Well, the member review forum is up, but there's no content. How sad.

Time to change that!

Without further ado, here is my Member Review of the 3Com 3C16478 16-port Gigabit Ethernet Switch.

----------------------------------------------------

With the onslaught of Gigabit Ethernet enabled motherboards and associated equipment, many have asked, is it really worth it? Will it make my web experience that much faster? Should I upgrade?

This review of sorts will attempt to answer this question.

As most of you know, I work for a large insurance company. Fortunately I have access to networking equipment that most do not. Recently, I procured a 3Com 16-Port Gigabit-Ethernet unmanaged switch for deployment to a department that is notorious for transferring large (10GB+) database files between themselves. Ever curious, I was eager to see just how much of a difference Gig-E would make compared to the standard 100-base switches most of the users are connected to.

Just for fun, I tested and benchmarked a Crossover cable and a 8-port workgroup switch (no integrated router) as well, just to see how they fared.

First off, let me show off the new Gig-E switch. It arrived in my office via courier – hand courier, as the company I buy my gear from is literally down the street. They are kind enough to drop my stuff off themselves.

Packaging was standard fare, brown-box w/ logos.



It was packed reasonably well, ensuring it was delivered in the same condition it left the factory:



Removing the foam packing material revealed a nice touch – a 3-Com Quality sticker seals the unit in the plastic bag. It’s a feel-good item, and a nice touch.



HereÂ’s the unit out of the packaging. You can clearly see that itÂ’s designed to be a 24-port unit; 3Com has just removed 8 of the ports.



Flipping the unit over to take a look at the rear reveals something that’s always annoyed me with 3Com switches. For some reason, they feel the need to cover the power outlet with a sticker that says ‘power socket’. Why they do this is absolutely beyond me. If a network administrator or cabling specialist doesn’t know what’s supposed to go here then they DESERVE whatever happens to them or their equipment. Considering the audience that is receiving these units, it seems like a complete waste of 3Com’s time to apply this sticker, and my time to remove it. Notice it’s split in half horizontally, contributing to the annoyance factor.



Right beside the ‘I’m so retarded that I don’t know what a power socket is’ sticker are 2 25mm fans. These puppies are LOUD - louder than the Stock Intel Heatsink fan at 2500RPM. Considering that these units are designed to go in a server room or wiring closet, it isn’t really an issue. The fans are blowing out the back, creating negative pressure on the inside of the switch housing. Intake is via the same grating along the sides of the unit.



3Com was kind enough to include a nice, heavy gauge power cord...



...as well as rack-mount dog-ears (& screws to mount them), and rubber feet, in case this unit will be mounted on a table-top instead of a rack. Note you canÂ’t use the rubber feet if youÂ’re installing the unit in a rack, as there is no clearance to allow this.



Here are the other 2 units included in this review:

The D-Link DSS-8+:



The Allied Telesyn ATFS724i:



Setup of an unmanaged switch is fairly simple. Plug it into the wall, plug in your clients. Make sure youÂ’re using compatible IPÂ’s.

The test is simply copying 2 large files between 2 computers. Both computers are identical except for their IP addresses. Here is the hardware:



Clients:

Intel Pentium 4 2.8c
2x256MB Kingston Value RAM
Intel D865GLC Motherboard, with Integrated Gig-E LAN & Video
160GB Western Digital SE SATA drive
Patch Cables: Custom made and certified Gig-E connections by Anixter Canada (This is important as some cable manufacturers test only the wire, and not the actual connections after the cables have been made.)

Switches:

Anixter Crossover Cable – connections certified Gig-E by Anixter Canada
3Com 3C16478 16-port Gigabit Ethernet Unmanaged Switch
Allied Telesyn ATFS724i 24-port 10/100 Ethernet Unmanaged Switch
D-Link DSS-8+ 8-port 10/100 Ethernet Unmanaged Switch

Software:

Windows XP Professional, Corporate Edition, SR-1 (NOT A WAREZ COPY!)
- A fresh install was performed on one machine, with all current drivers from Intel.com and all current Windows Updates. The machine was then Imaged using sysprep and PowerQuest DeployCenter to provide an identical installation on both machines.

Robust File Copy vXP010 (Available in the Windows Server 2003 Resource Kit)

Unreal Server 2004 .zip file (Downloaded from www.unrealtournament2004.com) – 777MB

Far Cry DVD .nrg file (Ripped from an original Far Cry DVD using Nero 6 Ultra) – 3.4GB

I copied the files 4 times each – twice from each machine to the other, using Robocopy. There was one exception – during the D-Link tests, a user hub upstairs failed, and needed to be replaced immediately. As a result, I could only get 3 tests done using that switch.

Robocopy is a command-line utility that gathers and statistical data of your file transfer while transferring. ItÂ’s great for scripting as there are many switches available for configuration. My personal experience has shown that it is as fast, if not faster than Xcopy, although those comparisons are out of the scope of this review. For more information on Robocopy, please see the Windows Server 2003 Resource Kit.

All testing was done OFF LINE – tests were NOT done on my public network, as user traffic may have skewed the results.

Results

A note concerning the results. Throughput is measured below in MB/minute or second. That is MegaBYTES per minute or second, as opposed to Mb, which is MegaBITS. Note the change from a capital ‘B’ to a lower case ‘b’. 1MB = 8Mb (8 bits in a byte). A typical cable modem will transfer data at the rate of 1.5Mb/second, or 0.2MB/second.

The first testing I did was with the D-link switch. Results from this switch are impressive, considering it only costs $30.00 Canadian. Because the unit only has a 1.6Gb back plane, using this unit in a high traffic environment is not recommended. D-link claims that this unit is capable of 200Mbps per port (25MB/s). Based on my testing, the D-Link unit did not live up to that specification. Throughput was consistent, averaging 593.535 MB/min, or 9.892MB/sec, well under the claimed 25MB/s per port. Here are the individual results:

UT copy #1: 1:19, 590.325MB/min
FC copy #1: 5:57, 593.543MB/min

UT copy #2: 1:18, 594.552MB/min
FC copy #2: 5:56, 595.339MB/min

UT copy #3: 1:18, 593.727MB/min
FC copy #3: 5:57, 593.726MB/min

Next, I moved on to the Allied Telesyn switch, which is full-duplex 10/100 auto-config switch for corporate/enterprise use. I use this switch to connect all users to our core Cisco unit. ItÂ’s a good performer, and has proven to be reliable. Once again, throughput was consistent, averaging 592.572MB/min, or 9.876MB/second. Again, here are the individual results:

UT copy #1: 1:19, 589.743MB/min
FC copy #1: 5:58, 591.372MB/min

UT copy #2: 1:18, 594.317MB/min
FC copy #2: 5:56, 595.260MB/min

UT copy #3: 1:19, 589.743MB/min
FC copy #3: 5:58, 591.321MB/min

UT copy #4: 1:18, 593.373MB/min
FC copy #4: 5:56, 595.444MB/min

After that, it was time to test my new baby, the Gig-E 3-Com switch. This too, is a full duplex, auto-config unit, capable of 10/100/1000 speeds. I was a little surprised by the inconsistency of this unit. I got results ranging from almost 2300MB/min to as low as 1200MB/min (still, more than double the results from the 100-base models).

It could be something as simple as the switch mapping paths on the initial transfer (where I got the 1200MB/min result) though. Average results for this switch were an amazing 1652.440MB/min, or 27.54MB/second – an almost 300% increase over the 100-base models. Once more, the individual results:

UT copy #1: 0:30, 1556.278MB/min
FC copy #1: 2:54, 1213.994MB/min

UT copy #2: 0:20, 2295.525MB/min
FC copy #2: 2:16, 1558.999MB/min

UT copy #3: 0:27, 1668.398MB/min
FC copy #3: 2:31, 1396.558MB/min

UT copy #4: 0:23, 1954.490MB/min
FC copy #4: 2:14, 1575.276MB/min

The last test was done using a crossover cable, to judge just how much (if any) interference the 3Com switch introduced into the file transfer. These results were a bit surprising, to say the least. Again, the performance inconsistency was there, as with the 3Com unit. Average throughput on the crossover cable was 1599.423MB/min, or 26.66MB/second – SLOWER than the 3-Com switch! Finally the individual results from the crossover test:

