ValantarWhich is why teasing a router is absurd

And yet they do it. Why?

Honestly, I get it that you think teasing a router is absurd but I really couldn't manage to extract anything apart from that. Sorry, maybe I haz the dumb? :)

mechtechhmmmm

When are we going to get a consumer router with 4 lan ports at 5gig? It would be nice to have ssd's in a NAS and be able to transfer over 100MB/s on a cable..............

What year did 1gig come out in consumer space again?

Talk about stagnation...............

If graphics slot development was like that we'd all still be on AGP......................


probably why they changed the name to Ryzen ;)

1999 was when the 1Gbps Base-T standard was ratified. AGP was actually surprisingly short lived at a mere 7 years.

5Gbps is way more expensive compared to 2.5Gbps, as it's closer to 10Gbps when it comes to all the "bad" things involved in making a chip.
Intel's list price for a 2.5Gbps Ethernet controller is $2.40 and their chips have a TDP if less than 2W. Intel doesn't have a 5Gbps product, neither does RealTek, which leaves Marvell/Aquantia. They're not cheap, although I don't know what their third gen chips cost, their previous chips were in the $20 range. They also run hot and require passive cooling. Personally I believe 5Gbps will be a short lived niche, as there are no switches, whereas there are several of 2.5Gbps by now. The price difference between 5 and 10Gbps is also too small in most instances that it doesn't make sense investing in 5Gbps, as you need suitable switches/routers etc.

I have two Aquantia 10Gbps cards, one in my NAS and one in my PC. For large file transfers it's unbeatable.

zlobbyYou can always get a 5 QSFP (10Gig) port CRS305. Many new NAS already have 10Gig port, or a couple of 2.5Gig ones. Then you can use bonding on the 2.5Gig ports.

Bonding doesn't work between a single client and server and requires a managed switch. You need multiple clients to see any benefit from bonding on the "server" side and bonding on clients is pointless.

@Valantar I'm enthusiastic about routers :(
I understand why most people aren't though, as they barely understand what the box from their ISP does, nor how vital it is. If people did, they'd invest in better gear.

zlobbyI'm not sure if you got my point. For as long as manufacturers are selling, they will spit out products.

As for this ASUS, I doubt its BOM is more than $35. The rest is just profit margin. Build cheap, sell expensive. Check the fist paragraph, second sentence.

If you truly think that's the BOM cost of a modern router, then you're really clueless.
Yes, your could make a MT7620 based router for $15, but it would be 2.4GHz only and have 100Mbps Ethernet.

The AC1750 router I was involved in making some 7-8 years ago had a BOM cost of around $100, although that was a bit special, as it had a 3.5" touch screen, ZigBee and Z-Wave, as well as a discrete USB 3.0 controller that increased the BOM cost. It was also the first consumer router to a heatpipe for cooling and possibly the first router with a mezzanine board.

My point here though is that now, we have routers with much more expensive SoCs, I know QCA charges $30+ for their higher-end router SoCs, more RAM, more flash, a third 5/6GHz radio and so on. Heatpipe cooling is not rare any more and multiple PCBs are also common. This is impossible to make for $35. Just look up the cost of a good power amplifier or LNA and you'll see that alone accounts for a minimum of $10-15 of the BOM cost in a higher-end router.

So that's $40-45 without the Wi-Fi chips, PCB, ports, housing, antennas, switch IC, etc. and doesn't even take packaging, production or certification costs into account. So yeah, you're way way off the target here.

zlobbyAnd yet they do it. Why?

Because marketing people love to hype things well beyond what is reasonable, and generally tend to veer on the side of the absurd if given the funding.

zlobbyHonestly, I get it that you think teasing a router is absurd but I really couldn't manage to extract anything apart from that. Sorry, maybe I haz the dumb? :)

... that's the point. Not marketing it, as you responded as if I was talking about, but teasing it. Specifically. Which to be effective requires that this is a type of product that people get sufficiently enthusiastic about to follow up on, as a tease purposely gives you next to no information. In other words, teasing is a hype-building exercise waiting on more marketing. How many people are going to keep F5-ing their favorite tech site in anticipation of the release of this router? My guess: nobody, or close enough as makes no difference. But of course it's entirely possible I'm completely out of touch and routers are the new hot tech buzz that everyone upgrades every two months?

Marketing a router is obvious and simple. Essentially "Hey, look at our cool new router. It's fast and covers your entire home. It's really gamery/Pro/[insert buzzword]." That makes sense when you're selling a borderline indistinguishable commodity infrastructure product. Teasing? No.

TheLostSwede@Valantar I'm enthusiastic about routers :(
I understand why most people aren't though, as they barely understand what the box from their ISP does, nor how vital it is. If people did, they'd invest in better gear.

I'm sure there are some of you out there! I've considered DIYing routers to get something better, but the cost and time investiture is always way beyond what's reasonable. (Not to mention the size, unless you go really expensive). And the few brands that seem to be actually good are either wildly expensive or impossible to get a hold of. Which leaves me (and most people) with a "pick what looks best of all the crap" situation.

ValantarI'm sure there are some of you out there! I've considered DIYing routers to get something better, but the cost and time investiture is always way beyond what's reasonable. (Not to mention the size, unless you go really expensive). And the few brands that seem to be actually good are either wildly expensive or impossible to get a hold of. Which leaves me (and most people) with a "pick what looks best of all the crap" situation.

Well, I have a Netgear R7800, which was a high-end router in it's day that sold for about $250.
It's based on QCA hardware, has third party firmware support thanks to Voxel and the only time I ever touch it, is when there's a firmware update. That last part is what I'm enthusiastic about, as it doesn't give me any hassle, it simply works. This is also why a lot of people forget they even have a router, as the ones that "just works" are often ignored, which means they are also not updated, which leads them to be potential targets for hackers...
This is why I want to see better support overall from all the router makers and there should be regulation that forces them to provide updates for a five year period, as that's an acceptable lifespan of a router. Many router manufacturers (TP-Link, D-Link and even Asus for a lot of models) issue a couple of updates and a year after they launch it, they drop it like a hot potato and makes revision 1.1 and repeat.
It's a really shitty market to be honest and consumers are either using unsecure hardware or are forced to buy new hardware every couple of years in a best case scenario to have a product that's at least somewhat secure.
This is why I only buy routers that are supported by the likes of Voxel and Merlin.

TheLostSwedeBonding doesn't work between a single client and server and requires a managed switch. You need multiple clients to see any benefit from bonding on the "server" side and bonding on clients is pointless.

Utter BS! In MikroTik you just add the bonded ports to a single bridge and voilá. NAS' support bonding by default. Please stop spreading FUD.

zlobbyUtter BS! In MikroTik you just add the bonded ports to a single bridge and voilá. NAS' support bonding by default. Please stop spreading FUD.

Yeah, you can do that, no problem, but have you actually tested it?
I'm not spreading FUD, as I have actually tested this, several times to see if anything has changed.
First time was well over a decade ago. If you don't have multiple clients, bonding of Ethernet links will not give you more throughput. Sorry, but it's a fact, a single client using the same speed interfaces can not gain additional network performance from the bonded link using industry standards.
There might be some workaround in Linux, but that's not following industry standards for bonding of Ethernet.

TheLostSwedeYeah, you can do that, no problem, but have you actually tested it?
I'm not spreading FUD, as I have actually tested this, several times to see if anything has changed.
First time was well over a decade ago. If you don't have multiple clients, bonding of Ethernet links will not give you more throughput. Sorry, but it's a fact, a single client using the same speed interfaces can not gain additional network performance from the bonded link using industry standards.
There might be some workaround in Linux, but that's not following industry standards for bonding of Ethernet.

Now you really lost it. LACP, bonding, etc. are real standards. Emphasis on standards.

I don't know what experience you have but I claim it's highly limited. I'm using various aggregation standards both at home and in various professional VM and networking scenarios. Shocker, they all work as advertised.

zlobbyNow you really lost it. LACP, bonding, etc. are real standards. Emphasis on standards.

I don't know what experience you have but I claim it's highly limited. I'm using various aggregation standards both at home and in various professional VM and networking scenarios. Shocker, they all work as advertised.

Lost it huh?

Ok, so did you even read what I wrote?
Once again, bonding/LACP based on 802.3ad does not give you any performance improvements for a single client against a server using bonding.
This is per the standard. This is how Ethernet works.
Even if you perchance have bonding on the client, you will see ZERO performance improvements, as the Ethernet network flow can't be split over two bonded connections, following current standards. The client will only use one link in an instance like this.

However, if you have multiple clients, you will see more throughput on the server end, but you might need more than two clients, as the loads are rarely perfectly balanced a bonded connection, so instead of getting 2Gbps throughput, you might get 1.4 or 1.65Gbps or less/more, depending on the exact test, payload and how many clients you're using to test with.
This is partially because there's still a 50% two clients use the same network connection, since bonding is pretty dumb over Ethernet.

If you still don't believe me, please read this thread, as it shows multiple examples how it's not possible for a single client to go faster than its rated Ethernet speed for a single NIC.
serverfault.com/questions/626368/why-iperf-still-reports-1gbps-performance-when-using-bonding-over-two-1gbps-adap

As I said, I have tested this several times. The first time with a Thecus NAS over a decade ago, then with a Synology NAS, a managed switch and even bonding on the client as the system I had at the time had two identical Intel NICs and finally when I worked at QNAP where I had a discussion with some of the engineers there as to why it didn't work as one would think it should work.

And the next time you want to belittle people, maybe you should actually test what you have implemented first, to make sure it works the way you think it works.

Keep it nice, people.
No insults, just discussion of the topic, please.
Remember... discuss the topic and not the member.
Thank You!