications for 10G networking. It is not new to the computing world and has existed for quite some time in it's fiber form. It is used primarily as a backbone network link but now is moving out to connect the servers to the network. The new part is the standardization and the on-going perfection of a copper variety of physical interface which prommises to reduce the cost of adoption of 10G accross many different platforms. While the average cost of fiber continues relatively high compared to copper (eg. 1G ethernet and now 10G follows this model) there has been much made of the size and power consumption of the copper interface. With the advent of 40nm silicon technologies Aquantia's quad chip (4/2009)reduces the power to 3.5W per port at 100M distances or low power mode of just 2 watts. This is more than enough to drive the port density up to the point that commercial switches will now start to appear in volume. In 2008 Intel's dual port NIC was priced around $500US and there has been a number of advances in functionality with server Nic's to make them attractive. This is a very competitive price port compared to any fiber based NIC on the market.

The advent of 10GBase-T Lan on Mother board (LOM) is now predicted to occur in 2012 for servers at least. This would mark an important step in gaining wide based acceptance and further reducing per port costs. The question on 10G networking is not if, it is when. —Preceding unsigned comment added by 198.90.65.175 (talk) 09:33, 2 June 2009 (UTC)

Market share
What is the real-world breakdown of which 10GE technology is deployed?

Eg: 10GBASE-SR (5%), 10GBASE-LR (4%), 10GBASE-CX4 (15%), ...

Much interested, thank you Full Decent (talk) 15:53, 17 September 2009 (UTC)

>> This is an interesting question to which I don't know the answer. You could take a look at the HP Procurve website and see what ports HP offer on their 10G switches. I expect there is a lot less 10GBASE-CX4 than there was and a lot more SFP+. — Preceding unsigned comment added by 216.21.32.60 (talk)

Compatibility ?????
With all the forms of MMF is there a definitive chart outlining compatibility? And for that matter is the new 10 Gig fiber compatible with older laser and or LED transceivers? —Preceding unsigned comment added by 68.90.31.2 (talk) 13:20, 7 April 2010 (UTC)

Dead link 2
During several automated bot runs the following external link was found to be unavailable. Please check if the link is in fact down and fix or remove it in that case!


 * http://lw.pennnet.com/display_article/249488/13/ARTCL/none/none/1/10GBase-LX4-vs-10GBase-LRM:-A-debate/
 * In 10 Gigabit Ethernet on 2011-05-23 02:08:01, 404 Not Found
 * In 10 Gigabit Ethernet on 2011-05-31 22:29:21, 404 Not Found

--JeffGBot (talk) 22:29, 31 May 2011 (UTC)

wan phy edits
Previously at the end of the wan phy section there was the sentance " Over time, the local area options became more popular. ". 213.131.238.28 changed this to "The WAN PHY is a niche device and is rarely used." (removing the ref in the process) and then 209.87.247.18 removed it completely.

Any thoughts on what should happen with this sentance? should it be put back the way it was before the IP edits? should it be left removed? can a cite be found for the "rarely used" claim by 213.131.238.28 ? Plugwash (talk) 08:09, 9 September 2011 (UTC)


 * First, it would be nice if contributors would leave edit comments. Second, it looks like removing that sentence was the right thing to do. The claim is not supported in the "cagr" reference. --Kvng (talk) 20:25, 11 September 2011 (UTC)

powered Ethernet and energy-efficient ethernet
What is going in with integrating the 802.3at powered Ethernet standard at 10G (it's presently at 1G) and energy-efficient Ethernet (which organizes data more efficiently into bursts so the transmitting and receiving circuitry spends most of its time off. These are extremely important standards especially as one-wire power-and-data cabling has become the norm (USB, powered Ethernet 1G) for whole classes of devices including VoIP, fixed wireless access points and rackmount devices (using PoE for these gets AC power supply out of the rack and often out of the room where it can be more efficiently cooled).  For those applications where 10G is not a profound gain over 1G because of bottlenecks, a system integrator will be wise to choose 1G at present for the flexibility of the powered Ethernet standard.  What we want from this article is when is this going to change, if ever, so that we aren't giving up 10G to get 802.3at power? —Preceding unsigned comment added by 142.177.112.142 (talk) 22:24, 11 December 2009 (UTC)
 * I suspect the simple answer is noone cares too much at this stage (though from the links in our energy efficient ethernet article it looks like they are working on energy efficient ethernet at 10G and may even have standardised it). The main application of energy efficient ethernet would seem to be desktops and similar low usage devices that are unlikely to get 10 gigabit any time soon though I could see some value for servers with light usage patterns. The main applications of POE are VOIP phones and wireless access points (I find your claim of using it to power rackmount hardware unlikely) which are unlikely to need 10 gigabit ethernet any time soon (if ever). Plugwash (talk) 10:52, 13 September 2011 (UTC)

Assumptions and other silliness.
For some odd reason someone wrote that LX4 is obsolete without citation. There is also an entry that states LRM is a market failure with a citation that indicates nothing more than LRM isn't known to be used in servers. — Preceding unsigned comment added by 163.150.10.201 (talk) 00:46, 7 February 2012 (UTC)

SFP+ direct attach
an anon recently renamed the SFP+ direct attach section to 10GBASE-CR. Should these edits be reverted? Plugwash (talk) 00:36, 15 January 2013 (UTC)


 * Yep, as far as I can tell 10GBase-CR doesn't actually exist as an IEEE standard, but I'll look into this more closely. I've found a reference that states "Some vendors incorrectly call these 10GBase-CU or 10GBase-CR cables, but there is no IEEE standard at this time." --Rwh86 (talk) 11:02, 15 January 2013 (UTC)


 * See IEEE 802.3-2008 44.1.1: "Physical Layer entities such as 10GBASE-SR, 10GBASE-LX4, 10GBASE-CX4, 10GBASE-LRM, 10GBASE-LR, 10GBASE-ER, 10GBASE-SW, 10GBASE-LW, 10GBASE-EW, and 10GBASE-T." – no mention of anything -CR (Cost Reduced?). Zac67 (talk) 19:52, 15 January 2013 (UTC)


 * "-CR (Cost Reduced?)" Following the IEEE MAU/Phy naming convention of nTYPE-LLLm-Eo (which even they do not strictly follow) 10GBASE-CR would be 10 Gigabit on twin axial copper with scrambled 64b/66b (en)coding. That is an accurate description of SFP+DA, albeit not an IEEE standard. See Slide 5. Zero10one (talk) 03:42, 17 January 2013 (UTC)

Move from '10-gigabit Ethernet'
Wouldn't it be "10 gigabit Ethernet" instead of "10 Gigabit Ethernet" as per WP:MOSNUM? Might as well do another move while at it. -- intgr 23:31, 10 August 2006 (UTC)


 * Well, noone has opposed so far, I'm going to do the move. -- intgr 12:26, 23 September 2006 (UTC)


 * One does not normally capitalize gigabit. So unless we consider our subject to be a proper noun, 10 gigabit Ethernet is the correct title. I've already changed body text to reflect this. I intend to submit a WP:RM to change the title. --Kvng (talk) 18:47, 7 December 2010 (UTC)


 * Normally you'd also want to hypenate a compund adjective as in 10-gigabit interface. So from a purely gramatical perspective 10-gigabit Ethernet might be most correct. --Kvng (talk) 18:55, 7 December 2010 (UTC)


 * See Talk:100_Gigabit_Ethernet for further discussion of Ethernet naming conventions. ~Kvng (talk) 22:14, 3 April 2015 (UTC)

10 GbE NICs section
IMHO, the section can be deleted after being challenged for a while and it should. The provided information is trivial and the article doesn't benefit from it. It's completely sufficient to mention NIC in the lead. --Zac67 (talk) 17:42, 18 September 2015 (UTC)


 * Agreed per WP:V and WP:BURDEN; the burden of proof is on the uploaders and reuploaders, not those wanting to remove uncited or dubious content. It's implied that such content is subject to removal—even without warning—by the disclaimer at the top of all mainspace edit pages: "Encyclopedic content must be verifiable." I'll remove it again. – void  xor  21:39, 18 September 2015 (UTC)


 * has removed "10GbE network interface cards are available from several manufacturers. These plug into ordinary computer servers using PCI express and provide one or more PHY module, LC or 8P8C connectors." and reverted my restore of the material citing WP:BURDEN. A citation is necessary here if the material is "challenged or likely to be challenged". So, Voidxor, have you misread WP:BURDEN are you really challenging the availability of 10 Gb NICs or is there something I'm missing? ~Kvng (talk) 15:50, 21 September 2015 (UTC)


 * Please note that BURDEN states "Any material lacking a reliable source directly supporting it may be removed and should not be restored without an inline citation to a reliable source". As I understand the policy, any removal of the material counts as it being challenged.


