The modules have only PHYs to covert typically XFI to xBaseT. The newer PHYs tend to support NBaseT, where the older ones did not. The newer PHYs 'should' also be slightly lower power. For performance, it's almost SOLELY based on cable performance, how well it can get the signals down the wire and reject CM. As for other things like Jumbo frames, typically PHYs don't discriminate anything over L1 (some exceptions apply). Going back to the host interface, another thing to look at is if the modules solely support XFI using flow control to support the lower speeds on the line interface, or if they support USXGMII, which is also a 10G interface (the same interface actually @ 10.3125Gbs & 64b/66b), but support the lower rates by multiplying the data (like how SGMII runs 100Mbs, still runs at 1.25Gbs, but duplicates 10x to slow it down). I would have been very interested in knowing which host interfaces these have, and for the future the same applies to switches with SFP+ (XFI only, or USXGMII). The final part you're missing is cable performance, you should try short and long cable, and also see what the limit is. Thanks & cheers
This has rapidly turned into my favorite tech channel on youtube - no hype, all fact, but still entertaining and interesting. Keep up the Lord's Work, Patrick!
Worth noting that the reason the Mikrotik S+RJ10 use a little more power than the other adapters is because they are enhanced range adapters, if you have one at either end then the cable run can be up to 200m for 1GBe or 2.5GBe (instead of the regular 100m).
"Do they support jumbo frames?" - the only valid answer to that question is "No, they don't.". They also don't support VLANs or BGP or HTTP/2, because they are Layer 1 devices and don't care what the bits they transport actually mean.
That was my initial thought as well, however jumbo frames could potentially be an issue. These contain PHY chips that could potentially do some retiming and idle insertion/removal. Jumbo frames mean there are fewer opportunities to insert/delete idle control characters between frames as the frames are 6x longer than 1500 byte MTU frames, and as such jumbo frames could possibly cause some issues. As for the other stuff, that's just data in the frame, so it's irrelevant from the standpoint of the PHY chip. The bigger question is why is there any difference in speed at all, since these are physical layer devices.
@@JasperJanssen They are PHY layer devices, SFI/XFI on one end and 10GBASE-T MDI on the other end. The interface on both sides of the device is physical layer, which must operate within around 100 ppm of the specified line rate, otherwise the link will not work at all. Unless this is injecting pause frames or something towards the switch (which they should not be, but it should be possible to rule this out), the only possible cause for a speed discrepancy is packet loss, either due to signal integrity or differences in the FEC decoding. 10GBASE-T uses an LDPC FEC, and it's possible that some of the modules do not implement the LDPC decoding correctly/completely, resulting in more bit errors and hence more dropped frames on some modules. What I would like to see is a dump of the switch packet counters (tx, rx, fcs error) from both ends of the link during these tests, that may shed some light on the discrepancy.
i know Im asking the wrong place but does any of you know of a tool to get back into an instagram account? I stupidly forgot the account password. I love any help you can offer me.
@Ledger Maximiliano I really appreciate your reply. I got to the site on google and Im in the hacking process atm. Seems to take quite some time so I will reply here later when my account password hopefully is recovered.
Thanks for this video it helped me with a purchase of a Wiitek SFP+ to RJ45 adapter. It worked fantastically in my Netgear 10g switch going to a QNAP 5gb USB adapter which Netgear said couldn’t be done with my switch but since the negotiation seems to happen in the adapter everything works great.
Thank you so much for your well done reviews! Thanks to you I bought a CSS3226 with a Mikrotik SFP+ to RJ45 10GBase-T Adapter a couple weeks back and I couldn't be happier! :)
Great review, the only addition I would consider here is choose lower wattage over everything else, especially if the longer term goal is to populate most of the ports. 20x 10GBase-T ports at .4 watt more sure does add up in heat and power draw. Also for the more serious consumers, get fibre where possible as the power draw is considerably less. Also fibre has much lower latency than 10GBase-T 6x10Gbase-T ports has a latency of ~15 opposed to 0.6 with fibre. On my router 10GBase-T port reports ~54c and the 10GSFP SMF fibre reports 32c. that's a whooping 20+c per module difference, thankfully I only have two Base-T modules the rest are fibre.
Great comparision! One thing you could have measured is latency over these adapters. If I've understood correctly, high quality SPF+ to 10GBase-T adapters should cause about 25 µs extra latency over fiber and it would be interesting to see if the cheapest models cause more.
Also a nice information i figured out the hard way. They are Hard capped at 30M wire length if the other device on the other end is a low power device like the Asus U2008 10g Switch. If you use an INTEL NIC on the other end with a 100M power output capacity it works fa some longer ranges about 40-50M. But that is one thing to consider when using these module with other low power devices on the other end that you may not be able to wire through your entire home. I would like to see some more imperial testing on your end if you could, i can just speak for one or two runs i made through my home which exceeded 30m.
My own tests show 45 meter is the maximum (with Cat5e cables). My longest Cat5e cable here in my house is 30 meters and works perfect with 10Gbase-T. Off course when you are a little less Scrooge, you can use Cat6a.
The generation of PHYs used within these modules, would draw too much power from the cage if they supported over ~30M. We should start to see that change in the coming future.
The reason they note the ‘Cisco’ model number on their advertising is because allot of Cisco switches won’t accept ‘non-Cisco’ SFP modules. They just refuse to ‘enable’. After some protest by (home) users they developed a config option to make them work. They still generate warning and error logs in the device. At least, that’s what was happening 5 years ago (when I did networking). We even had a supplier that flashed specific Cisco firmware on the SFP to avoid warranty discussions or deployment issues.
Just FYI - I got some of the Wiitek 10G-BaseT SFP+ modules today, they've got an apparent manufacturing/configuration date of 2021-03-30, and come with country of origin (Made in China) markings on the box and module. As for temperature, after a few minutes they're as warm as the 10G-BaseT SFP+ modules I got from FS - reporting about 50-54C in my Mikrotik CRS309-1G-8S+
We were recently troubleshooting an issue in a Cisco switch and we noticed one of our modules was throwing an overheat warning in the logs. We focused on that module and it turned out we had inadvertently installed a cheap module that came with some other hardware in the Cisco switch. There ended up being nothing wrong with the module, other than it did not support temperature reporting to the switch, so it was making a lot of log entries that were distracting. So that's another feature you need to be on the lookout for, is will it report temperature to a switch that wants temperature information.
Food for thought, if you plug a Cisco compatible "10GbaseT" SFP+ into a Cisco switch and run "show inventory" or "show interface status" it will come up as a SFP-10G-SR= module, a multimode LC fiber transceiver. Yeah it's spoofing the switch with another model and probably won't matter in the home but if run in the enterprise environment, it's not technically "supported" but I doubt Cisco TAC will even find out.
They will when you give them the "show tech" output. (if they look, and their scripts do.) At worst, they'll tell you to put "genuine" cisco optics in it. (which is of course bunk, as Cisco doesn't make optics... they have manufacturers add the required signature to the flash. And there are 64 possible signing keys -- the code's been known for a long time.)
Thanks for the great info.. I went with Hi-fiber. Google Fiber Router to a (Brocade)icx-6450-48poe switch. 10g no issues. Then I got a Zyxel multi-gig unmanaged to have 2.5g to two desktops. On the Google router it has 10g RJ45. So it went from RJ-45 -- > cat6a --> SFP+ on the brocade. with no problems. HI-Fiber SFP worked on both.. the Zyxel model XGS1010-12. Hope this helps someone.. Again thanks for the video.
Why no 10GTek? Used the 10GTek for the fiber install in my house last month. Just a couple runs between the NAS and a couple PCs (100ft each). Was able to to get 1.1GB transfer rates.. Pretty happy with them. (using 5 Port SFP+ Mikrotik switch). Amazon has amazing deals on fiber. Cost was very reasonable.
This video is awesome I haven't seen good concise comparisons of 10Gb/Cu SFP+ anywhere else. Surprised that the mikrotik is the least energy efficient.
The reason the Mikrotik S+RJ10 use a little more power than the other adapters is because they are enhanced range adapters, if you have one at either end then the cable run can be up to 200m for 1GBe or 2.5GBe (instead of the regular 100m).
I know I'm late to the party on this, but those Cisco adapters did, at one point, exist, then they suddenly marked them EOL and seemingly scrubbed them from their site. I was looking to buy some at one point for the 12 port 3850s we were looking at for our office, but they were persona non grata in the middle of my research.
I would be curious to know what transceivers work fully on Cisco switches, namely in terms of if it can check temperatures and other various stats, properly
I love Mikrotik routers - but their switches, I've had so many of them cook on me in the data center. I've changed over to Juniper - but that can get a little pricey for the home.
It's funny how often network brands get pigeon holed for one segment. I love Routerboard as well but have been apprehensive to try the switches. I tend to go Mikrotik router if NGFW is not needed, Aruba or Unifi switch and Tplink EAP or UAP for wireless because nobody does them all to my liking at least not the budgets I prefer..
At some point I am going to upgrade my switch to something with a QSFP+ port or two and they all seem to be either all RJ45 or all SFP+ for the rest of the ports. I'd rather go SFP+ for the rest to be uplinks to smaller switches elsewhere.
In my estimation, the difference in power dissipation was probably more to do with variances in component tolerances. Keep in mind that many of these "cheap" off-brand SFP+ units are likely built using the reject IC's from an assembly line that supplies IC's for a major brand name like Cisco/HP/Etc, so while they are still functional, some are over certain thresholds to go into the good bin. With regards to the fact that they appeared to be grouped up as 6 and 3 with similar functionality, I'd be willing to be that this test of "9" SFP+ modules was really a test of 2 modules with 9 different brands slapped on them.
As mentioned in an earlier comment, that was something that we thought at the outset. It turns out all nine modules had different PCB/ markings near the exposed part by the SFP+ connector. The chipsets may be similar, but at least the PCBs are different.
@@ServeTheHomeVideo Aye! Either way great work on these reviews! STH is one of the best resources out there for us homelab and small/medium business IT guys. Your testing of EPYC Rome lead me to 7402P's on 2113S-WTRT's for our hyper-converged server cluster at work. (We're running proxmox w/ceph). Works beautifully!
A transceiver works at the physical layer (Layer 1 OSI model) and frame size (hence jumbo frames) is a property of datalink layer (Layer 2 OSI model), right? If that is the case, I cannot imagine how any transceiver could be capable of not supporting jumbo frames.
The vendor compatibility advertising is related to the fact that some check the firmware or use specific options. You did not mention if you checked them with different switch vendors/generations? Which switches do need coded modules?
Great idea. ConnectX-3 Pro cards are inexpensive these days for 40GbE. Usually, those are way overkill for home networks. 10GbE is a lot less expensive so we generally recommend that these days. If you think of anything you think we should cover, feel free to comment, we can see what Rohit can do.
Good review. Would have been nice if you could have done some distance testing. Specifically the ones that claim to be greater than 30m. I'm sure that would be a huge selling point if true.
What about long term reliability? In some cases they may be in a "plug it in and forget about it" type environment. I would be curious to see how they worked long term.
It is somewhat hard for us to do a longer term reliability piece on these. After 2-4 years we would expect a newer generation to arrive making the current generation obsolete by the time we get that done. We did hammer them for a week and did not see any issues, but that is not exactly "long-term."
@@ServeTheHomeVideo My colleague and I were discussing this video and he brought up a very good question. We are using a company for a dark fiber backbone for all of our schools. He has found that we've had issues when the same SFP was not being used on the same ends of the fiber. Did you do any compatibility testing to see if one brand of SFP would work with a different SFP on the other end? Or did you always use matching SFP's on both sides? My colleague found that non-matching SFP's would cause comm errors, because of subtle differences between manufacturers. Thanks.
@@ServeTheHomeVideo So all the brands worked well with not being matched? I'm not sure I was clear about the branding test. 30m spec is fine. But does brand A and brand B work together?
I suppose the one key element yet to be seen is their reliability and longevity... That's going to be the most difficult metric for mass imported goods from no-name manufactures. Perhaps there are a handful of legit ODMs overseas churning them out and they are just rebranded. But who knows... These are fantastic for a home setup or lab, but I wouldn't use them for critical datacenter applications because of potential risk of failure and lack of resolvable warranty.
It'a a common knowledge that those "cisco" or other claims on listings are ment to says for what vendor it can be encoded. It's actually a good thing, it makes it easier to find whatever you want. So there is no myth, it's common knowledge ;-] From all of the people you should know that.
The 3 slower ones are probably based on BCM chips, while the faster ones are AQC. The AQC chips are faster because they do less DSP and hence has shorter latency, while the BCM ones do more DSP and introduce more latency. The BCM chips do more DSP not for no reasons. The BCM chips are rated for 80 meters over CAT6A, while the AQC chips are rated for 30 meters. Test with longer cables and see if this is the case.
@@btudrus No. Processing costs something, in this case, silicon area and power consumption. First (AQC10x) gen AQC is hot because they used an older process node due to lack of funding, just like many startups. The second (AQC11x) generation should be better in this regard.
Thank Gerhard. We mentioned in the video that we are going to keep the main site guide updated with new modules as we get them. We simply got the top 9 results on Amazon for this project and the UniFi UF-RJ45-10G was not there. Hopefully we can get that and a few more as well.
@ServeTheHome though FS.com modules aren't available on Amazon, I would never buy a module from Amazon I do only buy my modules from FS.com. Those modules were testing work at best at 30m max. The FS.com ones, 80m (Lawrence Systems got his working at 90m).
Did you try them on longer runs? I tried two of the brands on my 15m cat7 cable, none of them worked(on both 10g/1g). Even though they all advertise 35m on 10g and 100m on gigabit.
As in 2024, I'm using ADOP's new 10G module which supports 1/2.5/5/10G speeds. They also claim that new module may connect up to 80M in distance if the cable condition is good (perhaps on CAT7 or CAT8).
The Transceiver market has been experiencing a long deserved correction for the past couple of years. The pinnacle of this development is what Flexoptix has been doing. Sure, in an enterprise setting, it's most of the time not feasible because you can easily lose support for the whole solution just because you're running third party transceivers. But in a small business setting, or in a homelab, they're a total gamechanger. I've installed at least 200+ transceivers, DACs and AOCs from Flexoptix for dozens of customers at this point, and they've always worked flawlessly. The programmability is key. But in most devices, they also work out of the box. Which is very fortunate, because the FlexBox programmer that is used to program transceivers, DACs and AOCs cost an absolute fortune. If you've got it, though, you can install thousands of transceivers for each about 1/10th of the cost of the original product (for 10G SFP+ SR transceivers).
Just out of curiosity, will this work well in an Aruba switch? I've got some fiber-optic SFP+ modules for my Aruba 3500 switches but sadly, I have an issue with one of my runs of fiber and it looks like I'll need to change the links between two switches from fiber to copper.
We thought about that. The problem we saw while doing this series was that there were modules we purchased at $35 then they would be $55 two weeks later, $48 when we did the review and so forth. They moved so much we stopped putting that information in reviews. That is why we recommend just searching whenever you need them.
What about all of the issues I have seen online about vendor locking the firmware so they won't work in some manufactures equipment. Like putting an off brand in a Cisco switch...????
Hi Martin, we used 20m and a lower quality cable which is mentioned in each individual review's testing. We actually later tried 30m and they were all working for a week hammered with traffic. Our sense was basically that the 30m rating was safe. You are right that we did not test beyond spec there.
My question is which is more efficient/faster/reliable for converting between 10GBase-T and 10G DAC : (a) A 10GBase-T cable plugged into an SPF+ transceiver then plugged into a slot on a multi-slot, multi-gig switch? _OR..._ (b) A 10GBase-T cable plugged directly into a *_dedicated_* 10GBase-T / SPF+ media converter? It would seem that the dedicated media converter *_should_* win every time, as it has no MAC address lookups to make; it's simply what goes in one port has to come out the other. But I know of no actual testing data or empirical evidence to support this. My reason for asking is that our company just got AT&T 5Gb/s service that comes in over the infamous BGW320's Ethernet port. But our gateway has a 10Gb/s SPF+ primary WAN port. Some Ethernet transceivers work in this port, while others do not. But even the ones that work get very, very hot very fast. I worry that they may one weekend blow a gasket, brining down the company's high-speed connection.
SFP have nothing to do with Layer 2 or Layer 3. It is up to NIC / CNA to support Jumbo Frames. Your question was asked in wrong way. Effectively you are asking if specific Optical Fibre Cable or UTP/STP/SFTP cable supports Jumbo Frames. This rather weird question. Especially taking in account that you can use your OM4 cable to transmit Fibre Channel only. Does it support Jumbo Frames? No, even though it is physically possible to transmit Jumbo Frames over OM4 cable. Does OM4 cable support NVME-over-Fibre-Channel? This as wrong question as yours
4:19 This is even weirder question and answer them previous one Your results are pointing to bottleneck at either CPU or RAM or NIC/CNA level Assuming 9000b jumbo frame you should get 1184MB/s, not 800MB/s Use faster glorified calculator for benchmark next time. SFP have no influence over performance (speed wise)
Hi do you guys know with NBASE-T, do you need to have both ends supporting NBASE-T for it to work? I'm kind of confused how these nbase-t sfp adapters work since the switches they are being plugged into expect a 10G connection and yet they somehow work only providing a 5g/2.5g connection. So does that mean the switch does not need to support nbase-t for it to work?
To anyone looking @ using these for 2.5G/5G. I'm not sure I'd buy any except the Mikrotik one. With the Ipolex and likely the other Marvell-based ones, you'll get a 2.5/5G link but the module only broadcasts 10G and real world RX performance is pretty bad.
@@ServeTheHomeVideo Are you sure? I've read some reports that some of these will report 10G Back to the switch but link at 2.5G. May make a great topic for a video. Adding a M.2 2.5G Intel NIC to the 1L PC's is easy
Has anyone been able to run 2.5/5 iperf tests with these modules to see if any can support full rate transmit (which seems to be the case) and receive - which to me is terrible (600mbps) at even 1G link rate. My tests were run on two Mikrotik switches crs326/crs309 with and without pause frames enabled. If a majority of your traffic is one way, I guess this is okay, but not for me.
DAC would yield similar performance, at a lower price. For short distance it's a no brainer unless you have a lot of cable running all together: fiber is a bit more robust, and immune tu rf interference. I would only consider these converters on short runs if sturdiness is a real concern. Otherwise, they are perfect for leveraging existing infrastructural cable
DAC's tend to work a bit better... no (or few) active components (very little power draw / heat generated), and there's no serialization and conversion delay. (even optical modules have a tiny encoding delay)
I was really surprised that it seems you didn't end up with a few that just ended up as re-badged from one manufacturer , are your really saying that ALL the 9x units were of unique design and manufacture and that none of them were actually idenical ?
There are a few rebadges as the data set has grown. For example, the Ubiquiti UF-RJ45-10G units seem to be very similar in components (yet different markings) to the older designs from a number of lower-cost 6COM, QSFPTEK, ADOP, FlyFiber, and Wiitek units. See www.servethehome.com/ubiquiti-uf-rj45-10g-sfp-to-10gbase-t-module-review/ - The general sense is that we are seeing two different generations of chips inside with the newer ones supporting 2.5/5G as well.
That was our guess as well when we started. We, however, saw that they all had different exposed PCB near the SFP+ connector. The chipsets might be the same, but the PCB is different.
@@ServeTheHomeVideo Interesting. There must be more manufacturers popping up recently. In the past, it has been my experience that most of the "name brand" optics were sourced the same as many of the "generic" ones. They differed only in programming, testing, and support.
@@QuadDerrick Sorry for my poor english. Yes. There is a manufacturer called 6com. It is a small company in Shenzhen, China. It has no relation with 3com.
No response from Cisco? If some manufacturer was using MY trade-name to advocate use of a particular product (presumably without permission), I'd be a bit perturbed!
I love these 10-15 min videos. They are easy to consume. I just had my breakfast with this video.
The modules have only PHYs to covert typically XFI to xBaseT. The newer PHYs tend to support NBaseT, where the older ones did not. The newer PHYs 'should' also be slightly lower power. For performance, it's almost SOLELY based on cable performance, how well it can get the signals down the wire and reject CM. As for other things like Jumbo frames, typically PHYs don't discriminate anything over L1 (some exceptions apply). Going back to the host interface, another thing to look at is if the modules solely support XFI using flow control to support the lower speeds on the line interface, or if they support USXGMII, which is also a 10G interface (the same interface actually @ 10.3125Gbs & 64b/66b), but support the lower rates by multiplying the data (like how SGMII runs 100Mbs, still runs at 1.25Gbs, but duplicates 10x to slow it down). I would have been very interested in knowing which host interfaces these have, and for the future the same applies to switches with SFP+ (XFI only, or USXGMII). The final part you're missing is cable performance, you should try short and long cable, and also see what the limit is. Thanks & cheers
This has rapidly turned into my favorite tech channel on youtube - no hype, all fact, but still entertaining and interesting. Keep up the Lord's Work, Patrick!
Worth noting that the reason the Mikrotik S+RJ10 use a little more power than the other adapters is because they are enhanced range adapters, if you have one at either end then the cable run can be up to 200m for 1GBe or 2.5GBe (instead of the regular 100m).
"Do they support jumbo frames?" - the only valid answer to that question is "No, they don't.". They also don't support VLANs or BGP or HTTP/2, because they are Layer 1 devices and don't care what the bits they transport actually mean.
That was my initial thought as well, however jumbo frames could potentially be an issue. These contain PHY chips that could potentially do some retiming and idle insertion/removal. Jumbo frames mean there are fewer opportunities to insert/delete idle control characters between frames as the frames are 6x longer than 1500 byte MTU frames, and as such jumbo frames could possibly cause some issues. As for the other stuff, that's just data in the frame, so it's irrelevant from the standpoint of the PHY chip. The bigger question is why is there any difference in speed at all, since these are physical layer devices.
@@AlexForencich the fact that there is a speed difference means that in fact they’re *not* just PHY devices.
@@JasperJanssen They are PHY layer devices, SFI/XFI on one end and 10GBASE-T MDI on the other end. The interface on both sides of the device is physical layer, which must operate within around 100 ppm of the specified line rate, otherwise the link will not work at all. Unless this is injecting pause frames or something towards the switch (which they should not be, but it should be possible to rule this out), the only possible cause for a speed discrepancy is packet loss, either due to signal integrity or differences in the FEC decoding. 10GBASE-T uses an LDPC FEC, and it's possible that some of the modules do not implement the LDPC decoding correctly/completely, resulting in more bit errors and hence more dropped frames on some modules. What I would like to see is a dump of the switch packet counters (tx, rx, fcs error) from both ends of the link during these tests, that may shed some light on the discrepancy.
i know Im asking the wrong place but does any of you know of a tool to get back into an instagram account?
I stupidly forgot the account password. I love any help you can offer me.