UT copy #1: 0:21, 2150.092MB/min
FC copy #1: 2:42, 1302.257MB/min

UT copy #2: 0:29, 1597.899MB/min
FC copy #2: 2:10, 1624.665MB/min

UT copy #3: 0:32, 1428.939MB/min
FC copy #3: 2:36, 1357.054MB/min

UT copy #4: 0:28, 1658.260MB/min
FC copy #4: 2:06, 1676.218MB/min

The results of this testing are most pleasing to say the least. A 3% increase in network throughput is very impressive. Keep in mind though, that the difference between Fast-E and Gig-E is 10%. Where is the extra 7%? The hardware used cannot keep up with the data transfer speeds of Gigabit Ethernet. Where the 100-base models topped out, and were consistent, the bottleneck was the switch. Remove that bottleneck (with a Gig-E switch or Crossover cable) and the bottleneck moves from the network back to the computer, likely the hard drive.

Currently, I’m buying the Allied Switches for $475.00. This 3Com switch was $975.00 for a 16-port model, and $1500.00 for the 24-port model – an expensive unit to be sure, and way out of the reach of home users. Even in an Enterprise situation, I wouldn’t recommend this to my superiors for anyone but those power users who are transferring large DB files around. IMHO, the price/performance ratio is currently just too high for the performance gains realized. I’m sure as more and more companies begin to adopt Gig-E the costs associated with moving to this platform will come down, as they did with Fast Ethernet in the late 90’s. Eventually, you might be able to buy that 8-port D-Link in Gig-E flavor, and experience the pleasure of transferring over 3GB of data in less than 150 seconds for less than $50.00.

What does this mean to everyday users? Absolutely nothing. Gig-E will have no bearing whatsoever on your internet downloads, surfing speed or on-line gaming experience, as most (if not all) home cable/DSL modems are capped at 10Mb/sec or slower. Gig-E might be of use in a large LAN gaming event; however, again, performance increases over Fast-E might be negligible, considering client hardware and/or server hardware. Nice to have, but likely not necessary.

As far as modding potential goes, thereÂ’s not much. I guess you could replace the green and amber LEDÂ’s with blue and red ones, but who really cares? ItÂ’s a closet switch. It belongs in a wiring closet. No oneÂ’s going to see it behind the rats nest of cables anyway.

I give this unit 7 pimp hats out of 10, because of the price/performance ratio. ItÂ’s an excellent unit in the right situation, but out of reach for the mortals paying for their own gear.

Nice review PJ

About time you got a new title though ...

Thanks Jon...

None of that matrix code makes sense to me lol
UT copy #1: 1:19, 590.325MB/min
FC copy #1: 5:57, 593.543MB/min

UT copy #2: 1:18, 594.552MB/min
FC copy #2: 5:56, 595.339MB/min

UT copy #3: 1:18, 593.727MB/min
FC copy #3: 5:57, 593.726MB/min


Good review tho

If you thought that was confusing, you should try putting it into a graph

Guess I should break it down for you:

UT/FC copy #1 = Unreal Tournament/Far Cry copy #1
1:19 = the time it took to copy the file
590.325MB/min = the network throughput in megabytes per minute. If you want the per-second value, divide it by 60 (590.325 / 60 = 9.839MB/s

Hope that helps...

Awesome Review PJ, I'm actually glad you didn't put it into a graph I like tables. I also like that you repeated the tests for consistency.

Brandon Cisneros

good work for our first member review. You get kudos.

Great job PJ!!

Yeah, nice job PJ. I am trying to talk them out of an HP switch at work that was used at our old Missouri plant which was closed down. They have no use for it since they have Cisco equipment here.

Verry impressive, great job PJ