Question Data transfer speeds? Cat 6 cables

nictry

Distinguished Member
Joined
Apr 28, 2002
Messages
6,179
Reaction score
749
Points
1,223
Location
Kent
Having started the process of hard wiring my house (see previous thread on weighing up pro vs DIY!) I have been doing some testing of cables/speeds etc. and despite doing a few searches cannot work out if my speeds are 'acceptable'

I have found that cabling through from the router (draytek) to a switch right next to it vs cabling through to another floor on a 3m cable into a 24 bit switch as zero impact on data transfer speeds so thats a good start I feel as the 'max' i get is still the max i get even when adding the longer cable and new switch into the equation makes no difference

I currently get between 50 MBps (409 Mbps) and 56 MBps (453 Mbps) write to a Synology NAS on approx 0.5 gb transfer tests. Read I get 97 to 105 MBps (773 Mbps to 850 Mbps) and this is all on gigabit switches and cat 6 cable

Is this in the right ball park or should I be seeing faster data transfer speeds? I am about to start cabling another 12 ports through to other rooms so really want to be sure this is as good as it gets before I do so :)

Thanks in advance
 
The end devices are you weakest link for testing networks. The switches will likely be non-blocking, i.e. you'll get 1Gb through each port but the end devices won't be able to fully utilise that rate in day to day operation, and using data copies isn't a great way of testing. Look into iperf instead.

Cat5E=Cat6 for 1 Gb, it makes no difference, assuming installed correctly. You could also daisychange a bunch of switches and it'll make no discernable difference to the throughput.

I don't know the NAS in question but those data rates look okay at first glance.

As long as you terminate your cables properly (which you won't know as I doubt you have a proper cable tester ;)) you'll get a 1Gb link when used with a Gig switch.
 
I've had a quick glance at the other thread and not got my head around it fully. If you're only talking of a few 3 - 10m lengths, you could just buy ready made patch leads.
 
Thats what I have done, there are 14 cables planned for 4 different rooms coming from the switch. I have already done the ground to 1st floor connections and running well (as well as the data above suggests) so all cables wioll be pre terminated flat Cat6 of differeing lengths. I have been testing with a basic cable tester and interestingly some of my old cables did show a missing number or two which was a bit strange but will not be connecting any up until checked as will be going under carpet and dont want to do this job more than once!
 
Good stuff. Note that I don't think you can get a 'proper' flat Cat6 cable as it needs the twists and the separating core. For a few meters I doubt it will matter, but those things are needed to reduce cross talk and interference.

Some cheap cables could only have two pairs (1, 2, 3 and 6) in them so they're only good for 10/100Mb links, whereas 1Gb needs all four pairs. That could be the reason for the failure or maybe you had a core break somewhere.

Have fun :)
 
Use something like iPerf as it takes out the hard drives out of the equation for testing speeds. As others have pointed out it either works or not.

If you are copying files there are too many variables, 500MB of small files will copy significantly slower than one 500MB file.
 
Thanks, to be honest I have been testing with a standard 1gb file and then a few 1-200mb files as a straight NAS to PC transfer and as long as they have come in around 100 mb/s i am happy enough that they are doing their job now.
 
As you've discovered, testing using files is "do-able" but - certainly in olden days when HDD's were much slower than today - you can often end up (effectively) testing the performance of the HDD's in the source and/or sink device.

It's OK for giving you some ball-park estimates, but the way to use is not to look at the absolute values, but (for ethernet) see whether you are exceeding the "fixed" rates ethernet works at. Viz, if the transfer is faster than about 11-12MB/s ("Bytes" not "bits") you know the link is running "faster than 100mbps," so it's probably 1000mbps. (Gigabit.) Likewise, 1000mbps vs 10G, 10mbps versus 100mbps and so on.

Testing with things like NetIO and iPerf doesn't use the HDD's and "synthesizes" a burst of data so we can be more confident we are actually stressing the network. (Technically the CPU, buses, NIC's in the hosts may offer a constraint, but it's been decades since a PC couldn't out perform 10/100/1000 ethernet - maybe not 10G, but I haven't tested any 10G.)

Ethernet works at fixed rates, it's 10/100/1000/10G etc or nothing, so speed tests and copy tests essentially allow you to discover which rate you've got as I've describe about. There's never any sense of ethernet "working but slowly" - so you'd never get a 1000mbps (gigabit) link working at "only" (say) 800mbps due to cabling, interference or anything else. A gigabit ethernet link is fundamentally 1000mbps or nothing at all.

Incidentally, Wi-Fi also works at fixed rates - it's just that there's a lot more of them to choose from depending on protocol used, number of antenna, signalling conditions, error correction, etc. and it varies from minute to minute. Given that there are so many potential Link Rates for Wi-Fi, it can give the illusion of being "continuously variable" but as with so much in IT, that's just a "trick of the light."
 
Last edited:

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom