Those joiners can easily reduce your maximum cable length by 5m. I've seen some cheap ones which internally had about 5cm of untwisted wires connecting the two ends. Those ones at least say "CAT6" so hopefully they have a PCB inside, but still just the RJ45 connector is quite poor and will loose some signal strength.
One thing to bear in mind is Ethernet is rated for maximum distance over the specified cable. While CAT5 is rated for 100M at Gb, it can do much better at significantly shorter distances. Also, you mention the switch port shut down due to errors. I expect it was a significant amount of error and you were getting errors long before you reached that point. You have to properly measure the error rate, though you can get some idea by watching the error count. There is test equipment called "Bit Error Rate Test" (BERT) that is designed just for that purpose.
@@DavidM2002 Gigabit Ethernet was designed for CAT5 before 5e was available. So yes, plain CAT5. CAT5e is CAT5 with tighter tolerances. However, it won't make much difference, unless you're getting close to 100M. I have plain CAT5 in my condo, which was installed by my ISP in the late 90s (I provided the cable). They pulled it in with the coax when I got a cable modem. Also, they actually fished the cable through a wall, next to air duct, over ceilings, etc.. They even patched the drywall when they were done. None of that staple to the baseboard nonsense for me. 🙂
@@James_Knott Thanks James. I thought for sure it was a typo. I ran 5e in my little house about 20 years ago when I had the walls open doing a reno. And that was cutting edge for a house back then. So you were really ahead of your time. Well done.
@@DavidM2002 Well, most of my career has been in telecom, but at that time I was doing 3rd level OS/2 support at IBM Canada, so I would hope I'm a bit ahead of others. 🙂 Perhaps I should have said tighter tolerances, rather than specs. BTW, my first work LAN experience goes back to early 1978, on a proprietary Rockwell Collins network, in the Air Canada reservation system. First Ethernet was with 10base5 "Thicknet" connecting VAX 11/780 computers and in the late 80s, I hand wired some Ethernet controllers, on prototype boards for Data General Eclipse computers. So yeah, I'm a bit more technically advanced than most people.
Nice video, short and to the point. This makes me think of an analogy: can you drive an average car at over 100 miles per hour on the highway? Yes. But then come the details. Is it safe? Is there other traffic? Is it reliable? Driving fast has its risks, other traffic may cause you problems or you for it. Lots of factors come into play. So be aware of the limitations of the overall situation, understand the interactions, the pros and cons, and work out the best solution. Then be willing to come back at some point to reevaluate it to see what needs to be changed to meet the requirements at that point in time.
My house is all cat5e and I've got it fully on 10GbE now. I did some research on this too. Although standard does not qualify cat5e for 10GbE, but by cable performance characteristics cat5e should be able to carry 10GbE up to about 45m. However this is purely the cable, without the plug and socket. In real life conditions it would be something about half of that, so 25m is totally expected.
My condo has plain CAT5 (not 5e) that was installed in the late 90s. It's doing fine with Gb, though I expect I would want to replace it, if I ever went 10Gb. That shouldn't be too hard, as the CAT5 was pulled in by my ISP, when I first got a cable modem, and so the cable is just lying loose and not anchored anywhere. Yes, it was properly pulled in, fished up a wall, along side air ducts, over ceilings and more, so that it's only visible where it crosses my laundry room ceiling. None of this crappy staple to the baseboards you'd get these days.
Let's say you have cat5e already built into the wall. If you extend it to a patch panel using wall plates and cat6 cable, would you still be limited to 45m total for 10GbE, or 45m only for the cat5e cable?
@@Mrperson0 It depends. Think of the signal in the wire is taking damage when travelling. For cat5e, it is like 10gbe signal can only live for 45m. Unless the signal reaches something that regenerates it, like a switch or router, it will not have any more HP gained going through anything else. Plus that transitions like sockets and plugs will do one time damage to the signal too. So it depends on your actual environment.
@@CCL13CN Ahh, thanks for the explanation! Makes sense that only a switch or router would regenerate the signal, not a keystone. Hopefully my 10gbe upgrade will go smoothly and I will be able to use a cat5e cable that is already installed to connect at that speed.
Worked in a DC that had generations of cabling going back to the late 90’s. Predominantly it was Cat5e so we’d just plug 10G into whatever happened to be in the patch panel, for cross connecting adjacent racks and whatnot it never presented a problem.
There is more than if the cable works or not. A user doesn't really care, but the manufacturer must meet all governmental requirement regarding radiated emissions (i.e. generation of electrical interference) In the US that's FCC class A (office) and class B (residential). A less superior cable will generate more radiated emissions up to the point of failing the test. The responsibility doesn't fall on the cable, but the manufacturer of the switch or SFP module. So to ameliorate this, the manufacturer will specify a higher grade cable.
Cable Category most likely doesn't have a huge impact on emissions. I would bet a CAT5E S/FTP (Shielded, Foil Twisted Pair) cable has significantly less emissions than a CAT6A UTP (Unshielded Twisted Pair) cable.
@@bumbixp I'm going to assume that by emissions we are referring the industry standard referral of cross talk. External shielding only takes into account external traversal of cross talk, it does nothing for the internal cross talk withing the cable itself. For that we have the extra twisted in the twisted pairs within the jacket. In addition to that, you are missing the factor of impedance/resistance to voltage. Cat 5 is a thinner copper gauge of 24, Cat6 is a thicker copper gauge of 23. Therefore, Cat6 can go a longer distance due to it's greater copper mass, it has tighter twisted pairs to prevent cross talk within the jacket. THAT is what's important. The shielding only matters if you are running it next to areas prone to electrical interference. Aside that, the external emissions wouldn't make any impact of any kind to UTP.
Yup, my 2005 house was wired completely with CAT5e and I have several 10G links running over it that work just fine. I also have a 2005 era OM1 fiber connection from the main house to an additional unit on the property and I was able to run 10G over that by using LRM optical modules with MCP (Mode condition patch) cables.
So we had fun on a pre-rip-out floor at work. The cable under test was Molex DataCAT-5 installed OG in 1993. We TDR'd lengths from 20-90M in the building. We were using Molex OG patch leads. The switch under test was a Cisco 9300UX with the integrated copper interfaces capable of 10Gb. We managed a full 90M run - but the switch wasn't happy about it. Lots of errors - but no UDLD.. Runs of OG CAT-5 were happy-enough to 50-60M without errors. But - it's all gone now - The Ripout sees Molex DataCAT-6 cable installed. That works nice to 90M without any worries.
At the end of the day, if it works and you're satisfied with it's performance and any potential fallbacks/issues, awesome job in testing it out yourself! This is not too far off the swan song of running cheap vs expensive HDMI cables over short distances.
I have done this with short runs, but there are a lot of caveats. If you want a reliable connection you can depend on, use 6a. That said, solid core cabling fares better than stranded in terms of signal quality, and if you can reduce or eliminate jacks or couplers, that will help. If the wire is going up into a hot attic, the impedance of the wire will change and may make an otherwise acceptable connection turn marginal or unacceptable. Had this issue with a 50ft unsplit run - was great all winter and spring, until the first hot day of summer when the switches disabled the ports for link flap.
I'm going to assume that by emissions we are referring the industry standard referral of cross talk. External shielding only takes into account external traversal of cross talk, it does nothing for the internal cross talk within the cable itself, where the twisted pairs are. For that we have the extra twisted in the twisted pairs within the jacket, thus 6 has more twisted pairs. In addition to that, is the factor of impedance/resistance to voltage. Cat 5 is a thinner copper gauge of 24, Cat6 is a thicker copper gauge of 23. Therefore, Cat6 can go a longer distance due to it's greater copper mass, it has tighter twisted pairs to prevent cross talk within the jacket. Shielding only matters if you are running it next to areas prone to electrical interference. Aside that, the external emissions wouldn't make any impact of any kind upon unshielded twisted pair "UTP". So Cat5 WILL work with 10gb or 2.5gb. With CAT5 on 10gb/2.5gb your speed to distance ratio will lower the further the cable is, you will produce and be susceptible to more emissions/cross talk, within the jacket. However, it will work and will be a none issue to home owners with Cat5/e.
Here's the wisdom of the day: You can run every signal over every cable, as long as the cable is short enough. Better shielding (or shielding at all), a bigger diameter and cable twisting will only extend the range the signal can travel and still have good quality but they never decide if you can run a signal over a cable at all. You can run Gbit Ethernet even over totally unshilded, untwisted, thin phone wires, as long as you only need cross a gab of less than a meter or occasional signal errors are okay (Ethernet will detect them and drop bad packets, higher protocol layers will then resend missing data that didn't make it). High quality plugs and cables only give you longer ranges and less signal errors, they don't enable you to do something you could not otherwise do.
Exactly. I've reused lots of old untwisted 4 wire phone cables and they are working fine for 100mb and even 1gb when using two 4 wire phone cables in one rj45 connector for shot distances.
It would've been useful to show whether any errored frames were recorded. By the time a port is error-disabled the error rate is pretty high. Perhaps there were errors not regular enough to affect data transfer speeds, but irregular packet loss will be bad for online gaming and streaming.
Then there is the potential problem of alien crosstalk. In other words the risk of multiple cables in a bundle, disturbing each other. Maybe not a big problem in a residential network, but definitely worth considering in a larger commercial system. C6a components are much less susceptible to that problem. In fact, the reason many U/UTP C6a cables are so bulky, is very deliberate. It’s simply just to separate the pairs in one cable from the pairs in the next one. Just to minimise alien crosstalk.
It really depends on the quality of cables as in how strictly have the cables been tested for. The monoprrice slimrun CAT 6A UTP claims 550Mhz tested but lengths as short as 10 feet have failed to give 10Gbps. It falls back to 5 Gbps. There'sa video on youtube showing that if you do a quick search.
One question I had but I never bothered to test is: If you plug multiple antennas into a wifi access point designed for MIMO but you separate them by a fair distance, do you get better coverage? E.g. one access point with 2 antennas 10 metres apart vs 1 access point in the middle
I have high hopes for an old ~37m run of CAT6A with a pair of $36 30m transceivers (which Amazon has lost) when I do eventually get them. If not, I'll run fiber before spending the money on high power Cu transceivers.
I'd say you have a very good chance of it working just fine. Please comment back once you give it a go. I'm interested to know if your result replicate mine.
As an expert in this field. It can negotiate at that speed but don’t expect it to actually perform consistently at that speed. Cross talk, packet drops and so forth are a very major reason why it’s not rated for 10gb. Cat6A and up is rated because the wire is thicker and so is the insulation. This greatly reduces cross talk, RF interference and allows the connection to run far more stable.
@@Daniel15au caT5e is rated from 30awg to 26 typically. It's quite thin and so is the insulation, that's why it's so prone to RF interference and why it's not rated beyond 1gbps and even at 1gbps it can be flakey.
The SFP+ “30M” rating on those modules is more to do with the power draw/heat dissipation of the module. SFP+ isn’t normally meant to run 10Gb copper Ethernet, as the SFP+ spec is very power budget limited, and getting 10Gb to go for 100M on ANY cable requires quite a bit of power per port.
Yes, with Cat5e cables, you can still achieve a maximum speed of 1 Gbps. However, over short distances, it is possible to get slightly higher speeds than 1 Gbps, although it may not be consistent or reliable
What speed will the router work on if you have a 2.5 gbs switch with a 10 gbs sfp+ port plugged in a router with a 10gbs port. The speed of the data is 1.6 gbs. My network card is intel x540-t2 and detects the 10 gbs network if I plugin the cable directly to the card but if the network is passed through my 2.5 gbs switch the card detects it as 1gbs only? My question is If I buy an sfp+ 10 gbs rj45 and plug it in the 2.5 gbs switch in the 10gbs port will it make a difference. i just realised that the intel x540 runs only at 1 gbs or 10. Not 5 or 2.5. Thanks
I think I found out that there will be 10gbs on that port but internally will be at 2.5gbs on the other port. That's why I will have input at 10 and gave a 10 going to computer and the rest at 2.5 or 1 gbs
No surprise if the answer is Yes, it can. Because all of 4 pair of twisted pair is connected, the pin connection is intact, and the auto negotiation on link speed is acknowledged. The real problem is stability. That's why Cat 6 cable has tons of shielding element, even the rj45 jack is shielded, and got bigger thickness. The stability got worsen with every port used, because of signaling interference. I've got enough of stress when setting up Gigabit switch with 24 ports onward, it break down the whole connection.
🎯 Key Takeaways for quick navigation: 00:00 🕵️♂️ You can use Cat5e copper cabling for 10 gigabit Ethernet, but the recommended cable is Cat6a for 100 meters and Cat6 for up to 55 meters. 01:10 🧪 Testing with 30-meter SFP+ copper modules showed that 25 meters is a practical limit for Cat5e cabling for 10 gigabit Ethernet. 01:51 💻 For practical home use, Cat5e cabling can likely handle 10 gigabit Ethernet without issues for considerable distances. Made with HARPA AI
Hello, at 2:01 minute mark, there are 2 NAS 8 bay stacked. I would like to know the model. I think they are Synology DiskStation DS1821+ units, but it's not very clear. Thank you.
I`m not sure if this is a non European thing, but why does everyone seem to use UTP on Cat5e and never consider SFTP / FTP / STP ? I know the UTP versions are cheaper but still the shielding could help somehow to reduces errors especially if there is shielding between the wire pairs. Also the couplers will add additional resistance on the wire and so the connection attenuation will go up. But still, good idea to put a video up telling people Cat5e could transfer 10GBe, if the conditions are right.
I think it is because parts of the US still use old and noisy electric infrastructure running on on 110v, whereas all of europe is on the same modern 220v grid and it is 'cleaner' hence electrical noise is less of a problem, hence less need for shielded pair. I might be wrong, but that's what I always thought. Plus, the British also have the best plug and socket system ever designed!
also twisted pairs are used because of differential signaling which cancels out noise induced on the lines. only if the noise amplitude is greater than a certain threshhold would you need extra sheilding. Something like an electric fence can be problematic for external network cables
UTP was the standard for telephone lines and slowly evolved towards today's CAT6a. The tools and techniques for working with UTP all came from the phone companies. Even before networking became a common thing, CAT5 was being run for commerical phone systems. So when networking started to take off, it was convenient to use that existing cabling. Ethernet was engineered to use this cable because it already existed and over the short distances(under 100m) found in most buildings, it was more than good enough. Modern CAT6a just improved earlier systems while maintaining backwards compatibility. For the shielding to be effective the equipment at both ends of the link really need to be designed for it and most stuff is not. The shielding can create ground loops which can be a major problem. Which is likely the major reason the telephone systems avoided it in the first place.
Why though? Fiber is very expensive for the equipment, hard to make in your own lengths and would need to be adapted to copper/implemented with add-in cards for just about any device that you would connect to it. Its a lot of added expense when cheap copper will future-proof you for a while. This person runs 10G on cat5e on in-house lengths, and I expect that there is nothing you actually use the internet for that will come near to saturating the 10G connection. Most storage would bottleneck the actual speed
@@srks12345 Very true, but fibre will be the future eventually! I still remember the Econet that BBC microcomputers used, and the BNC coax token ring networks with terminators on the end, running at 10Mb! (RM Nimbus networks with Winchester server!)
@@8skellerns Thats some pretty old gear :D I guess the only certainty is that in 50 years time, someone looking at the installation will think: they did things in such a quaint way back then
Probably because the voltage received at the input of the differential input receiver is out of limit and is too low. So the switch decides that whilst there is some kind of signal present is it out side of specifications and therefore there must be something wrong.
@@davepusey I mean, the other obvious other error type that could be detected is bit error. But the switch wouldn't need to shut down the interface on that port. A bit error being simple where a data bit has changed state. If using TCP protocol that would be detected, but not if using UDP. The switch itself wouldn't care about that, it would just increment an internal counter of the number of bit errors. And the TCP protocol handles it by packet retransmission.
What I don't see mentioned here is any reference to standard Electrical Engineering, specifically transmission line theory. The distance limit all boils down to one thing - attenuation. There is no such thing as a "digital" cable. Any electrical signal or power transmission over a wire is an ANALOG function. Computer and any other data signals are just an analog voltage pulse train resembling an FM radio signal. So you can run even 10Ge over CAT3 cable! Note I didn't sat anything about length. As long as the pulse train is not so distorted as to obscure the data recovery, it works. There are numerous factors that come into play here. The specific cable LCR paramaters, how many impedance mismatch bumps in the path (connectors), the quality of the sending and receiving circuit, the amount of RFI present. It's not a simple answer at all. However we can't expect the IT industry to consider all these minute electrical details on top of all their other higher level concerns. Just like we can't expect an Electrical Engineer to be a database developer. As In many industries, "rule of thumb" standards are developed to simplify selection of components including cable and connectors for systems integration. Such as "you need at least CAT6A for 10GE". That's a safe bet but in many cases these conservative rule-of-thumb standards can be challenged with success. I would strongly suggest any deviation form the standards be tested first before assuming it will work on a larger scale.
Attenuation really isn't the whole story. It's only half the equation, it's signal to noise ratio that is really important. The differences between CAT5e and CAT6a are mostly about reducing crosstalk which is a form a noise. When your using all 4 pairs to send data in full duplex you have pairs receiving data and transmitting data directly adjacent. The transmitting pairs have a very strong signal on them, while the receiving pairs are listening for a very weak signal due to the attenuation over the distance to the remote station(the switch in most cases). All that really matters is that the distant station's signal is stronger than the signal cross talk from your own transmitting pairs. CAT6a has additional spacing between the pairs to reduce this crosstalk. The noise floor also gets much worse when you have potentially hundreds of ethernet ports terminating into a single data rack. This is the enviroment the standards are built to work in. In residential applications where you have far fewer ethernet ports running in close proximity and over shorter distances, the standards are typically massive overkill. Another thing to keep in mind is the frequency components of typical noise sources are typically far lower than modern ethernet. So stuff like lighting ballast that would cause issues for something like 10mbit ethernet, are not really that much of a problem. It's other high frequency equipment, mainly other ethernet systems that are going to generate noise in the passband that 10gbit operates in. Also the lower frequency noise isn't going to couple well into the twisted pairs. Also stuff like wifi is far too high of a frequency to be an issue. Not many people are going to have high powered VHF transmitters near their homes.
I was looking for the answer after I bought a router and cat5e cable encluded with that. Now I got the answer with real life test without searching it.
The SMB file copies are a TCP transfer. Any packet loss at all would destroy the copy rate, because that's how TCP works, but I get a flat copy curve limited to the drives read and write speed. I also get zero errors on the Cisco switch end. If there is any packet loss, I cant work out how it's hidden.
Now if only we could get 10Gb equipment for the price of 1Gb equipment. Most of stuff doesn't even do 2.5Gb or 5Gb. Its been years and most everything is still 1Gb and even 100Mb equipment is still selling. Using Cat6a shielded for new cabling is somewhat economical now too. Just have to figure out what junk equipment to avoid. Lots of equipment use the same SERDES hardware with different protocols. Equipment could possibly even start using other protocols like external PCIe to chain and star between themselves.
Useful info, thanks for putting it up. When I moved in to my current place about 12 years ago, I took the view that I'd better take the pain early and go Cat6 because I'll probably upgrade at some point and I don't want to be pulling all the floorboards up again. It's nice to know that in an alternate universe, a cheaper version of me would've totally gotten away with it.
Yes, sort of. Ethernet has standards for 1, 2.5, 5 and 10Gits over twisted pair copper. There is no 2Gbit speed available. Cat 5e will support 2.5 and 5 Gbit connections out to 100 metres.
is there a difference between cat 5e and cat6, yes there is between two rooms in my apartment, around 15m, I had a run of cat 5e running gigabit - the iperf speed on that was ~800mbs; I then switched to cat6 and the iperf speed went to ~900mbs with the exact same hardware
10Gb link speed autonegotiation != 10Gb speed. Plus, the material of 5e cables - since these are the cheapest - may vary. Not sure if these were with solid copper or AL.
This isn't a home setup, it's a business. What you are looking at is just a patch panel where all the ports on a floor go back to get patched to switches for network connectivity. There are two PC's sitting on top and some Synonolgy NAS' sitting on top of them. It's pretty messy but that was the kind of business it was.
I think the file transfer test was to demonstrate successful packet transfers using 10Gbit connection. In other words, the 10g connection is stable. Real world you will never see 10Gbit file transfers from one machine to another due to hardware limitation (SSD speed, RAM size and speed). The file transfer can only happen as fast as the uploading machine can Read from its disk and the reciveing machine can Write to its disk.
@@maxbianculli4662 If i buy a gigabit hardware and can only transfer files over 10 or 100mbit/s ethernet, then i would assume my money was wasted. If i spent a lot to get a 10gbit/s hardware and my tests were measuring 5gbit/s i wouldn't declare it a success. Even knowing that 5gbit/s is very fast. For example, i use thunderbolt to sincronize very large databases between 2 machines. Both running high end SSDs. As my network-over-thunderbolt connection tops out at 10gbit/s, i can copy these files at 1GB/s between the computers. Real world. Any normal nvme ssds can get 3GB/s reads and writes, thats 3 times faster than network. A good pci 4.0 ssd can get 7 or 8 GB/s I may confess i have never used a 10gbit/s network, but i would really expect a troughput of 1GB/s.
Yes! Max Bianculli has the correct explanation. The copy was from a SATA SSD to network storage. The theoretical fastest my SATA drive could go is 550MB/s and I was getting ~460. Also, the 10GBit NICs I was using don't have 5 and 2.5 GBit connectivity options. It will only do 10 or 1. The copy is clearly faster than 1 so the link was working at 10.
@@matthewdaley7535 When benchmarking SMB/NFS/FTP create a ramdisk on both machines and copy files between them. When testing raw network you probably should use Totusoft LST or maybe spawn a few instances of HRPING and increase the flood before you start getting dropped packets. For more real-world-kinda testing I set up a ramdisk on "server" and ran CrystalDiskMark over SMB from the "client" to see if I could saturate 40Gbps ConnectX-3 but ran into CPU bottleneck first because setting up multi-threaded SMB was kinda too much pain and wasn't supported on TrueNAS (the intended purpose of server) anyway.
Unfortunately, in the real world, structured cabling is not perfect like your patch cable. 5e or any structured cabling a lot of times will have kinks, run too closely to electrical etc. These degrade performance, making the allowable lengths even shorter.
No. It'll run 1Gb over 100m at which point it degrades tro the standard Cat5 cable transfer speed which is Fast Ethernet (100Mb/s). You want 10G? You need cat 7 or 8.
Interesting, also I'd think those little cable couplers only degraded the signal even more.
Indeed, that was basically a worst of the worst scenario. Still, good results!
Those joiners can easily reduce your maximum cable length by 5m. I've seen some cheap ones which internally had about 5cm of untwisted wires connecting the two ends. Those ones at least say "CAT6" so hopefully they have a PCB inside, but still just the RJ45 connector is quite poor and will loose some signal strength.
One thing to bear in mind is Ethernet is rated for maximum distance over the specified cable. While CAT5 is rated for 100M at Gb, it can do much better at significantly shorter distances. Also, you mention the switch port shut down due to errors. I expect it was a significant amount of error and you were getting errors long before you reached that point. You have to properly measure the error rate, though you can get some idea by watching the error count. There is test equipment called "Bit Error Rate Test" (BERT) that is designed just for that purpose.
Do you really mean Cat 5 ( older and quite inferior to 5e) ? Or do you mean Cat5e ?
@@DavidM2002 Gigabit Ethernet was designed for CAT5 before 5e was available. So yes, plain CAT5. CAT5e is CAT5 with tighter tolerances. However, it won't make much difference, unless you're getting close to 100M. I have plain CAT5 in my condo, which was installed by my ISP in the late 90s (I provided the cable). They pulled it in with the coax when I got a cable modem. Also, they actually fished the cable through a wall, next to air duct, over ceilings, etc.. They even patched the drywall when they were done. None of that staple to the baseboard nonsense for me. 🙂
@@James_Knott Thanks James. I thought for sure it was a typo. I ran 5e in my little house about 20 years ago when I had the walls open doing a reno. And that was cutting edge for a house back then. So you were really ahead of your time. Well done.
@@DavidM2002 Well, most of my career has been in telecom, but at that time I was doing 3rd level OS/2 support at IBM Canada, so I would hope I'm a bit ahead of others. 🙂 Perhaps I should have said tighter tolerances, rather than specs.
BTW, my first work LAN experience goes back to early 1978, on a proprietary Rockwell Collins network, in the Air Canada reservation system. First Ethernet was with 10base5 "Thicknet" connecting VAX 11/780 computers and in the late 80s, I hand wired some Ethernet controllers, on prototype boards for Data General Eclipse computers. So yeah, I'm a bit more technically advanced than most people.
Straight to the point and conclusion. Much appreciated.
Nice video, short and to the point.
This makes me think of an analogy: can you drive an average car at over 100 miles per hour on the highway? Yes. But then come the details. Is it safe? Is there other traffic? Is it reliable? Driving fast has its risks, other traffic may cause you problems or you for it. Lots of factors come into play.
So be aware of the limitations of the overall situation, understand the interactions, the pros and cons, and work out the best solution. Then be willing to come back at some point to reevaluate it to see what needs to be changed to meet the requirements at that point in time.
My house is all cat5e and I've got it fully on 10GbE now. I did some research on this too.
Although standard does not qualify cat5e for 10GbE, but by cable performance characteristics cat5e should be able to carry 10GbE up to about 45m. However this is purely the cable, without the plug and socket. In real life conditions it would be something about half of that, so 25m is totally expected.
45 metres would be about my estimate for the 80 metre SFP+ modules or inbuilt ports which matches up with what you say.
My condo has plain CAT5 (not 5e) that was installed in the late 90s. It's doing fine with Gb, though I expect I would want to replace it, if I ever went 10Gb. That shouldn't be too hard, as the CAT5 was pulled in by my ISP, when I first got a cable modem, and so the cable is just lying loose and not anchored anywhere. Yes, it was properly pulled in, fished up a wall, along side air ducts, over ceilings and more, so that it's only visible where it crosses my laundry room ceiling. None of this crappy staple to the baseboards you'd get these days.
Let's say you have cat5e already built into the wall. If you extend it to a patch panel using wall plates and cat6 cable, would you still be limited to 45m total for 10GbE, or 45m only for the cat5e cable?
@@Mrperson0 It depends. Think of the signal in the wire is taking damage when travelling. For cat5e, it is like 10gbe signal can only live for 45m. Unless the signal reaches something that regenerates it, like a switch or router, it will not have any more HP gained going through anything else. Plus that transitions like sockets and plugs will do one time damage to the signal too. So it depends on your actual environment.
@@CCL13CN Ahh, thanks for the explanation! Makes sense that only a switch or router would regenerate the signal, not a keystone. Hopefully my 10gbe upgrade will go smoothly and I will be able to use a cat5e cable that is already installed to connect at that speed.
Worked in a DC that had generations of cabling going back to the late 90’s. Predominantly it was Cat5e so we’d just plug 10G into whatever happened to be in the patch panel, for cross connecting adjacent racks and whatnot it never presented a problem.
There is more than if the cable works or not. A user doesn't really care, but the manufacturer must meet all governmental requirement regarding radiated emissions (i.e. generation of electrical interference) In the US that's FCC class A (office) and class B (residential). A less superior cable will generate more radiated emissions up to the point of failing the test. The responsibility doesn't fall on the cable, but the manufacturer of the switch or SFP module. So to ameliorate this, the manufacturer will specify a higher grade cable.
Yes, manufacturers do have different problems to solve than the edge case users that are watching this video.
Cable Category most likely doesn't have a huge impact on emissions. I would bet a CAT5E S/FTP (Shielded, Foil Twisted Pair) cable has significantly less emissions than a CAT6A UTP (Unshielded Twisted Pair) cable.
@@bumbixp I'm going to assume that by emissions we are referring the industry standard referral of cross talk. External shielding only takes into account external traversal of cross talk, it does nothing for the internal cross talk withing the cable itself. For that we have the extra twisted in the twisted pairs within the jacket. In addition to that, you are missing the factor of impedance/resistance to voltage. Cat 5 is a thinner copper gauge of 24, Cat6 is a thicker copper gauge of 23. Therefore, Cat6 can go a longer distance due to it's greater copper mass, it has tighter twisted pairs to prevent cross talk within the jacket. THAT is what's important. The shielding only matters if you are running it next to areas prone to electrical interference. Aside that, the external emissions wouldn't make any impact of any kind to UTP.
Yup, my 2005 house was wired completely with CAT5e and I have several 10G links running over it that work just fine. I also have a 2005 era OM1 fiber connection from the main house to an additional unit on the property and I was able to run 10G over that by using LRM optical modules with MCP (Mode condition patch) cables.
So we had fun on a pre-rip-out floor at work. The cable under test was Molex DataCAT-5 installed OG in 1993. We TDR'd lengths from 20-90M in the building. We were using Molex OG patch leads. The switch under test was a Cisco 9300UX with the integrated copper interfaces capable of 10Gb. We managed a full 90M run - but the switch wasn't happy about it. Lots of errors - but no UDLD.. Runs of OG CAT-5 were happy-enough to 50-60M without errors.
But - it's all gone now - The Ripout sees Molex DataCAT-6 cable installed. That works nice to 90M without any worries.
Thanks for this, I have so much CAT5-E running around my building....so much work, so little time...
At the end of the day, if it works and you're satisfied with it's performance and any potential fallbacks/issues, awesome job in testing it out yourself! This is not too far off the swan song of running cheap vs expensive HDMI cables over short distances.
Thank you for the video! 2:00 that cabling is CATastrophic event caused by a previous installer.
Straight to the paint and no non-sense. You got something here if you want to grow the channel. Best wishes!
Thanks for doing the experiment. Lots of hype on cabling.
Great presentation. No bullshit, test and everything is clear. Well done!
Just discovered this channel....looks nice. I will stay surely.
Great to know, I have put off buying a 10gig switch, but looks like it's time to do some shopping LOL
I was expecting more technical analysis.
But it was interesting watch your video, and fun.
Thank you for this video.
Very helpful information! It's not so much a question of which cable, as "how far do you need to go?'
Wonderd about that - many Thanks!
Using all those ethernet couplers likely reduced your distance significantly, as well.
I have done this with short runs, but there are a lot of caveats. If you want a reliable connection you can depend on, use 6a.
That said, solid core cabling fares better than stranded in terms of signal quality, and if you can reduce or eliminate jacks or couplers, that will help. If the wire is going up into a hot attic, the impedance of the wire will change and may make an otherwise acceptable connection turn marginal or unacceptable. Had this issue with a 50ft unsplit run - was great all winter and spring, until the first hot day of summer when the switches disabled the ports for link flap.
Quick and informative!
big question !!!! thanks for answer it
Thanks for the video!
This is brilliantly useful to me as i'm about to move house to one that's cabled with cat5e everywhere. I'll let you know how I get on.
I'm going to assume that by emissions we are referring the industry standard referral of cross talk. External shielding only takes into account external traversal of cross talk, it does nothing for the internal cross talk within the cable itself, where the twisted pairs are. For that we have the extra twisted in the twisted pairs within the jacket, thus 6 has more twisted pairs. In addition to that, is the factor of impedance/resistance to voltage. Cat 5 is a thinner copper gauge of 24, Cat6 is a thicker copper gauge of 23. Therefore, Cat6 can go a longer distance due to it's greater copper mass, it has tighter twisted pairs to prevent cross talk within the jacket. Shielding only matters if you are running it next to areas prone to electrical interference. Aside that, the external emissions wouldn't make any impact of any kind upon unshielded twisted pair "UTP". So Cat5 WILL work with 10gb or 2.5gb. With CAT5 on 10gb/2.5gb your speed to distance ratio will lower the further the cable is, you will produce and be susceptible to more emissions/cross talk, within the jacket. However, it will work and will be a none issue to home owners with Cat5/e.
Here's the wisdom of the day: You can run every signal over every cable, as long as the cable is short enough. Better shielding (or shielding at all), a bigger diameter and cable twisting will only extend the range the signal can travel and still have good quality but they never decide if you can run a signal over a cable at all. You can run Gbit Ethernet even over totally unshilded, untwisted, thin phone wires, as long as you only need cross a gab of less than a meter or occasional signal errors are okay (Ethernet will detect them and drop bad packets, higher protocol layers will then resend missing data that didn't make it). High quality plugs and cables only give you longer ranges and less signal errors, they don't enable you to do something you could not otherwise do.
Exactly. I've reused lots of old untwisted 4 wire phone cables and they are working fine for 100mb and even 1gb when using two 4 wire phone cables in one rj45 connector for shot distances.
It would've been useful to show whether any errored frames were recorded. By the time a port is error-disabled the error rate is pretty high. Perhaps there were errors not regular enough to affect data transfer speeds, but irregular packet loss will be bad for online gaming and streaming.
Then there is the potential problem of alien crosstalk.
In other words the risk of multiple cables in a bundle, disturbing each other.
Maybe not a big problem in a residential network, but definitely worth considering in a larger commercial system.
C6a components are much less susceptible to that problem.
In fact, the reason many U/UTP C6a cables are so bulky, is very deliberate.
It’s simply just to separate the pairs in one cable from the pairs in the next one.
Just to minimise alien crosstalk.
Balanced signalling and differential input receivers are your friend...
It really depends on the quality of cables as in how strictly have the cables been tested for.
The monoprrice slimrun CAT 6A UTP claims 550Mhz tested but lengths as short as 10 feet have failed to give 10Gbps. It falls back to 5 Gbps. There'sa video on youtube showing that if you do a quick search.
As a tech and network nerd, I could confirm this is one of the most entertaining scientific experiments ever😂
Thank you!
That's good news! : )
One question I had but I never bothered to test is: If you plug multiple antennas into a wifi access point designed for MIMO but you separate them by a fair distance, do you get better coverage? E.g. one access point with 2 antennas 10 metres apart vs 1 access point in the middle
GOOD to know this!!
I have high hopes for an old ~37m run of CAT6A with a pair of $36 30m transceivers (which Amazon has lost) when I do eventually get them. If not, I'll run fiber before spending the money on high power Cu transceivers.
I'd say you have a very good chance of it working just fine. Please comment back once you give it a go. I'm interested to know if your result replicate mine.
i bought cat 8 ugreen cables from amazon after getting 10gig to my home , i should test it aswell
Thanks for sharing ;)
As an expert in this field.
It can negotiate at that speed but don’t expect it to actually perform consistently at that speed.
Cross talk, packet drops and so forth are a very major reason why it’s not rated for 10gb.
Cat6A and up is rated because the wire is thicker and so is the insulation. This greatly reduces cross talk, RF interference and allows the connection to run far more stable.
I thought the main difference was tighter twists and tighter manufacturing tolerances?
@@shaunclarke94 CAT5 usually uses 24AWG (0.52mm) wires whereas CAT6A and good CAT6 uses thicker 23AWG (0.57mm) wires.
@@Daniel15au caT5e is rated from 30awg to 26 typically. It's quite thin and so is the insulation, that's why it's so prone to RF interference and why it's not rated beyond 1gbps and even at 1gbps it can be flakey.
The SFP+ “30M” rating on those modules is more to do with the power draw/heat dissipation of the module. SFP+ isn’t normally meant to run 10Gb copper Ethernet, as the SFP+ spec is very power budget limited, and getting 10Gb to go for 100M on ANY cable requires quite a bit of power per port.
Yes, with Cat5e cables, you can still achieve a maximum speed of 1 Gbps. However, over short distances, it is possible to get slightly higher speeds than 1 Gbps, although it may not be consistent or reliable
What speed will the router work on if you have a 2.5 gbs switch with a 10 gbs sfp+ port plugged in a router with a 10gbs port. The speed of the data is 1.6 gbs. My network card is intel x540-t2 and detects the 10 gbs network if I plugin the cable directly to the card but if the network is passed through my 2.5 gbs switch the card detects it as 1gbs only? My question is If I buy an sfp+ 10 gbs rj45 and plug it in the 2.5 gbs switch in the 10gbs port will it make a difference. i just realised that the intel x540 runs only at 1 gbs or 10. Not 5 or 2.5. Thanks
I think I found out that there will be 10gbs on that port but internally will be at 2.5gbs on the other port. That's why I will have input at 10 and gave a 10 going to computer and the rest at 2.5 or 1 gbs
Very interesting...
Good to know.
Thank you for this video!
No surprise if the answer is Yes, it can. Because all of 4 pair of twisted pair is connected, the pin connection is intact, and the auto negotiation on link speed is acknowledged.
The real problem is stability. That's why Cat 6 cable has tons of shielding element, even the rj45 jack is shielded, and got bigger thickness. The stability got worsen with every port used, because of signaling interference.
I've got enough of stress when setting up Gigabit switch with 24 ports onward, it break down the whole connection.
🎯 Key Takeaways for quick navigation:
00:00 🕵️♂️ You can use Cat5e copper cabling for 10 gigabit Ethernet, but the recommended cable is Cat6a for 100 meters and Cat6 for up to 55 meters.
01:10 🧪 Testing with 30-meter SFP+ copper modules showed that 25 meters is a practical limit for Cat5e cabling for 10 gigabit Ethernet.
01:51 💻 For practical home use, Cat5e cabling can likely handle 10 gigabit Ethernet without issues for considerable distances.
Made with HARPA AI
Hello, at 2:01 minute mark, there are 2 NAS 8 bay stacked. I would like to know the model. I think they are Synology DiskStation DS1821+ units, but it's not very clear.
Thank you.
You are very close, they are a little older DS1819+'s. They dont have 10Gbit interfaces which made them only useful for archive.
@@matthewdaley7535 You can buy 10gbps card into these if I am not wrong...
I`m not sure if this is a non European thing, but why does everyone seem to use UTP on Cat5e and never consider SFTP / FTP / STP ?
I know the UTP versions are cheaper but still the shielding could help somehow to reduces errors especially if there is shielding between the wire pairs.
Also the couplers will add additional resistance on the wire and so the connection attenuation will go up.
But still, good idea to put a video up telling people Cat5e could transfer 10GBe, if the conditions are right.
stp is more work to terminate plugs etc.
I think it is because parts of the US still use old and noisy electric infrastructure running on on 110v, whereas all of europe is on the same modern 220v grid and it is 'cleaner' hence electrical noise is less of a problem, hence less need for shielded pair. I might be wrong, but that's what I always thought.
Plus, the British also have the best plug and socket system ever designed!
also twisted pairs are used because of differential signaling which cancels out noise induced on the lines. only if the noise amplitude is greater than a certain threshhold would you need extra sheilding. Something like an electric fence can be problematic for external network cables
UTP was the standard for telephone lines and slowly evolved towards today's CAT6a. The tools and techniques for working with UTP all came from the phone companies. Even before networking became a common thing, CAT5 was being run for commerical phone systems. So when networking started to take off, it was convenient to use that existing cabling.
Ethernet was engineered to use this cable because it already existed and over the short distances(under 100m) found in most buildings, it was more than good enough. Modern CAT6a just improved earlier systems while maintaining backwards compatibility.
For the shielding to be effective the equipment at both ends of the link really need to be designed for it and most stuff is not.
The shielding can create ground loops which can be a major problem. Which is likely the major reason the telephone systems avoided it in the first place.
Really makes me think though that we should be starting to wean ourselves off copper cables. Run fibres in the walls?
Why though? Fiber is very expensive for the equipment, hard to make in your own lengths and would need to be adapted to copper/implemented with add-in cards for just about any device that you would connect to it. Its a lot of added expense when cheap copper will future-proof you for a while. This person runs 10G on cat5e on in-house lengths, and I expect that there is nothing you actually use the internet for that will come near to saturating the 10G connection. Most storage would bottleneck the actual speed
@@srks12345 Very true, but fibre will be the future eventually! I still remember the Econet that BBC microcomputers used, and the BNC coax token ring networks with terminators on the end, running at 10Mb! (RM Nimbus networks with Winchester server!)
@@8skellerns Thats some pretty old gear :D I guess the only certainty is that in 50 years time, someone looking at the installation will think: they did things in such a quaint way back then
where is conclusion?
I'd be interested to know the exact reason the switch gives in it's logs for err-disabling the port.
Probably because the voltage received at the input of the differential input receiver is out of limit and is too low.
So the switch decides that whilst there is some kind of signal present is it out side of specifications and therefore there must be something wrong.
@@deang5622 Hmm, I guess that could constitute "port flapping" if it keeps dropping the link too rapidly. Makes sense.
@@davepusey I mean, the other obvious other error type that could be detected is bit error. But the switch wouldn't need to shut down the interface on that port.
A bit error being simple where a data bit has changed state. If using TCP protocol that would be detected, but not if using UDP.
The switch itself wouldn't care about that, it would just increment an internal counter of the number of bit errors. And the TCP protocol handles it by packet retransmission.
What I don't see mentioned here is any reference to standard Electrical Engineering, specifically transmission line theory. The distance limit all boils down to one thing - attenuation. There is no such thing as a "digital" cable. Any electrical signal or power transmission over a wire is an ANALOG function. Computer and any other data signals are just an analog voltage pulse train resembling an FM radio signal. So you can run even 10Ge over CAT3 cable! Note I didn't sat anything about length. As long as the pulse train is not so distorted as to obscure the data recovery, it works. There are numerous factors that come into play here. The specific cable LCR paramaters, how many impedance mismatch bumps in the path (connectors), the quality of the sending and receiving circuit, the amount of RFI present. It's not a simple answer at all.
However we can't expect the IT industry to consider all these minute electrical details on top of all their other higher level concerns. Just like we can't expect an Electrical Engineer to be a database developer. As In many industries, "rule of thumb" standards are developed to simplify selection of components including cable and connectors for systems integration. Such as "you need at least CAT6A for 10GE". That's a safe bet but in many cases these conservative rule-of-thumb standards can be challenged with success. I would strongly suggest any deviation form the standards be tested first before assuming it will work on a larger scale.
Attenuation really isn't the whole story. It's only half the equation, it's signal to noise ratio that is really important. The differences between CAT5e and CAT6a are mostly about reducing crosstalk which is a form a noise. When your using all 4 pairs to send data in full duplex you have pairs receiving data and transmitting data directly adjacent. The transmitting pairs have a very strong signal on them, while the receiving pairs are listening for a very weak signal due to the attenuation over the distance to the remote station(the switch in most cases). All that really matters is that the distant station's signal is stronger than the signal cross talk from your own transmitting pairs. CAT6a has additional spacing between the pairs to reduce this crosstalk.
The noise floor also gets much worse when you have potentially hundreds of ethernet ports terminating into a single data rack. This is the enviroment the standards are built to work in. In residential applications where you have far fewer ethernet ports running in close proximity and over shorter distances, the standards are typically massive overkill.
Another thing to keep in mind is the frequency components of typical noise sources are typically far lower than modern ethernet. So stuff like lighting ballast that would cause issues for something like 10mbit ethernet, are not really that much of a problem. It's other high frequency equipment, mainly other ethernet systems that are going to generate noise in the passband that 10gbit operates in. Also the lower frequency noise isn't going to couple well into the twisted pairs. Also stuff like wifi is far too high of a frequency to be an issue. Not many people are going to have high powered VHF transmitters near their homes.
I was looking for the answer after I bought a router and cat5e cable encluded with that. Now I got the answer with real life test without searching it.
What about packet loss tests though?
The SMB file copies are a TCP transfer. Any packet loss at all would destroy the copy rate, because that's how TCP works, but I get a flat copy curve limited to the drives read and write speed. I also get zero errors on the Cisco switch end. If there is any packet loss, I cant work out how it's hidden.
Now if only we could get 10Gb equipment for the price of 1Gb equipment. Most of stuff doesn't even do 2.5Gb or 5Gb. Its been years and most everything is still 1Gb and even 100Mb equipment is still selling. Using Cat6a shielded for new cabling is somewhat economical now too. Just have to figure out what junk equipment to avoid. Lots of equipment use the same SERDES hardware with different protocols. Equipment could possibly even start using other protocols like external PCIe to chain and star between themselves.
Useful info, thanks for putting it up. When I moved in to my current place about 12 years ago, I took the view that I'd better take the pain early and go Cat6 because I'll probably upgrade at some point and I don't want to be pulling all the floorboards up again. It's nice to know that in an alternate universe, a cheaper version of me would've totally gotten away with it.
How about internet fibre ?
Cat 5e capable of running 2gbps ?
Yes, sort of. Ethernet has standards for 1, 2.5, 5 and 10Gits over twisted pair copper. There is no 2Gbit speed available. Cat 5e will support 2.5 and 5 Gbit connections out to 100 metres.
If you have Good Decent Cat5e, you should be good.
is there a difference between cat 5e and cat6, yes there is
between two rooms in my apartment, around 15m, I had a run of cat 5e running gigabit - the iperf speed on that was ~800mbs; I then switched to cat6 and the iperf speed went to ~900mbs with the exact same hardware
Could be terminations, 5e does gig for 50m no issues.
10Gb link speed autonegotiation != 10Gb speed.
Plus, the material of 5e cables - since these are the cheapest - may vary. Not sure if these were with solid copper or AL.
technically the answer is yes. But you should use a better grade to avoid crosstalk
What do you use this server stack for? It's massive!
This isn't a home setup, it's a business. What you are looking at is just a patch panel where all the ports on a floor go back to get patched to switches for network connectivity. There are two PC's sitting on top and some Synonolgy NAS' sitting on top of them. It's pretty messy but that was the kind of business it was.
@@matthewdaley7535 I see. Looks like it's a home-run business 🙂
yes
But your test is getting a little less then 5gbit/s (460 MB/s). Was you able to get 10gbit/s?
I think the file transfer test was to demonstrate successful packet transfers using 10Gbit connection.
In other words, the 10g connection is stable. Real world you will never see 10Gbit file transfers from one machine to another due to hardware limitation (SSD speed, RAM size and speed). The file transfer can only happen as fast as the uploading machine can Read from its disk and the reciveing machine can Write to its disk.
@@maxbianculli4662
If i buy a gigabit hardware and can only transfer files over 10 or 100mbit/s ethernet, then i would assume my money was wasted.
If i spent a lot to get a 10gbit/s hardware and my tests were measuring 5gbit/s i wouldn't declare it a success. Even knowing that 5gbit/s is very fast.
For example, i use thunderbolt to sincronize very large databases between 2 machines. Both running high end SSDs. As my network-over-thunderbolt connection tops out at 10gbit/s, i can copy these files at 1GB/s between the computers. Real world. Any normal nvme ssds can get 3GB/s reads and writes, thats 3 times faster than network. A good pci 4.0 ssd can get 7 or 8 GB/s
I may confess i have never used a 10gbit/s network, but i would really expect a troughput of 1GB/s.
Yes! Max Bianculli has the correct explanation. The copy was from a SATA SSD to network storage. The theoretical fastest my SATA drive could go is 550MB/s and I was getting ~460. Also, the 10GBit NICs I was using don't have 5 and 2.5 GBit connectivity options. It will only do 10 or 1. The copy is clearly faster than 1 so the link was working at 10.
@@maxbianculli4662 How many errors occurred?
@@matthewdaley7535 When benchmarking SMB/NFS/FTP create a ramdisk on both machines and copy files between them.
When testing raw network you probably should use Totusoft LST or maybe spawn a few instances of HRPING and increase the flood before you start getting dropped packets.
For more real-world-kinda testing I set up a ramdisk on "server" and ran CrystalDiskMark over SMB from the "client" to see if I could saturate 40Gbps ConnectX-3 but ran into CPU bottleneck first because setting up multi-threaded SMB was kinda too much pain and wasn't supported on TrueNAS (the intended purpose of server) anyway.
Cat5e transmits up to 1gbps. [Not ideal for gaming]
Cat6 transmits up to 10gbps. [Ideal for gaming]
If you turn off every electro smog devices you can run over 100m. If you fit it along power cables, you're not going to get 5m
Of course you can... lol
1:36 "Copper... Optical Transceiver Module" - what a nonsense...
Haha, I never noticed that before.
Your Cable Joiners are the biggest Problem. Without them with a 1:1 30m Cable everything should work fine.
Unfortunately, in the real world, structured cabling is not perfect like your patch cable. 5e or any structured cabling a lot of times will have kinks, run too closely to electrical etc. These degrade performance, making the allowable lengths even shorter.
Huh.
No. It'll run 1Gb over 100m at which point it degrades tro the standard Cat5 cable transfer speed which is Fast Ethernet (100Mb/s). You want 10G? You need cat 7 or 8.
I've seen 2.5gbe work at 50 meters without any problems on cat5 and 10gbe on 10m, without any issue.
THIS SHOULD HAVE BEEN A CZcams SHORT!
That's a good idea but I have no idea how to cut it down from 2 minutes 15 seconds to under 1 minute without losing too much content.
I hate CZcams Shorts. Much better longer.
I hate shorts and just block the category. Why video in portrait mode when most watch with a landscape mode monitor ?
@@DavidM2002 Most videos are video's in wide-screen, but CZcams Shorts crops it to vertical video. Vertical video is a pet hate for me!