@Ledger Maximiliano I really appreciate your reply. I got to the site on google and Im in the hacking process atm.
Seems to take quite some time so I will reply here later when my account password hopefully is recovered.
Thanks for this video it helped me with a purchase of a Wiitek SFP+ to RJ45 adapter. It worked fantastically in my Netgear 10g switch going to a QNAP 5gb USB adapter which Netgear said couldn’t be done with my switch but since the negotiation seems to happen in the adapter everything works great.
Thank you for doing important work like this saving consumers real cash. Boring results are just as important and valid as explosive ones.
Thank you so much for your well done reviews! Thanks to you I bought a CSS3226 with a Mikrotik SFP+ to RJ45 10GBase-T Adapter a couple weeks back and I couldn't be happier! :)
Thanks so much for the work you put into these! Much appreciated insight into a sometimes overlooked niche of homelab or IT/home users.
Great review, the only addition I would consider here is choose lower wattage over everything else, especially if the longer term goal is to populate most of the ports. 20x 10GBase-T ports at .4 watt more sure does add up in heat and power draw. Also for the more serious consumers, get fibre where possible as the power draw is considerably less. Also fibre has much lower latency than 10GBase-T 6x10Gbase-T ports has a latency of ~15 opposed to 0.6 with fibre. On my router 10GBase-T port reports ~54c and the 10GSFP SMF fibre reports 32c. that's a whooping 20+c per module difference, thankfully I only have two Base-T modules the rest are fibre.
Great content. Always was looking for SFP Adapter Reviews. Most YT channels are only consumer orientated. Your content is from pro to pros. 👍
Great comparision! One thing you could have measured is latency over these adapters. If I've understood correctly, high quality SPF+ to 10GBase-T adapters should cause about 25 µs extra latency over fiber and it would be interesting to see if the cheapest models cause more.
Also a nice information i figured out the hard way. They are Hard capped at 30M wire length if the other device on the other end is a low power device like the Asus U2008 10g Switch.
If you use an INTEL NIC on the other end with a 100M power output capacity it works fa some longer ranges about 40-50M. But that is one thing to consider when using these module with other low power devices on the other end that you may not be able to wire through your entire home.
I would like to see some more imperial testing on your end if you could, i can just speak for one or two runs i made through my home which exceeded 30m.
We did testing in the reviews of longer cable lengths, however, these are all rated at 30m. Going beyond that there are a lot more variables.
My own tests show 45 meter is the maximum (with Cat5e cables). My longest Cat5e cable here in my house is 30 meters and works perfect with 10Gbase-T. Off course when you are a little less Scrooge, you can use Cat6a.
The generation of PHYs used within these modules, would draw too much power from the cage if they supported over ~30M. We should start to see that change in the coming future.
Since I'm looking to build a lab starting in a few weeks and might wire it for 10G, this video was very useful. Thank you for posting it.
Nicely done video; informative and no fluff.
The reason they note the ‘Cisco’ model number on their advertising is because allot of Cisco switches won’t accept ‘non-Cisco’ SFP modules. They just refuse to ‘enable’. After some protest by (home) users they developed a config option to make them work. They still generate warning and error logs in the device.
At least, that’s what was happening 5 years ago (when I did networking). We even had a supplier that flashed specific Cisco firmware on the SFP to avoid warranty discussions or deployment issues.
Just FYI - I got some of the Wiitek 10G-BaseT SFP+ modules today, they've got an apparent manufacturing/configuration date of 2021-03-30, and come with country of origin (Made in China) markings on the box and module. As for temperature, after a few minutes they're as warm as the 10G-BaseT SFP+ modules I got from FS - reporting about 50-54C in my Mikrotik CRS309-1G-8S+
We were recently troubleshooting an issue in a Cisco switch and we noticed one of our modules was throwing an overheat warning in the logs. We focused on that module and it turned out we had inadvertently installed a cheap module that came with some other hardware in the Cisco switch. There ended up being nothing wrong with the module, other than it did not support temperature reporting to the switch, so it was making a lot of log entries that were distracting. So that's another feature you need to be on the lookout for, is will it report temperature to a switch that wants temperature information.
Food for thought, if you plug a Cisco compatible "10GbaseT" SFP+ into a Cisco switch and run "show inventory" or "show interface status" it will come up as a SFP-10G-SR= module, a multimode LC fiber transceiver. Yeah it's spoofing the switch with another model and probably won't matter in the home but if run in the enterprise environment, it's not technically "supported" but I doubt Cisco TAC will even find out.
They will when you give them the "show tech" output. (if they look, and their scripts do.) At worst, they'll tell you to put "genuine" cisco optics in it. (which is of course bunk, as Cisco doesn't make optics... they have manufacturers add the required signature to the flash. And there are 64 possible signing keys -- the code's been known for a long time.)
Thanks for the great info.. I went with Hi-fiber. Google Fiber Router to a (Brocade)icx-6450-48poe switch. 10g no issues. Then I got a Zyxel multi-gig unmanaged to have 2.5g to two desktops. On the Google router it has 10g RJ45. So it went from RJ-45 -- > cat6a --> SFP+ on the brocade. with no problems. HI-Fiber SFP worked on both.. the Zyxel model XGS1010-12. Hope this helps someone.. Again thanks for the video.
Why no 10GTek? Used the 10GTek for the fiber install in my house last month. Just a couple runs between the NAS and a couple PCs (100ft each). Was able to to get 1.1GB transfer rates.. Pretty happy with them. (using 5 Port SFP+ Mikrotik switch). Amazon has amazing deals on fiber. Cost was very reasonable.
Great video as always! Thank you and your team for doing the comparison.
This video is awesome I haven't seen good concise comparisons of 10Gb/Cu SFP+ anywhere else. Surprised that the mikrotik is the least energy efficient.
The reason the Mikrotik S+RJ10 use a little more power than the other adapters is because they are enhanced range adapters, if you have one at either end then the cable run can be up to 200m for 1GBe or 2.5GBe (instead of the regular 100m).
@@llynellyn Ah. I was ready to not buy them. Now. I still won't buy them as I have a small apartment.
I know I'm late to the party on this, but those Cisco adapters did, at one point, exist, then they suddenly marked them EOL and seemingly scrubbed them from their site. I was looking to buy some at one point for the 12 port 3850s we were looking at for our office, but they were persona non grata in the middle of my research.
Cinematic and color graded videos about IT? Thanks youtube! SUBSCRIBED
I would have watched it too if he did one in a server room with a mobile phone and torch during a storm.
I would be curious to know what transceivers work fully on Cisco switches, namely in terms of if it can check temperatures and other various stats, properly
I love Mikrotik routers - but their switches, I've had so many of them cook on me in the data center. I've changed over to Juniper - but that can get a little pricey for the home.
It's funny how often network brands get pigeon holed for one segment. I love Routerboard as well but have been apprehensive to try the switches. I tend to go Mikrotik router if NGFW is not needed, Aruba or Unifi switch and Tplink EAP or UAP for wireless because nobody does them all to my liking at least not the budgets I prefer..
The fiber SFP+’s use much less power then cooper. In a former life we swapped all the SPFs to fiber and save thousand per month on electricity.
Thank you Patrick, great review, it would be great to know what is the distance these SFP+ adapters can do, thank you for your contributions!
Great video! how about a video on going from 10GB to 40GB, ur thoughts on 40Gb QSFP+ to 4x 10Gb SFP+ DAC Breakout Cables
We use this a ton. In the 100GbE generation, we do not even use 25GbE switches, only 100G to 4x 25G breakouts.
thanks for the demo and info, have a great day
You too!
At some point I am going to upgrade my switch to something with a QSFP+ port or two and they all seem to be either all RJ45 or all SFP+ for the rest of the ports. I'd rather go SFP+ for the rest to be uplinks to smaller switches elsewhere.
In my estimation, the difference in power dissipation was probably more to do with variances in component tolerances. Keep in mind that many of these "cheap" off-brand SFP+ units are likely built using the reject IC's from an assembly line that supplies IC's for a major brand name like Cisco/HP/Etc, so while they are still functional, some are over certain thresholds to go into the good bin. With regards to the fact that they appeared to be grouped up as 6 and 3 with similar functionality, I'd be willing to be that this test of "9" SFP+ modules was really a test of 2 modules with 9 different brands slapped on them.
As mentioned in an earlier comment, that was something that we thought at the outset. It turns out all nine modules had different PCB/ markings near the exposed part by the SFP+ connector. The chipsets may be similar, but at least the PCBs are different.
@@ServeTheHomeVideo Aye! Either way great work on these reviews! STH is one of the best resources out there for us homelab and small/medium business IT guys. Your testing of EPYC Rome lead me to 7402P's on 2113S-WTRT's for our hyper-converged server cluster at work. (We're running proxmox w/ceph). Works beautifully!
Love STH website and youtube channel. always great info
A transceiver works at the physical layer (Layer 1 OSI model) and frame size (hence jumbo frames) is a property of datalink layer (Layer 2 OSI model), right? If that is the case, I cannot imagine how any transceiver could be capable of not supporting jumbo frames.
The vendor compatibility advertising is related to the fact that some check the firmware or use specific options. You did not mention if you checked them with different switch vendors/generations? Which switches do need coded modules?
Sadly the Wiitek adapter is no longer available on Amazon.
Have you checked if the switches overheat with the modules that consume the most power? Can you make a compatibility list? Nice list @09:25 :)
i didnt know who i was expecting but when i saw him, "HEY ITS THE GUY FROM LINUS "PATRICK FROM SERVETHEHOME!"
Can you guys do a roundup of some QSFP+ PCIe cards at some point? i.e. is it really necessary to spend $300+ on one card for home network purposes?
Great idea. ConnectX-3 Pro cards are inexpensive these days for 40GbE. Usually, those are way overkill for home networks. 10GbE is a lot less expensive so we generally recommend that these days. If you think of anything you think we should cover, feel free to comment, we can see what Rohit can do.
What about errors? Input/CRC/etc? You mentioned Jumbo, I'm surprised you didn't mention those other 3.
Good vijay-eo.
Exactly what i was looking for!
Nice
Some of those modules become really hot - did you try to compare them by their temperatures?
Good review. Would have been nice if you could have done some distance testing. Specifically the ones that claim to be greater than 30m. I'm sure that would be a huge selling point if true.
What about long term reliability? In some cases they may be in a "plug it in and forget about it" type environment. I would be curious to see how they worked long term.
It is somewhat hard for us to do a longer term reliability piece on these. After 2-4 years we would expect a newer generation to arrive making the current generation obsolete by the time we get that done. We did hammer them for a week and did not see any issues, but that is not exactly "long-term."
@@ServeTheHomeVideo My colleague and I were discussing this video and he brought up a very good question. We are using a company for a dark fiber backbone for all of our schools. He has found that we've had issues when the same SFP was not being used on the same ends of the fiber. Did you do any compatibility testing to see if one brand of SFP would work with a different SFP on the other end? Or did you always use matching SFP's on both sides? My colleague found that non-matching SFP's would cause comm errors, because of subtle differences between manufacturers. Thanks.
@@leesully1669 We did not see that with these although we were generally staying below the 30m rated specs. That is generally a good practice.
@@ServeTheHomeVideo So all the brands worked well with not being matched? I'm not sure I was clear about the branding test. 30m spec is fine. But does brand A and brand B work together?
I suppose the one key element yet to be seen is their reliability and longevity... That's going to be the most difficult metric for mass imported goods from no-name manufactures. Perhaps there are a handful of legit ODMs overseas churning them out and they are just rebranded. But who knows...
