Gigabit Ethernet: Dude, Where's My Bandwidth?

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
1.000.000.000 bits / 8 = 125.000.000 bytes
125.000.000 bytes / 1024 = 122.070 KiloBytes
122.070 KiloBytes / 1024 = 119,21 MegaBytes per second theoretical max bandwidth.

On average there's about a 10% overhead, so you could easily say that a gigabit network performs around 100MB/s
 
[citation][nom]zetone[/nom]@all complaining about the technical aspects of this article: I think the target audience is NOT network administrators.[/citation]


***EXACTLY***

Gentlemen, I apologize to those of you who knew this stuff already, but on occasion I strive to make accessible articles for folks who posess no knowlesge whatsoever about a subject.

This is a top-down primer, nothing more, and I apologize if it implies more. It is a lightning quick investigation of a real-world gigabit network in the home and the most obvious factors limiting it's speed, to give a layman some basic understanding of what's going on here. I thought if including a lot of other information like MTUs, frame sizes, stripe sizes, etc, but for this kind of article I thought it best to keep it simple.

We're hapy to accomodate you folks who are looking for more, and we certainly target articles toward that audience as well. So if you'd like to see an in-depth investigation about a particlular aspect of networking, feel free to suggest one and we'll be happy to consider it.

[***NOTE***] I edited the first page to make it obvious that this is for folks who are interested in the basics. I apologize for not doing this in the first place, and if I've wasted anyone's time.

Sorry for any confusion and not making the target audience clearer. Peace out fellows!

- Don Woligroski
 
That was definitely lite reading. Not even the basics ware covered. On this level I explain how things work to my mom. Give as some credit.
 
@MartenKL
"Gbit is actually 10^9 bits per second, ie about 119 MB/s."

Nope.

kilo = 1024 (2^10)
mega = 1024x1024 (2^20)
giga = 1020x1024x1024 = 1 073 741 824 bits (2^30)
=> 1 Gbit is actually 2^30 bits.


Also
1 Gbit = 1024 Mbit => 1 Gbit/s translates to 128 MB/s.
 
you young kids ... use ttcp to test your network bandwidth (wsttcp on windows)

You can use it to confirm each of your PC's are running properly (and network switches and routers). You can also use it to confirm your wireless performance.
 
Test 3: Shorten Cable length from 50 ft. to 28 ft.
In this test, we removed one of the 25 ft. cables and replaced it with a 3 ft. Cat 5e cable.

Come on... take 30 seconds to read what you wrote, gosh...
 
[citation][nom]computabug[/nom]Come on... take 30 seconds to read what you wrote, gosh...[/citation]

*sigh*

Do the math, lad. We removed ONE of the 25 ft cables. (implying two 25 foot cables)

25 + 25 = 50

25 + 3 = 28
 
Perfect article for the general population and me included. I loved to see how it's actually the HDD causing the problem and not the typical home network.

If someone is a network administrator or wants more in dept knowledge on this issue, you either know more already and don't need any of this research or you're a poor network administrator. Either way, you're not the target audience of this nice short article.
 
This column is not at all about networking... it is about disk transfers. He doesn't even mention streaming data, real-time data acquiring, the benefits of Gbit networking when it comes to multiple machines on your LAN (office or home). What a tremendous waste of my time reading it and my Gigabit network downloading it.
 
If you really wanted to test out the HD theory, which it certainly seems to be... why not try out some SSDs? Or atleast 3-4 drive RAID0 arrays, perhaps Raptors? On an ICH8R, I get about 100MB/s *write* speeds with two Seagate 320GB drives.

My point is that you proved RAM can be fed to the GbE controller at near-peak, but you never proved if it can be fed from the SATA controller equally fast when its known the hard drive subsystem is capable of such performance. Since the SATA controller on many boards is off the Southbridge, this data needs to be transferred from the Southbridge to a PCI-e lane with the GbE controller (generally off the Southbridge). This is starkly different than RAM, where the data is being fed from the high-speed Nortbridge down to the GbE controller.
 
[citation][nom]neiroatopelcc[/nom]Could you run a test of cabling difference in a more large scale place? ie. does it make a difference on a hp 4108 switch ... [/citation]

I think the whole point of the article was about a home gigabit network. Anyone using an HP4108 for their home network is A) Gullible, has too much bloody money, and talked to the wrong salesperson. They will never need that. B) Living in a house with as many rooms as some hotels. C) For being held at someone's house, putting on a very large LAN Party.

Also to expect the author to test every single possible configuration of switches/routers/cables etc.. and to post the results just so that they cover the situation that each of us readers is considering/running is absurd.

I do agree that the test needed to be on a larger scale (as has already been mentioned several times). Using 25 ft cables would barely get me from the office to the next room (and thats a straight shot through the door), let alone from the office, upstairs to the room at the other end of the house and on the opposite side. Not that I have a big home by any means, but if I wanted to run cabling from the office to the Home theater PC it would take at least 80-100 ft of cable(thats a guess).