 * Also, this paragraph made other non-trivial claims besides "the availability of 10 Gb NICs". -- intgr [talk] 17:19, 21 September 2015 (UTC)


 * A "10 GbE NIC" is a (somewhat ordinary) NIC with a 10 GbE PHY or SPF+ slot. What makes it special? What actual information was provided? I've added the keyword "NIC" to the lead and that was about it. However, I challenge the statement "These plug into ordinary computer servers using PCI express" as well in that XG NICs also plug into non-ordinary computers, or non-servers, or sometimes not using PCIe. The paragraph wasn't very good to start with, so it's not really worth fighting over. Any information it did provide is probably better off elsewhere. So, what are we missing? --Zac67 (talk) 17:34, 21 September 2015 (UTC)


 * Intgr is correct. No challenge is required, but the removal of the content (initially by 108.181.220.196) thrice was a challenge (three challenges, actually). Furthermore, I'd point out that the tag was another challenge which predated removal.


 * Let's not forget that you're welcome (even encouraged) to restore the section with references . Just because something has been removed from an article doesn't mean that it's permanently expunged. – void  xor  19:18, 21 September 2015 (UTC)


 * Come to think of it, Zac67's comment of 18 September was a fifth challenge. – void  xor  17:15, 22 September 2015 (UTC)


 * Clearly there's a consensus to delete the section. Thanks for weighing in everyone. ~Kvng (talk) 04:58, 25 September 2015 (UTC)

10gbase-T section
It should also be noted that one reason few 10gbase-T implementations exist is because they dissipate too much power. Also 10gbase-T does not support Cat 5e and may not support Cat 6. New wiring standards Category 6e and Category 7 have been created for 10gbase-T. The claim of products in 2008 at less than $500 per port is not substantiated. --Quanstro 06:27, 21 October 2007 (UTC)

It's 2009 - just as 1Gig devices before these 10G devices were expensive and power hungry 10G devices track similar improvements in power consumption and cost. A 1G nic when the technology was new cost over $1000 and consumed over 7W. Now they are free on the mother board and consume less than a watt. Your choice of course - use it or not. It now looks like 10G copper will follow this same pattern...history repeats. Around about 2012-2013 you will probably see Lan on Mother Board to to 10G just like 1G did in the past. It will then be your choice, buy a CX4 or Fiber adapter or just use what came free with the server...

Cabling has also developed. There is no Category 6E perse - Category 6A has been on the market for a few years and most vendors are now in phase two to phase three designs. It is cheaper and easier to implement as a result. Category 6A is now proposed as the minimum standard for data centres by the ISO document.

It is always easier to be looking back - When I saw this post vintage 2007 I thought it would be interesting to note just how much this technology has progressed. I would say that whenever you fail to recognize trends (exponential ones at that) you might just miss the important points. Here the important point is more bandwidth will be needed in the future. People will want it to be cheap and easy and hopefully backward compatable. This is why ethernet is so popular. This is why 10G ethernet will also be popular. The fact that it will come free on a motherboard is always an important selling point in my book. —Preceding unsigned comment added by 198.90.65.175 (talk) 05:19, 22 July 2009 (UTC)

It is now 2016, and I haven't seen 10GbaseT devices on motherboards, or other places. I am talking about for home use, I presume for big data centers, it isn't hard to find. On the other hand, I pretty much see 1000baseT everywhere. My cable modem now has it, even though I am pretty sure it can't get the bits through that fast in either direction. My home network is pretty much gigabit to all devices that could come close to using it, though the wireless network needs updating. I do see gigabit devices in Goodwill stores, though, which tells you were it is now. Gah4 (talk) 08:33, 13 August 2016 (UTC)

"Center defect"???
What is a "center defect"? I've never heard that term.

Due to recent edits, the article now claims that FDDI fiber (62.5 μm fiber) has lower bandwidth due to a "center defect". I have no idea what to make of that. FDDI fiber has a lower modal bandwidth spec (i.e., higher dispersion) so it can't do 10 Gb/s at any significant distance. That's all there is to it as far as I know.

A reference and explanation would be good. Paul Koning (talk) 18:22, 7 March 2008 (UTC)


 * Check out slide 14 of http://bicsi.org/Events/Conferences/Spring/2005/GeorgePRES.pdf —Preceding unsigned comment added by 213.131.238.25 (talk) 13:28, 10 March 2008 (UTC)


 * Ok, but that slide doesn't talk about FDDI fiber. Apart from that one picture without any explanation, is there anything else?  I still don't think it's a good idea to refer to a term that is essentially never used in the industry and isn't defined anywhere. Paul Koning (talk) 16:59, 10 March 2008 (UTC)


 * One of the significant things about OM3 from the 10GBE viewpoint is the lack of need for a mode conditioning patch cord (MCPC) for -LX4. The 'center defect' makes explaining the operation of the MCPC really straight forward. I don't object if you want to remove 'center defect' however I think the mention of the MCPC for -LX4 needs to be kept and why it is not needed for OM3. Also OM3 is really significant for 10GBE as it allows the use of low cost -SR and will maybe kill off adoption -LRM. A decent description of OM3 and how it differs from FDDI would be helpful in this article. I am new to wikipedia and I saw the 10GBE entry was way out of date and I thought I would make a stab at bringing it up to date. —Preceding unsigned comment added by 77.97.110.196 (talk) 20:26, 10 March 2008 (UTC)

Centre defect was germain some time ago. Today we have moved on. FDDI fiber is also a discussion of the past as current technologies have long since diminished interest in this issue. In short bandwidth on multi mode fiber is now an issue with the advent of 10G technology. The effect of dispersion and loss as it relates specifically to VCSEL based transcievers is the issue of the day. This discussion is not based on "Center Dip" or "Center Defect" of MM glass but looks to come up with ways to effectively model the performance of MM systems when used in high speed networks. The advent of OM3 is based on a dispersion measurement system that compares signal delay across the entire fiber (a small amount of the core is not included). These measurements are then used to calculate the effective modal bandwith of the system. The measurement techniques and capabilities have since been refined and now OM4 fiber is being standardized. The advantages of these fibers include the support of longer distance for 10G along with greater tolerance for system defects like loss and transmitter bandwidth. OM3 can support a distance of 300 meters @ 10G for ethernet while OM4 will provide over 500 meters of reach. This distance can be traded for additional loss (like cross connections for example) which make it quite practicle for use in even large data centers. Compare the reach of 62.5 micron fibers at approximately 30 meters and you can see that this is just not viable. The issue is not mode conditioning patch cords...

All of these fibers are based on 50 micron laser optimized Multi Mode glass. These fibers are also backwards compatable with legacy systems. There is just no good reason left to be thinking about deploying 62.5 micron fiber in a network.


 * See for an explanation. It comes from using lasers designed for SMF with short distance MMF. Some MMF has a defect right in the center that has no effect with LED sources, but causes problems with laser sources. Gah4 (talk) 08:47, 13 August 2016 (UTC)

Citation needed?
10G Ethernet can also run over twin-axial cabling, twisted pair cabling, and backplanes.[citation needed] As I understand it, when someone adds a [citation needed], they are supposed to explain it in the Talk page. In this case, I don't know if they want a citation for twin-axial, twisted-pair, backplanes, all three, or something else. Any ideas? Gah4 (talk) 08:51, 13 August 2016 (UTC)


 * I have removed this tag. Existence of all three, with references to specific standards, is established in earlier #Standards section. ~Kvng (talk) 20:34, 17 August 2016 (UTC)

No host to host links?

 * 10 gigabit Ethernet defines only full duplex links which can be connected by switches.

Doubt this is exactly true, since two computers work fine connected back-to-back with a cross-over cable, although that is rarely done in practice. I think it is trying to say that "hubs" are not allowed because all links must be full duplex. Generally this a detail, and should be moved out of the lead. As indicated above. a bit more context and background could be given for the general reader without becoming a tutorial I hope. W Nowicki (talk) 16:50, 7 May 2011 (UTC)

10GBE was the first Ethernet speed to be full duplex only. Bearing in mind Ethernet used to be known as CSMA-CD with the MA standing for 'multiple access' this was a significant change at the time. Full-duplex implies point to point connections such as two computers connected together. The 'can be connected by switches' text is pointing out that more than two nodes can be connected together with the use of a switch. I agree this seems obvious now but might not have been obvious in the year 2000 when this text was probably written. —Preceding unsigned comment added by 68.233.206.78 (talk) 17:04, 21 May 2011 (UTC)


 * The lead now contains a description of this with adequate context. A generally connected by qualifier has been added, for instance. ~Kvng (talk) 19:09, 6 November 2022 (UTC)

Dead link
During several automated bot runs the following external link was found to be unavailable. Please check if the link is in fact down and fix or remove it in that case!


 * http://www.lightcounting.com/lighttrends/1004/
 * In 10 Gigabit Ethernet on 2011-05-23 02:08:01, 404 Not Found
 * In 10 Gigabit Ethernet on 2011-05-31 22:29:03, 404 Not Found

--JeffGBot (talk) 22:29, 31 May 2011 (UTC)


 * This page is archived at Wayback but all archived pages are 404 error messages. ~Kvng (talk) 19:09, 6 November 2022 (UTC)