These are fantastic for a home setup or lab, but I wouldn't use them for critical datacenter applications because of potential risk of failure and lack of resolvable warranty.
No 10Gtek? Seems like they always top the amazon list for me
It'a a common knowledge that those "cisco" or other claims on listings are ment to says for what vendor it can be encoded. It's actually a good thing, it makes it easier to find whatever you want. So there is no myth, it's common knowledge ;-] From all of the people you should know that.
The 3 slower ones are probably based on BCM chips, while the faster ones are AQC. The AQC chips are faster because they do less DSP and hence has shorter latency, while the BCM ones do more DSP and introduce more latency. The BCM chips do more DSP not for no reasons. The BCM chips are rated for 80 meters over CAT6A, while the AQC chips are rated for 30 meters. Test with longer cables and see if this is the case.
what about the power consumption?
can it be that lower latency leads to higher power consumption?
@@btudrus No. Processing costs something, in this case, silicon area and power consumption. First (AQC10x) gen AQC is hot because they used an older process node due to lack of funding, just like many startups. The second (AQC11x) generation should be better in this regard.
Nice review. Would have been nice to include Unifi's 10G solutution the UF-RJ45-10G.
Thank Gerhard. We mentioned in the video that we are going to keep the main site guide updated with new modules as we get them. We simply got the top 9 results on Amazon for this project and the UniFi UF-RJ45-10G was not there. Hopefully we can get that and a few more as well.
@@ServeTheHomeVideo Sounds great. Keep up the good work!
@@ServeTheHomeVideo Please also try the two FS.com ones.
would a teardown of the modules be of benefit? i haven't read the reviews, so maybe the asic's are already known?
Did STH do similar testing for optical SFP+?
We have not done that yet
Which revision of the Mikrotik module was used in the testing? Also, could you update to include the temperatures of the modules while running.
Did you notice/measure any heat variations between them? They seem to get incredibly hot, 130F-150F hot.
Yes that is a limitation of the SFP form factor. Get a RJ45 10Gb switch if you don't want that heat (and want longer distance 10Gb runs)
Thanks for the hustle!
This is great, feeding the algorithm.
dude, the wiitek one has the china export logo right there.
Oh man, you really weren't joking when you said the old studio setup echoed a lot
Yes!
@ServeTheHome though FS.com modules aren't available on Amazon, I would never buy a module from Amazon I do only buy my modules from FS.com. Those modules were testing work at best at 30m max. The FS.com ones, 80m (Lawrence Systems got his working at 90m).
Wait. No tearing them apart to see what's inside? (experience says they're all very likely from the same few factories in china.)
Thank You Sir ....
Did you try them on longer runs? I tried two of the brands on my 15m cat7 cable, none of them worked(on both 10g/1g). Even though they all advertise 35m on 10g and 100m on gigabit.
We did but only to 30M
As in 2024, I'm using ADOP's new 10G module which supports 1/2.5/5/10G speeds. They also claim that new module may connect up to 80M in distance if the cable condition is good (perhaps on CAT7 or CAT8).
The Transceiver market has been experiencing a long deserved correction for the past couple of years. The pinnacle of this development is what Flexoptix has been doing. Sure, in an enterprise setting, it's most of the time not feasible because you can easily lose support for the whole solution just because you're running third party transceivers. But in a small business setting, or in a homelab, they're a total gamechanger. I've installed at least 200+ transceivers, DACs and AOCs from Flexoptix for dozens of customers at this point, and they've always worked flawlessly. The programmability is key. But in most devices, they also work out of the box.
Which is very fortunate, because the FlexBox programmer that is used to program transceivers, DACs and AOCs cost an absolute fortune. If you've got it, though, you can install thousands of transceivers for each about 1/10th of the cost of the original product (for 10G SFP+ SR transceivers).
Just out of curiosity, will this work well in an Aruba switch? I've got some fiber-optic SFP+ modules for my Aruba 3500 switches but sadly, I have an issue with one of my runs of fiber and it looks like I'll need to change the links between two switches from fiber to copper.
Hey, I didn't catch what you said at 4:50 in "that support 2 1/2 Gig and 5Gig internet" Is that what you said?
Thank you sir
I don't need one but I watched the whole thing. I blame Patrick. And RUclips. And myself too.
Would have liked to see a price graph in there as well so i dont have to click every link to get the price.. or maybe that was your intention ;)
We thought about that. The problem we saw while doing this series was that there were modules we purchased at $35 then they would be $55 two weeks later, $48 when we did the review and so forth. They moved so much we stopped putting that information in reviews. That is why we recommend just searching whenever you need them.
gtek work great i have those they work super fast have you tested those i dont see them listed.
What about all of the issues I have seen online about vendor locking the firmware so they won't work in some manufactures equipment. Like putting an off brand in a Cisco switch...????
thx, but why have you not tested the link stability with long range links?
Hi Martin, we used 20m and a lower quality cable which is mentioned in each individual review's testing. We actually later tried 30m and they were all working for a week hammered with traffic. Our sense was basically that the 30m rating was safe. You are right that we did not test beyond spec there.
ServeTheHomeVideo thx for the test!
My question is which is more efficient/faster/reliable for converting between 10GBase-T and 10G DAC :
(a) A 10GBase-T cable plugged into an SPF+ transceiver then plugged into a slot on a multi-slot, multi-gig switch? _OR..._
(b) A 10GBase-T cable plugged directly into a *_dedicated_* 10GBase-T / SPF+ media converter?
It would seem that the dedicated media converter *_should_* win every time, as it has no MAC address lookups to make; it's simply what goes in one port has to come out the other. But I know of no actual testing data or empirical evidence to support this. My reason for asking is that our company just got AT&T 5Gb/s service that comes in over the infamous BGW320's Ethernet port. But our gateway has a 10Gb/s SPF+ primary WAN port. Some Ethernet transceivers work in this port, while others do not. But even the ones that work get very, very hot very fast. I worry that they may one weekend blow a gasket, brining down the company's high-speed connection.
Maybe with the cisco thing the just mean that it supports cisco e.g has the correct phy address + id that cisco accepts.
SFP have nothing to do with Layer 2 or Layer 3.
It is up to NIC / CNA to support Jumbo Frames.
Your question was asked in wrong way.
Effectively you are asking if specific Optical Fibre Cable or UTP/STP/SFTP cable supports Jumbo Frames. This rather weird question. Especially taking in account that you can use your OM4 cable to transmit Fibre Channel only. Does it support Jumbo Frames? No, even though it is physically possible to transmit Jumbo Frames over OM4 cable. Does OM4 cable support NVME-over-Fibre-Channel? This as wrong question as yours
4:19
This is even weirder question and answer them previous one
Your results are pointing to bottleneck at either CPU or RAM or NIC/CNA level
Assuming 9000b jumbo frame you should get 1184MB/s, not 800MB/s
Use faster glorified calculator for benchmark next time.
SFP have no influence over performance (speed wise)
What chips they are made with?
Hi do you guys know with NBASE-T, do you need to have both ends supporting NBASE-T for it to work? I'm kind of confused how these nbase-t sfp adapters work since the switches they are being plugged into expect a 10G connection and yet they somehow work only providing a 5g/2.5g connection. So does that mean the switch does not need to support nbase-t for it to work?
Yes. Both ends do.
@@ServeTheHomeVideo So how do you know if an SFP+ switch/NIC supports NBASE-T? It's really hard to find documentation on this.
To anyone looking @ using these for 2.5G/5G. I'm not sure I'd buy any except the Mikrotik one. With the Ipolex and likely the other Marvell-based ones, you'll get a 2.5/5G link but the module only broadcasts 10G and real world RX performance is pretty bad.
Does my switch need to support 2.5G, or will it just work? I have a desktop that has 2.5 and I'd love to be able to use that
Yes it does.
@@ServeTheHomeVideo Are you sure? I've read some reports that some of these will report 10G Back to the switch but link at 2.5G. May make a great topic for a video. Adding a M.2 2.5G Intel NIC to the 1L PC's is easy
@@ServeTheHomeVideo I just ordered one to try, I will let you know if it works in my Cisco switch which does not do 2.5/5
All power consumption of 10Gbase-t SFP+ are greater than 2W that MSA required. They could cause hotter.
They don't market cisco part replacements but they market cisco compatible firmware / deviceid,
The mikrotik one only does jumbo frames at 10gbit
Latency?
Has anyone been able to run 2.5/5 iperf tests with these modules to see if any can support full rate transmit (which seems to be the case) and receive - which to me is terrible (600mbps) at even 1G link rate. My tests were run on two Mikrotik switches crs326/crs309 with and without pause frames enabled. If a majority of your traffic is one way, I guess this is okay, but not for me.
Hi, where can I buy this modules for 30,40,50$?
Would the performance between using one of these modules or a DAC be similar or would a DAC perform slightly better?
DAC would yield similar performance, at a lower price.
For short distance it's a no brainer unless you have a lot of cable running all together: fiber is a bit more robust, and immune tu rf interference. I would only consider these converters on short runs if sturdiness is a real concern.
Otherwise, they are perfect for leveraging existing infrastructural cable
DAC's tend to work a bit better... no (or few) active components (very little power draw / heat generated), and there's no serialization and conversion delay. (even optical modules have a tiny encoding delay)
Can I install sfp transceiver on sfp+ NIC port?
The Wiitek actually has a China Export symbol on it.
Why are they only doing ~800MB/s?
If I buy microtic 10g sfp+ switch, I need 2 10g base t adapter for connect 10g sfp network card via rj45 cable Am I right?
How did you miss 10GTek?
We did that as well www.servethehome.com/10gtek-asf-10g-t-10gbase-tx-sfp-to-10gbase-t-module-review/
I was really surprised that it seems you didn't end up with a few that just ended up as re-badged from one manufacturer , are your really saying that ALL the 9x units were of unique design and manufacture and that none of them were actually idenical ?
There are a few rebadges as the data set has grown. For example, the Ubiquiti UF-RJ45-10G units seem to be very similar in components (yet different markings) to the older designs from a number of lower-cost 6COM, QSFPTEK, ADOP, FlyFiber, and Wiitek units. See www.servethehome.com/ubiquiti-uf-rj45-10g-sfp-to-10gbase-t-module-review/ - The general sense is that we are seeing two different generations of chips inside with the newer ones supporting 2.5/5G as well.
That Cisco product might be an unlisted product for big installations.
Can this adapter use in san switch that support sfp+
Some of these 10GBe go all the way to 1GBe; did you test for 1GBe?
It is likely that many of these modules are actually the same with just a different brand sticker on them.
That was our guess as well when we started. We, however, saw that they all had different exposed PCB near the SFP+ connector. The chipsets might be the same, but the PCB is different.
@@ServeTheHomeVideo Interesting. There must be more manufacturers popping up recently. In the past, it has been my experience that most of the "name brand" optics were sourced the same as many of the "generic" ones. They differed only in programming, testing, and support.
@@ServeTheHomeVideo Could you make pcb shots?
6com ? it exists a network card producer named 6com now ? it used to be 3com... i feel older and older every day..
No. 3com was taken by HP in 2010. 3com does not change it's name to 6com
@@catchnkill 10:18 , look at the lower card, it says 6com ,, dont it ?
@@QuadDerrick Sorry for my poor english. Yes. There is a manufacturer called 6com. It is a small company in Shenzhen, China. It has no relation with 3com.
No response from Cisco? If some manufacturer was using MY trade-name to advocate use of a particular product (presumably without permission), I'd be a bit perturbed!
That one eyebrow hair doesn't want to be there.