[citation][nom]neiroatopelcc[/nom]We're using cat6 at work whenever we do something new, but we have plenty of old cat5e cables as well. Not sure if it actually makes a difference.btw. cat 5 without the e mark does NOT work very well with gigabit. We've got a bunch of those old cables in one of the offices I am working, and we're constantly getting crc/allignment warnings because it's trying to run gigabit over an old cable... [/citation]

When I was at the university, we were finding the same problems. If the user was complaining about problems, when we got to the wiring closet if they had cat 5, the first thing we did was replace the cable with cat6. 99% of the time that would fix the users complaints about speed/errors.
 
The assumption that just because 25 feet CAT5e worked the same as CAT6, 50 feet would probably work the same is dead wrong. In a previous job at a data center I spent some time benchmarking the network and it became apparent that anything more than 25-30 feet of CAT5e would degrade to 100Mb/s. Please do your readers a favor and run the CAT5e comparison to CAT6 over 50' and 100', to actually prove your theory.
 
I must chime in and ask "Why don't you talk about Jumbo Frames?!?" If anything boosted the speeds on my network (and by quite a bit), it was getting switches that support JF and setting all the NICs to use JF.
 
Fantastic article for the novice not knowing much about networks and computer mechanics, I think If you went the whole hog in explaining in indept you would only be preeching to the converted only the so called experts what understand. Great JOB
Although it would br nice to follow up in anouther article to check up on SSD's,coax,fiberoptics,wireless,and a combination of wireless and cable.I know there are alot of readers out there that already know the answers and could probably write this article but for people who are covices or people like myself who no matter how much they read can never remember what they read from one day to the next it would be a nice refresher.
I have set up my own home network of 3 computers years ago useing a Netgear Duel speed hub,cable and arigenaly I used netgear cards but nowdays its all in the motherboard and all I have been doing is just changeing the computers and or just updateing the components in the computer,so it has been a refreshing reminder for me but I would still like to read more
Gazz
 
[citation][nom]SpadeM[/nom]For all tech people out there, the title of the article should have been a dead give away about the technical aspect of this piece. Sure they could have used a server platform with a server os, SSD's and ram disks, and why not some tech language what most people don't understand. But, as the titles states in a very suggestive way, this article is for people that have simple questions and seek simple answers. I'm ok with this piece, it isn't and injustice or it isn't wrong in any way IF you look at who it is addressed to. Remember the KISS rule guys.[/citation]

O Im sorry, I haddnt noticed the site name change to 'Tech4Dummies'.

Tom's is SUPPOSE to be for enthusiasts. i.e. geeks, sysadmins, and the kind of people who want the nitty gritty.

At this point, we may as well be reading cnet...
 
I like this article. Hopefully more people will read it and at least take away the difference between Mb and MB's! Sure there were some inaccuracies, but otherwise this was a well done article.

Jumbo packets, mtu's, etc are irrelevant to the audience of this article, not to be forgotten though! I've got numerous RAID 1,0, and 5's at my house, I can get over 100MB's pushed from my 4 disk RAID5 if I stream HD video to my three TV's and a couple computers all at once. How sweet that is! This is with a Cisco 2960G 24 port switch and all Intel Gigabit NIC's.
 
For those of you with cat5e not running gig, over any length to 100m, check were it's running. I've had runs a hair over 100m run perfectly fine at gig speeds. Hell I've had 100mb run full speed at over 700ft!

Course the quality of cable is also important. Hitachi and Belden is some of the best in my experience.
 
As others have mentioned, the article spend a lot of time talking about disk speed. And not enough time talking about the networking configuration side of things.

There should have been no need to discuss disk speeds in any great detail. A network transfer does not have the result in a disk read or write! There are tools available that allow the testing of network speed WITHOUT using a disk (RAID, RAM disk or otherwise).

For example,
http://www.passmark.com/products/pt_advnet.htm
 
[citation][nom]B-Unit[/nom]Tom's is SUPPOSE to be for enthusiasts. i.e. geeks, sysadmins, and the kind of people who want the nitty gritty. At this point, we may as well be reading cnet...[/citation]

Our mandate isn't to be exclusively for tech-snobs.

I go out of my way to write articles for the lowest common denominator now and again. We also strive to write in-depth articles, as well, for those who can appreciate them.

There's no need or mandate to completely ignore folks who have a lower level of expertise. This article was for them and I make no apologies for that.
 
I honestly think the people complaining just look at the first page and perhaps a few graphs and race to get the first "correction" in. I read multiple responses targeted at the fact that there would be network overhead and that you won't ever be able to reach the theoretical max. Well, that was stated specifically at least twice in the article. You try so hard to look smart on the internet and just embarrass yourselves.
 
I have no idea how this guy sees results this quick. RAM drive to RAM drive on a pair of updated and tweaked Vista systems with CFS, the best I've ever seen is around 60 MB/s. That's with a pair of Marvell Yukon PCIe gigabit NICs. The onboard NICS on my circa 2006 systems were much worse.
 
Networking is an area where Linux excells. Windows simply CANNOT match the networking performance of Linux.

Nice article Tom's
but
leaving out the performance driving OS, Linux and BSD as in Apple is a MISTAKE.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom