Question Anyone moving from AM4 platform to Z490 (Intel)? For the 2.5Gbit LAN?

[VirtualLarry]

I know I'm tempted. Even the $200 ASRock Extreme-whatever budget board has 2.5Gbit LAN onboard. Most Z490 boards do.

Whereas, what about the X570 boards? Still very few.

I don't know about B550 boards, maybe the mobo makers will have a change of heart, and add 2.5Gbit LAN to most of those, too.

Yeah, I know, I know, you can buy PCI-E x1 8125 RealTek LAN cards for $20-30 to drop into a PCI-E x1 slot (if you have one left, after all of your GPUs and NVMe SSDs on AM4), or a USB3.0 dongle with a RealTek 2.5Gbit LAN from Cable Matters for $30 off of Amazon. (I have some of each, already.)

But there's both something powerful, and utilitarian, about 2.5Gbit LAN being built-in, to me.

 

[mxnerd]

Wait for Ryzen 4000 desktop APU and new motherboard?

AMD Ryzen 4000 'Renoir' APU Specifications Leak Out - Ryzen 7 4700G 65W Flagship, Pro Series, Low-Power '35W' SKUs With Zen 2 CPU and Enhanced Vega GPU Detailed

AMD's Ryzen 4000 'Renoir' desktop 65W & 35W APUs have leaked out along with their proper model names and final clock speeds.

wccftech.com

 

[VirtualLarry]

I was wondering about an 8C/16T Renoir APU, on the ASRock DeskMini 300(W) unit. That might be pretty sweet! (*with RAID-0 NVMe, even)

 

[piokos]

But there's both something powerful, and utilitarian, about 2.5Gbit LAN being built-in, to me.

Absolutely. But 2.5Gbps remains 2.5Gbps. So you either get a mobo with a faster NIC included or you buy an extension. It's just another $30 or so that one has to consider in a budget.

Since this year Intel adds 2.5GbE to both desktop and mobile platforms, new motherboards for AMD platform will probably follow.

Guess what, they heard your demand and (almost) all of them have 2.5Gbit LAN, but mostly from Realtek though. Even Biostar follow suit.

It's really not as much about demand as just being forced by Intel, since they're making this a standard on their platforms.
So next year we'll probably get 2.5GbE in most PCs.

But PCs are just one side of the problem. 2.5GbE is extremely rare in other products.
So most people won't be able to use this until routers, switches, NASes etc. follow.

Synology example: 2020 models already launched (DS120j, DS220j, DS620slim) are all 1Gbps. And DS620slim is a "premium" model.
But an announced, but not yet launched, DS1620xs will come with 2.5GbE.
AFAIK it hasn't been confirmed for other models launching this year (DS220+, DS720+, DS920+).

But the biggest issue will probably be with ISP-provided routers, which are very rarely updated and - obviously - you're usually forced to use them for 2 years.
So if you sign for a service this year, and get a 1Gbps router, you may be stuck on 1 Gbps even if everything else you have can go faster.

Wow, that's a hoot. RealTek 2.5GbE-T performs better than Intel, this time around.

I'm more concerned about the "packet loss" mentioned. I'm a firm believer that "packet loss" on a local wired LAN is a Bad Thing.

Well, Intel's NIC works bad with some routers and perfectly fine with others. So if you're able to choose all network elements, it should be fine.

2.5GbE is a little forced and - for lack of better wording - not as robust standard as 1 and 10 Gbps. And it's really young. Intel simply failed to properly test it.
2.5GbE integrated in Comet Lake looks a little rushed and experimental.
Fixed chips will arrive later this year with Tiger Lake and Rocket Lake, which were meant to be the first generation offering this.

 

[VirtualLarry]

But PCs are just one side of the problem. 2.5GbE is extremely rare in other products.
So most people won't be able to use this until routers, switches, NASes etc. follow.

Synology example: 2020 models already launched (DS120j, DS220j, DS620slim) are all 1Gbps. And DS620slim is a "premium" model.
But an announced, but not yet launched, DS1620xs will come with 2.5GbE.
AFAIK it hasn't been confirmed for other models launching this year (DS220+, DS720+, DS920+).

Asustor has been shipping 2.5GbE-T capable NAS units since last year. No wonder Synology is behind in market-share. They (Asustor) even released 2.5GbE-T USB3.0 dongles, for their NAS units that shipped with only 1GbE-T ports directly on them. (I should know, I upgraded mine.)

But the biggest issue will probably be with ISP-provided routers, which are very rarely updated and - obviously - you're usually forced to use them for 2 years.
So if you sign for a service this year, and get a 1Gbps router, you may be stuck on 1 Gbps even if everything else you have can go faster.

You must not be from the USA. There's no provider that I know of, that "forces" you to use their router for 2 years. (If you rent a router, you can exchange it for a new one at any time, AFAIK. And if you own your own, you can also change it at any time.)

Edit: But that doesn't affect your LAN speed, or your NAS speed, if you put your LAN and/or NAS on faster-than-1GbE-T LAN switches. Your WAN connection can remain at 1GbE-T or slower. So I'm not even sure why you're bringing this up.

Well, Intel's NIC works bad with some routers and perfectly fine with others. So if you're able to choose all network elements, it should be fine.

2.5GbE is a little forced and - for lack of better wording - not as robust standard as 1 and 10 Gbps. And it's really young. Intel simply failed to properly test it.
2.5GbE integrated in Comet Lake looks a little rushed and experimental.
Fixed chips will arrive later this year with Tiger Lake and Rocket Lake, which were meant to be the first generation offering this.

Gosh. I hope that they don't take this attitude with PCI-E support. "Well, this was the first release of PCI-E 4.0. It only causes corruption of SOME data. Recommended work-around is to force PCI-E 3.0 mode in BIOS..."

Sounds a bit of Intel-apologist to me. ("Well, it was a new standard, you can't expect them to get it right, out of the starting gate." Yeah, remember the Z68 recalls?)

 

[piokos]

Asustor has been shipping 2.5GbE-T capable NAS units since last year. No wonder Synology is behind in market-share.

Well, Asustor websites lists FOUR devices as 2.5GbE.
I'm also not entirely sure how their market share looks. I've always seen Asustor as a very niche manufacturer (enthusiasts?) in a niche market (NASes).
Are they really leading right now? In which market segments? Source?

Also, aren't Synology and Asustor somehow connected? Asustor just being a more enthusiast-oriented front company?

You must not be from the USA.

Well, 95% of population isn't, so there's a rather high probability.

There's no provider that I know of, that "forces" you to use their router for 2 years. (If you rent a router, you can exchange it for a new one at any time, AFAIK. And if you own your own, you can also change it at any time.)

I know very little about ISP specifics in USA (other than that services are rather on the expensive side).

Where I live we usually get a router with the service and it's usually something customized specifically for that provider.
It's rented and set up to work just with that service. And you can't used anything else.

I believe this is pretty common practice worldwide.
Furthermore, my ISP is owned by a US-based Liberty Global.

Edit: But that doesn't affect your LAN speed, or your NAS speed, if you put your LAN and/or NAS on faster-than-1GbE-T LAN switches. Your WAN connection can remain at 1GbE-T or slower. So I'm not even sure why you're bringing this up.

Because that adds complexity and what I started with is 2.5GbE becoming a mainstream standard. Buying switches is not exactly "mainstream". To be honest, even using ethernet isn't at that point. So I would expect this to in a direction of routers and NASes supporting 2.5/5 GbE and a fast enough WiFi for PCs.

But you're obviously right. I can run my LAN via a switch and get that 2.5Gbps.
Also, of course I would like my internet service to go past the 1Gbps barrier at some point. Providers don't feel the pressure to offer anything faster because consumer devices wouldn't handle it. 2.5GbE LAN is a good starting point.

Sounds a bit of Intel-apologist to me.

Absolutely not. It's a consumer-grade product that's supposed to be robust - but isn't. But it happened to Intel in favourable circumstances, when clients mostly can't utilize it so they won't complain.

 

[zir_blazer]

I insist. Ignore 2.5G. If you need higher speed networking, look around to get previous generation used 10G discrete PCIe NICs and Switches on the cheap. Consider that we're seeing Chipset integrated 1000BASE-T since at least the nForce 4 Chipset from 2004, if you waited 16 years to get a mere 2.5x speed increased and pay a ridiculous premium for it, you're doing something wrong. I hate how 2.5GBASE-T pretty much only purpose is to milk consumers, as the only useful thing it can do is allow you to reuse Cat5E cabling, at the cost of new NICs and Switches anyways. And if you intend on cabling again, check for SFP+ Ports and fiber SFP+ transceivers and cables. If I had the need and money, I think I would go full fiber.

 

[piokos]

I insist. Ignore 2.5G.

What's the point of this crusade against 2.5GbE?

If you need higher speed networking, look around to get previous generation used 10G discrete PCIe NICs and Switches on the cheap.

So how much for a used 10GbE NIC for a laptop? USB-C? Thunderbolt?

Seriously, 10GbE is very expensive and problematic. Premium NICs, cables and wall outlets (if you're not on Cat6 already), switches.

if you waited 16 years to get a mere 2.5x speed increased and pay a ridiculous premium for it, you're doing something wrong.

There's absolutely no explicit premium. It runs over Cat5e-compliant infrastructure, so something most already have for 1GbE.
By the end of this year it'll be supported in most new desktops and laptops sold.
So, as you gradually replace your equipment (like you usually do), you'll move to 2.5GbE automatically.
Hence, while you're speaking about "milking", most consumers' experience will be that LAN just started working 2.5x faster on it's own.

If you want 2.5GbE instantly - sure, there's an entry cost. But it's still much lower than for 10GbE.

And most of the time your home infrastructure will still work for 5GbE as it becomes the next mainstream standard.

I hate how 2.5GBASE-T pretty much only purpose is to milk consumers, as the only useful thing it can do is allow you to reuse Cat5E cabling, at the cost of new NICs and Switches anyways. And if you intend on cabling again, check for SFP+ Ports and fiber SFP+ transceivers and cables. If I had the need and money, I think I would go full fiber.

Well, actually 2.5GbE arrived exactly to save us money. 10GbE has been around for years and hardly anyone uses it (even in offices).
So we can stay on 1GbE for another decade or we can slowly move to 2.5GbE in 2-3 years and 5GbE later on (still Cat5e for short distances).

 

[VirtualLarry]

What's the point of this crusade against 2.5GbE?

I agree. I think that it's a needed and useful standard. 10GbE-T, on the desktop, sounds nice, but there are both some technical, as well as practical, and price issues with 10GbE-T, that there aren't with an incremental, practically-free (as included on this new round of mobos, as well as B550 mobos) new standard, that can use existing cabling.

In fact, I would argue that 2.5GbE-T is MORE important than 5GbE-T, which I see as more niche. Not the least because the market has been nearly flooded with relatively inexpensive PCI-E x1 and USB3.0 and/or Type-C adapters for 2.5GbE-T, powered by chipset made by RealTek. (Aquantia apparently now has 2.5GbE-T chips too, as obviously does Intel.)

All we really need to complete the picture, is routers/switches/gateways (even ISP devices), that support 2.5GbE-T, if not just for WAN, also for LAN connection ports. (Note that Comcast's XB7 Gateway is supposed to have 2.5GbE-T ports on it.)

 

[zir_blazer]

So, as you gradually replace your equipment (like you usually do), you'll move to 2.5GbE automatically.
Hence, while you're speaking about "milking", most consumers' experience will be that LAN just started working 2.5x faster on it's own.

If you want 2.5GbE instantly - sure, there's an entry cost. But it's still much lower than for 10GbE.

And most of the time your home infrastructure will still work for 5GbE as it becomes the next mainstream standard.

Well, actually 2.5GbE arrived exactly to save us money. 10GbE has been around for years and hardly anyone uses it (even in offices).
So we can stay on 1GbE for another decade or we can slowly move to 2.5GbE in 2-3 years and 5GbE later on (still Cat5e for short distances).

That is precisely why I'm against it. You're NOT going to move to 2.5G automatically because even if you build a new computer with 2.5G integrated, you will not change the Switch unless you are actually wanting to buy a new one and pay the price premium for a faster 2.5G one compared to all the 1G gear that is already in the market. And you still have to add NICs to the older computers. That is the CURRENT cost to implement 2.5G. Sure, it doesn't looks that bad. Yet, when the vendors decides to introduce 5G in two or three years, you will have to purchase new computers/Switches/NICs AGAIN, and maybe cables. And we're still not using 14 years old 10G, yet, which you will eventually have to buy AGAIN, with mandatory cabling.
Basically, you are going to be milked three times in a row with two middle point standards instead of attemping to make a longer jump to 10G and don't even touch the network again, as giving the track record of around 15-16 years with mainstream 1G you can bet that the extra effort to deploy 10G will last two decades. No wonder that network gear manufacturers decided to introduce for consumers a standard that was faster than 1G but not too fast...
The thing that most amuses me is when people are always waiting for a new Socket that supports new PCIe gen/DDRwhatever so that they can "future proof" their computers but ignore that networking is at least a realistic goal if you want to future proof something.


As I have stated before, since 2017, Zen introduced 4 built-in 10G MACs that supports 10GBASE-KR. Sadly, AMD decided to not expose it in any of their products but the EPYC Embedded line. Actually, that is precisely one of the multiple reason about why I'm angry about how AMD didn't exploited everything that Zen could do to wreak havoc in multiple markets at once. Had AMD wanted to, you could couple two links with a Dual Port PHY like Microchip VSC8254, that cost 52 U$D in 100 volume quantities (Note that I don't know how much quality or feature difference there is between PHYs because I haven't looked into them too much. There are cheapre Single Port or more expensive Quad Port PHYs, too) and have pretty much the most expensive components for a pair of SFP+ Ports, Motherboard-side.
Then you need a pair of 20-30 U$D fiber transceivers, fiber cabling, which is cheaper than CAT6A cabling, plus a Mikrotik Switch like the 149 U$D CRS305-1G-4S+IN or the 269 U$D CRS309-1G-8S+IN, and voila, you have the baseline for a rather high end network that you will never have to touch again until you become bald or your hair turns gray. Sounds like a good investment to me. Is simply that some vendors didn't wanted to offer such products to consumers. Realistically, they don't seem too expensive to me considering how many years they will last.

 

[VirtualLarry]

As I have stated before, since 2017, Zen introduced 4 built-in 10G MACs that supports 10GBASE-KR. Sadly, AMD decided to not expose it in any of their products but the EPYC Embedded line.

I at least agree with you there. AMD adding 10GbE MACs to Zen SoC, and then not pinning them out such that they could actually be utilized, on AM4, or at the very least, on TR4, seems like such a total waste to me.

Zen-based systems could have stood head-and-shoulders above all of the rest of the CPUs on the market, in terms of networking, had they pinned-out those 10GbE MACs, wired up some PHY chips on the boards, and gone to town.

Like you said, single or dual 10GbE on your average gaming PC's mobo, would have been AMAZING, and IMHO, sparked a much-needed revolution in networking. At least on the LAN side.

Now, we're finally (with Intel Z490 platform, and AMD B550 platform, and to a lesser partial extent X570) getting 2.5GbE-T ports on some boards. (And yes, on some really high-end boards, 10GbE-T, but those are just as pricey as buying an add-in card for 10GbE-T, if not more so.)

Edit: I'm using that Microtik 4-port 10GbE (SFP+) + 1-port 1GbE-T switch on my LAN, and it works well thus far. It runs really, really, hot with 4x copper SFP+ transceivers plugged in, so I've got a Rosewell external USB 120mm fan strategically-positioned on top of it, blowing air through it, and now temps are way more manageable. I recommend that setup highly. (Plugged into a nearby USB hub on my desk.)

 

[zir_blazer]

Zen-based systems could have stood head-and-shoulders above all of the rest of the CPUs on the market, in terms of networking, had they pinned-out those 10GbE MACs, wired up some PHY chips on the boards, and gone to town.

Like you said, single or dual 10GbE on your average gaming PC's mobo, would have been AMAZING, and IMHO, sparked a much-needed revolution in networking. At least on the LAN side.

Actually, there are Intel Processors with built-in 10G MACs, the Xeon D. Some recent Atoms also have them, too. Moreover, Intel (And I'm sure that AMD too, since they have at least one Embedded Ryzen model capped to 1G speeds) 10G MACs supports to operate both 10GBASE-KR and 1000BASE-KX modes. Thus, if 10G PHYs are still too expensive for your taste, that capability could have been used to go with 1G PHYs, which are more convenient since they are cheaper than full blown integrated PCIe NICs like you have on AM4, as AMD doesn't have a Chipset integrated MAC to rely on like Intel does to pair its own 1G and now 2.5G PHYs.
Also, I have seen some Dual Port and Quad Port 10G PHYs that are pin compatible between them like the Inphi CS4227 (Dual Port) and Inphi CS4223 (Quad Port), so you can switch chips and not populate two Ports to make Dual Port and Quad Port versions of the same basic Motherboard design. I don't know whenever something like that could have been done with 1G and 10G PHYs so that you could use the same Motherboard design with pin-compatible PHYs to cover both standard 1G and 10G models. This design consolidation would have made them even easier to implement in mainstream...
Note that the AMD MACs have a few cons, as apparently they are dull and basic in their feature sets. You have things like SR-IOV in the Intel X550 cards (And Xeons D also supported that on the built-in MACs). This means that AMD couldn't have tackled with those boring MACs the Server market, but they were completely satisfactory for consumer.


Instead of all that, what we've got? 2.5G. Which Intel couldn't even get right at platform launch. And 3 years too late. I'm not thrilled about it after seeing that 10G was THIS CLOSE to go mainstream. Instead, I'm angry, very angry.

Edit: I'm using that Microtik 4-port 10GbE (SFP+) + 1-port 1GbE-T switch on my LAN, and it works well thus far. It runs really, really, hot with 4x copper SFP+ transceivers plugged in, so I've got a Rosewell external USB 120mm fan strategically-positioned on top of it, blowing air through it, and now temps are way more manageable. I recommend that setup highly. (Plugged into a nearby USB hub on my desk.)

I'm surprised that you actually have 10G gear.

Is there any reason why you settled for copper? If I had to do everything from scratch (And Motherboard manufacturers provided some damned SFP+ Ports instead of Ethernet, which is perhaps the single reason why is expensive to get fiber on standard desktops. You always require a card), I would certainly go fiber. Note that fiber besides being cheaper also consumes like 1/3 the power than copper does. I think that copper only advantage is being able to use it for PoE (Power Over Ethernet), if you have proper gear.
Also note that not all 10GBASE-T transceivers works with NBASE-T like 2.5 and 5G. It may be possible that you have to spend another 50 U$D for a SFP+ copper transceiver. The good thing is that old SFP 1G 1000BASE-T transceivers are like 5 U$D each, so it would be cheap to interface modern SFP+ switches with copper while aiming to fiber.

 

[bigboxes]

My next project on my road map is an upgrade to 10GbE whenever I finish my home server project. I believe the server board will have onboard Intel 10GbE.

 

[VirtualLarry]

Just a small note, with that 10GbE SFP+ Microtik switch, I am using 10G copper transceivers that do step down to 2.5GbE-T. I didn't buy them for that, but it turns out, I lucked out.

Ipolex ASF-10G-T and SFP-10G-T-S Review 10Gbase-T and Nbase-T Adapter

In our Ipolex ASF-10G-T review we see a 10Gbase-T and Nbase-T adapter compatible with the Cisco SFP-10G-T-S and found a nice surprise

www.servethehome.com

There is one other downside. When we put two of these into a low-cost MikroTik CRS326-24G-2S+RM switch they would heat up to levels well beyond the newer revision MikroTik S+RJ10. We did not get any drops in two days of testing, but we swapped one of the Ipolex ASF-10G-T / SFP-10G-T-S modules for a newer revision S+RJ10 and the Ipolex was noticeably warmer. In a home setting and running 25-30m 10Gbase-T that would get us nervous in a passively cooled switch.

Yes, with four of those transceivers, my Microtik passively-cooled switch was getting REALLY HOT. Even without much network load. I added a Rosewill external USB fan ($15 USD), and set it to "Med", and put it over the end of the switch, blowing down into it. I plugged it into a USB hub on my desk next to it.

 

[piokos]

You're NOT going to move to 2.5G automatically because even if you build a new computer with 2.5G integrated

Soon there won't be no "even if". 2.5GbE will (likely) be available on every new PC. That's what I meant.

you will not change the Switch unless you are actually wanting to buy a new one and pay the price premium for a faster 2.5G one compared to all the 1G gear that is already in the market.

Assuming you use a switch at all, yes - you'll have to replace it. So for people who have one, that's an extra cost - still way smaller than going 10GbE.

And you still have to add NICs to the older computers. That is the CURRENT cost to implement 2.5G.

I precisely used the word: gradually.
Yes, if you want 2.5GbE tomorrow, on every PC - it's an extra cost.

But that's not what this is about. We're talking about 2.5GbE becoming a new mainstream standard. So that 2, 3 or 5 years from now (hard to say how long) every new mainstream device - for both home and office - comes with a 2.5GbE in place of 1GbE that it used to come.

I'm not sure if you remember how we moved from 100Mbps to 1Gbps in early 2000s. The standard was finalized in 1999, but even as late as 2006 1GbE was a "premium" feature that you didn't get on some cheaper PCs.
And yeah, 10GbE NICs appeared in 2003 (twisted pair 10GBASE-T in 2006). Hardly anyone cared.

The fact that today:
- we're relying more on integrated NICs
- new standard uses the same cables
already mean the transition should be much faster.

So in my household example: I have a few PCs and all of them should be replaced by 2022 (with Tiger Lake or later). So all my PCs will be 2.5GbE-compliant.
I'm more worried about the ISP-provided router, which means I may be forced to buy a 2.5GbE or 5GbE switch - they'll hopefully cost $30 by then.

Other than that - absolutely no extra cost.

Yet, when the vendors decides to introduce 5G in two or three years, you will have to purchase new computers/Switches/NICs AGAIN, and maybe cables.

I won't *have* to purchase anything. It'll be just as above. 5GbE-ready devices will launch and after another replacement cycle I'll be on 5GbE. And with the distances I have at home, CAT5e will still work.

If I wanted to get 2.5 or 5GbE as soon as possible - sure, there's an extra cost (still smaller than 10GbE). But I don't. I'm perfectly happy getting 2.5GbE in e.g. 2023 and 5GbE in 2028 or something - INSTEAD OF STAYING ON 1GbE.

And we're still not using 14 years old 10G, yet, which you will eventually have to buy AGAIN, with mandatory cabling.

10GBASE-T is expensive and there's a good chance it'll never be a mainstream solution. It was introduced pretty much together with 1GbE - as an enterprise-grade solution for connecting servers and workstations.
Especially today, as 5GbE becomes possible over standard CAT5e, going over all the necessary fuss for 10GbE looks even less worthwhile.

Also, we now move to PCIe 4.0 and USB 4.0, which will make 40GbE possible, but that's already Cat8 cabling.
So if someone is on 1GbE today, he can benefit from 2.5 and 5GbE - to move directly to 40GbE at some point.

And USB 4.0 / 40GbE is an important milestone because it'll lead to modular systems.
2.5, 5, 10 gigabit ethernet - functionally that's just faster transfers, so not everyone feels the urge.

And of course I really hope to see stable 1Gbps over WiFi 7 (802.11be) by 2024 at which point the only ethernet cable I'll need is the one going to the NAS.

As I have stated before, since 2017, Zen introduced 4 built-in 10G MACs that supports 10GBASE-KR. Sadly, AMD decided to not expose it in any of their products but the EPYC Embedded line.

Because it only makes sense in EPYC embedded. For the same reason Intel only does it in Xeon D.

Then you need a pair of 20-30 U$D fiber transceivers, fiber cabling, which is cheaper than CAT6A cabling, plus a Mikrotik Switch like the 149 U$D CRS305-1G-4S+IN or the 269 U$D CRS309-1G-8S+IN, and voila, you have the baseline for a rather high end network that you will never have to touch again until you become bald or your hair turns gray. Sounds like a good investment to me. Is simply that some vendors didn't wanted to offer such products to consumers.

No offense, but you're talking about it as an enthusiast. So you want the fastest ethernet... because it's the fastest ethernet. I'm talking about a common mainstream standard.

Also, the suggestion of running fibre at home... Was it supposed to be a joke?
Fibre can't be bent as much as twisted pair, so you can't put them in existing conduit.
So now I'm either running cables on the floor (lovely!) or doing a home renovation. So I really doubt it's cheaper than Cat 6.

Realistically, they don't seem too expensive to me considering how many years they will last.

That is a total guess.
We may end up with a different 10GbE standard since this one is already old and wasn't created with consumer equipment in mind.
As such, the first standard that I may be inclined to really invest into is 40GbE - and that's only if modular computers follow (i.e. a serious functional gain). If it's just for file transfers, I'll be perfectly fine on 5GbE. Probably for the rest of my life. That's 600MB/s. SATA SSD speed.

 

[mxnerd]

2.5Gbps switch for $30 by 2022? You are absoultely dreaming. You will be lucky to buy them for under $100.

USB 4.0 can operate at 40 Gbps, but the length probably is same as Thunderbolt 3

• 18 inches - 40GBps
• 2 meters - 20GBps

USB is for peripherals, never meant for networking. It's always a hack to use an ethernet to USB adapter to achieve the purpose.

 

[Rigg]

I think multi-gigabit is the answer for residential and small business networks. Now that we have affordable multi-gigabit Aquantia NICs and MikroTik switches there aren't really any major cost hurdles. If a home or small business has Cat5e structured cable 10G should be fine up to 45 meters.

Aquantia’s Gamer Edition AQtion AQN-107 10 GbE NIC Available

www.anandtech.com

CRS312-4C+8XG-RM | MikroTik

Switch of the future: the first MikroTik product with 10G RJ45 Ethernet ports and SFP+

mikrotik.com

FAQ Cat5e vs. Cat6 vs. Cat6a Cabling

Advice: Difference between cat5e, cat6, and cat6a cabling?

www.kit-communications.com

 

[piokos]

2.5Gbps switch for $30 by 2022? You are absoultely dreaming. You will be lucky to buy them for under $100.

Today - yes, because there are only enterprise products. It's not that hard to buy a 1GbE switch for $100 as well.

As we get demand for consumer 2.5GbE switches, they'll appear. You can get a decent unmanaged 5-port 1GbE switch for $15.

But most importantly, we need mainstream consumer routers to go 2.5GbE, because most people don't buy additional switches.

USB 4.0 can operate at 40 Gbps, but the length probably is same as Thunderbolt 3

• 18 inches - 40GBps
• 2 meters - 20GBps

USB is for peripherals, never meant for networking. It's always a hack to use an ethernet to USB adapter to achieve the purpose.

I obviously never said about any dirty workarounds. Normal communication over ethernet. For example: eGPU over ethernet. Or storage (but seen as local).

I wouldn't worry about distance.
What's the limit for HDMI cables? 15m? And that's probably for 1080p.
You can run 4K over HDBaseT for 100m on Cat 6a cabling.

 

[mxnerd]

A lot of solutions use ethernet cables, but not for networking.

HDBaseT is for extending AV signals for AV devices.

 

[piokos]

A lot of solutions use ethernet cables, but not for networking.

HDBaseT is for extending AV signals for AV devices.

Semantics.

Yes, HDBaseT isn't using Ethernet protocols. It's not Ethernet. If you plug a cable to HDBaseT on one side and Ethernet NIC on the other, it won't work.

But it is proper networking - i.e. communicating over a network using adressed packets. The model is different (OSI, not TCP/IP), the packets are different.
But it actually shares 10GbE physical layer, i.e. it encodes digital signal in the same way (PAM-16).
And, as a result, it also has roughly the same bitrate limit: ~10Gbit/s

 

[aigomorla]

2.5gb and 5gb are pointless, unless we start having 2.5 / 5 gbe switches that are priced affordably.
I can see why they would be popular as they probably do not require as much power going though the lan cable that 10gbe does.
But the switchs are the main point in any network, and so far why would you buy an expensive 10gbe switch to use it on a 2.5 or 5gbe?
Just get it done with and go full 10gbe.

 

[VirtualLarry]

Search for "Zyxel 12 port switch" on Amazon, and you'll find their newest consumer switches, with 8x 1GbE-T, 2x 2.5GbE-T, and 2x 10GbE SFP+ (which I don't know if they will support multi-gig as well, with an appropriate multi-gig supported copper transceiver).

Managed is $179.99, unmanaged is $149.99. For 2x 2.5GbE-T, 2x 10GbE, and a smattering of 1GbE-T ports. Not horrible.

Now that both Z490 mobo platform, and B550 mobo platform, are going to support 2.5GbE-T by default in most of their models, the time and opportunity is ripe for TrendNet or TP-Link or someone to come out with an 8-port 2.5GbE-T switch, preferably with one or two 10GbE SFP+ or -T uplinks, for hopefully not more than $200-250 to start with, and trend downward year after year after that.

I already own 2x of those big rack-style D-Link 8-port 2.5GbE-T with two (can be loud, starting up, or under load, or on a hot day) 40mm fans, but that's still not quite consumer gear.

 

[mxnerd]

QNAP just announces/releases 5port 2.5gbe switch QSW-1105-5T

QSW-1105-5T | Instantly upgrade your network to 2.5GbE connectivity

With five 2.5GbE ports, the QNAP QSW-1105-5T switch instantly upgrades your network to take advantage of 2.5GbE-enabled devices and NAS without requiring any new network cables. Featuring a near-silent fanless design and plug-and-play set up, the QSW-1105-5T provides network loop detection and...

www.qnap.com

No price yet but should be available next month.

 

[Rigg]

Search for "Zyxel 12 port switch" on Amazon, and you'll find their newest consumer switches, with 8x 1GbE-T, 2x 2.5GbE-T, and 2x 10GbE SFP+ (which I don't know if they will support multi-gig as well, with an appropriate multi-gig supported copper transceiver).

Managed is $179.99, unmanaged is $149.99. For 2x 2.5GbE-T, 2x 10GbE, and a smattering of 1GbE-T ports. Not horrible.

Now that both Z490 mobo platform, and B550 mobo platform, are going to support 2.5GbE-T by default in most of their models, the time and opportunity is ripe for TrendNet or TP-Link or someone to come out with an 8-port 2.5GbE-T switch, preferably with one or two 10GbE SFP+ or -T uplinks, for hopefully not more than $200-250 to start with, and trend downward year after year after that.

I already own 2x of those big rack-style D-Link 8-port 2.5GbE-T with two (can be loud, starting up, or under load, or on a hot day) 40mm fans, but that's still not quite consumer gear.

Why would you bother with a $200-250 8 port 2.5 when you can buy this for $500-600?

MikroTik CRS312 Cloud Router Switch 12-Port 10G RJ45 and 4-Port SFP+ L5

MikroTik 12-Port 10G RJ45 and 4-Port SFP+ MikroTik cloud router switch is a practical solution for those who work with huge data files and need fast, capable connection 24/7.

www.balticnetworks.com

 

[aigomorla]

I absolute love MikroTik.
Its not noobie friendly tho, but if you know how to work your way around a console its ubiquiti class at half the price.
My current 10g switch is MikroTik, but i am running it DAC cables and not fiber with transceivers.
Although i could run it with transceivers since you can get them for like 10 dollars on ebay, but i dont trust myself with fiber cables as i know i will bend them too far and have them go SNAP.

This is my 10g backbone.

Amazon.com: MikroTik 9-Port Desktop Switch, 1 Gigabit Ethernet Port, 8 SFP+ 10Gbps Ports (CRS309-1G-8S+IN): Computers & Accessories

Buy MikroTik 9-Port Desktop Switch, 1 Gigabit Ethernet Port, 8 SFP+ 10Gbps Ports (CRS309-1G-8S+IN): Switches - Amazon.com ✓ FREE DELIVERY possible on eligible purchases

www.amazon.com

Its quiet as hell as its is passive.
Show me another 10g switch that is PASSIVE.

QNAP just announces/releases 5port 2.5gbe switch QSW-1105-5T

I heard QNAP switches are terribly unreliable.
Look at the reviews for there 10g one. Its almost on PAR with ASUS.
I can pretty much say it that switch will be hard press to go against mikrotik 4 port SFP+ one in price.
The problem with SFP+ is you need expensive RJ-45 10G transceivers unless your running a SFP+ card.

 

[piokos]

2.5gb and 5gb are pointless, unless we start having 2.5 / 5 gbe switches that are priced affordably.

Not every user needs a switch (other than the one in the router).
I bought my first switch 20 years after I got an Internet connection. Only because I wanted an RPi in the bedroom and its WiFi is not stable enough.

I can see why they would be popular as they probably do not require as much power going though the lan cable that 10gbe does.
But the switchs are the main point in any network, and so far why would you buy an expensive 10gbe switch to use it on a 2.5 or 5gbe?
Just get it done with and go full 10gbe.

10GbE remains a lot more expensive. Even more so when you already got NICs for 2.5GbE "bundled" with the PCs.

QNAP just announces/releases 5port 2.5gbe switch QSW-1105-5T

As I said earlier: 2.5GbE coming as standard in mainstream platforms will convince switch makers to launch consumer-grade 2.5GbE models.
That said, this happened a few weeks earlier than I expected.

No price yet, but looking at the features... probably $40-50.

Why would you bother with a $200-250 8 port 2.5 when you can buy this for $500-600?

MikroTik CRS312 Cloud Router Switch 12-Port 10G RJ45 and 4-Port SFP+ L5

MikroTik 12-Port 10G RJ45 and 4-Port SFP+ MikroTik cloud router switch is a practical solution for those who work with huge data files and need fast, capable connection 24/7.

www.balticnetworks.com

I'm not sure if this is meant to be a joke or what.
We started from discussing consumer platform. Now people are linking enterprise-grade massive switches that cost as much as an average PC.

I heard QNAP switches are terribly unreliable.

It's the only consumer-grade switch they offer right now. Everything else is more about small business segment, so punching in a different category.
Competing with cheap consumer stuff, this may be adequate. We'll see soon enough.

That said, given the choice, I'd probably go for something made by company with a good history in consumer products, e.g. TP-Link, D-Link or Netgear.