I'm making my own high-end router. From scratch. 

I always thought it meant Chinese Export 🤣 nah, just kidding. 😅

@@maticjelovcan Fun fact: there's also Chinese Export, the logo is slightly different (basically different kerning).

Indeed... or use of the shelf extruded aluminium cases ;)

Yeap. There's billet aluminum case - what he built - and extruded, stamped, and cast. Stamping is quick and cheap (once the molds are made), but it's not the best looking output. (no sharp corners.) And then there's the issue of low volume; anyone who would/could do this would never invest the time in making the molds for just a few hundred, or even thousand units.

Yeah pressed and formed don't have the unibody feel of the CNC machined but I do get your points

If the project is supposed to be for home users then 1U rack is way too big. For commercial use, its fine, but for home use it has to be small enough to fit on a bookshelf, or on a desk next to you computer.

@@hubertnnn If it's smaller (as planned) than it will also fit inside the rack (with adaptor or whatever, but it will fit.

​@@hubertnnnthere already a ton of routers that much smaller than 1U in width, but make that up with long adapters

Please make this have rack ears. I didnt buy a 22u rack to use as a shelving unit.

Aluminum is a much better heat sink than the construction grade sheet metal.

It just looks really nice. That’s it

When he said he's going to fix the price to value ratio by upping the internals cost is when I hooked off. Please don't focus that much on the external casing, while plastic is mostly disliked it is used by almost everything for a reason.

a cnc case isn't really that much of an issue. there are ways to make a cnc aluminum case quite easilly without enormous costs. with scale prices go down and the cost of production also goes down. it fhe case was made out of like 1/8 aluminum plate it would be even cheaper but still maintain most of the benefits of cnc milling the entire thing. you can bend and weld the plate togeather for the outer ring and then weld the top on, and then cnc the finalized touches for the case to accept the bottom piece.

@@OctagonalSquare How often do you look at your router, though?

@@ashleyzinyk399 some people hang them on a wall in a bedroom or office, they gotta at least look ok.

I'm very reluctant to develop a new product in x86 architecture when the whole world seems to be moving away from it (Apple Silicon). Plus, the CPUs you're mentioning have no networking peripheral built in, so I imagine separate controllers would add significantly to the overall cost if I want to add two 10Gb ports. As for the height, yes, it'll actually be even narrower/slimmer but more on that in a separate video. And no, there will be no Realtek network chips. We'll use Marvell. There will most likely be a PCIe port on it in one for or another - those details are still being worked on. Thanks for the feedback!

Intel's 2.5G eth NICs have had issues lately, be sure to get fixed chipsets. Also I think for this use case Realtek is fine. Can't even remember having problems with Realtek cards *ever* (unlike, say, Aquantia (that got bought by Marvell), where I was unable to get their multigig cards working under Linux), and it's not like they're a new player in the market.

@@tomazzaman except for using Intel, otherwise I agree with the comment above. 2.5G is already mainstream and of course wifi 7 near enough

Agreed, Intel as an option if possible. And definitely an option to rack mount it would be awesome. Simple folded steel rack ears like Mikrotik use would work.

Heavily disagree on architecture. A good ARM has TONS of advantages over x86, not the least of which being power efficiency. Plus things like box64 can run most x86_64 software on ARM now IIRC for the things without ARM variants. I'd rather an over specced ARM chip that can double duty as a low-load homelab or media server than an x86 chip limited by power draw, heat dissipation, etc.

SUPER SUPER SECOND THIS! Pllease!!!

I thought he said there was already 2 or more RJ45 2.5 gig ports?

@@beakerwsw Please re-read the original post. If the FTTH is coming in on a 2.5Gb SFP connector (like mine does), and the SFP cages on this router only handle 1Gb and 10Gb (like many do), then I would need to plug the 2.5Gb SFP into a media converter or a switch to convert it into a 2.5Gb RJ45 connection to be able to plug it into this router. If the SFP cages support 2.5Gb, then I could plug it directly into the SFP cage without any additional hardware.

Thank you for the feedback - yes, it'll be rackmountable.

I don't understand why something like this would serve the purpose? ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-AKUTzjA1grE.htmlsi=XXNmu0fFajtDMEcW

what's wrong with the dozen or more existing options?

@@cpanic1153 too expensive, no support, no documentation, ugly, vendor lock-in, lack of features, pick any.

You're most welcome! There will be a lot of content like this, I want to share the whole journey!

Also design lifecycle because computer parts typically have a very short lifecycle from leading edge to obsolete. By the time the product is released your CPU or networking chip could be close to last orders / lifetime buy. If you are lucky the next generation will be pin compatible.

Actually, all the parts we're discussing both with engineers and distrubutors have *at least* 10 years left on the lifecycles. I won't have it any other way.

Thank you for your kind words! And welcome on board!

You are so welcome! Glad you find them useful. More to come!

Thank you! Yep, the CE understanding was wrong on my part, thanks for the clarification!

Yep, this topic comes up quite a lot, because I too would like to lower the cost while trying to retain the look of it. Supposedly there's a process, similar to welding, that can achieve a similar outcome, however I have no experience with it, and will definitely need to do more homework on the topic. Will report when and if I learn more. Thanks!

Yeah, that certainly sounds like feature creep. Instead of putting an existing device in a box, now he has to design the motherboard from scratch too. I think this is the kind of stuff you do when you expect to sell millions of units, but it's a big undertaking for a small new company.

Thank you! and welcome on board!

Thank you! And welcome on board! A lot has already happened and I'm working on an update video, likely to come out next week (or worst case, a week after).

Thank you! And welcome aboard!

Glad you enjoy it!

Thank you!

Thank you! My thougts exactly. I was pitching to an investor yesterday and he asked some similar questions (to challenge me, which I like!). As in, "you know you can get some chips much cheaper?" And I responded: "Yes, but those chips are worse, have poor thermal management and will be unstable in any kind of non-ideal environments. It's not a compromise I'm willing to do, because in a race to the bottom, nobody wins." He seemed happy with my response :D

Thank you! In fact, we did briefly look at the possible tech to achieve the same result, such as welding, but no results/solutions yet. Will dig further! :)

​​@@tomazzaman I was thinking this too. If you need something to screw into you could hypothetically machine a frame with the leftover aluminum. That itself may be a useful tool, being able to machine additional parts in parts of the aluminum you'd otherwise be getting rid of. This doesn't solve the underlying issue though. One thing you could consider is using separate pieces to your advantage aesthetically. Aka, machine each side of the router and possibly leave blocks for screwing into. Then for some sides you could use a different color or material for flat sections, e.g. after that rounded corner you could have long flat strips of a different color, or even a different material. This may be useful for things like an I/O shield for your motherboard and ports, which you can machine separately. Separating the sides results in a potentially much smaller overall volume, because you cut out most of the middle that you are discarding. Smaller bounding boxes = less material = cheaper, at the cost of having to machine addition parts. However, it is possible, likely even that you could machine the separate parts from the same block of aluminum, sticking them together as close as you can reasonably get (maybe an inch of separation). This would result in you being able to machine the four long skinny sides together. The big front and back could also be machined together, because they are both of the same size. If you can also make parts identical e.g. the top/bottom parts, left/right, and front/back, you can instead of machining them together simply machine 2 of them. And considering the necessity for some kind of area for I/O and ports, you could probably just leave out an entire side and machine an I/O shield with a simpler technique. If you design this with traditional I/O shields in mind for whatever motherboard form factor you use, this would allow other motherboards to be used. Batching different similarly sized parts together where the amount of empty space is minimized, and creating duplication for re-use of parts will mean you can just either use less material for multiple parts, or if truly identical, simply machining more of the part instead of having to machine multiple or machine them together would likely work great. You could even go a step further, and instead of machining the sides separately, also even machine the corners separately. You might even be able to create spots for something like tempered glass so you can show off the insides of your fancy new RGB gaming router fad starter. I can assume a tempered glass sheet is likely cheaper to produce than a part cut from a big block of aluminum, while also potentially benefitting the visual appeal. Lots of interesting ideas on how to try to reduce machining overhead, and create re-use where possible without sacrificing the aesthetic appeal of the case too much. Side note, but this is somewhat analogous to optimization in programming, in fact it is a reasonably accurate analogy even. Each part is a function, the desired shape is the code, and the 1 inch envelope is sort of like the overhead of running the code. Each part itself has an O(n^3) complexity. Like in programming breaking the pieces up too small is impractical and you add extra overhead unnecessarily. Also like in programming, re-use can lead to performance gains by being able to make optimizations (like cutting out the empty center of the router case as a cost by splitting it into separate pieces, or batching similarly sized parts together to reduce overhead). In programming there are often aesthetic preferences at play too, so a similar balancing game between cost and sleekness/cleanliness. Your budget or target cost is also like the hardware requirements you're trying to meet, and that governs how aggressively you will try to optimize. Pretty fun!

@@Hexcede @tomazzaman: Exactly, look the furnitures of Ikea, for example. How smart they are. Only the visible parts are painted or dedicated for visual and taktile appereance. The rest is only as less as possible, they should just functioning, not more. And this is the rules, what are working in the business. Be smart, be clever!

That's stupidly overpriced. Best solution is to rough cast and then jig machine it to lower processing cost for mount points etc

Welcome on board! 🙌

Thank you! And welcome on board!

Thank you! Welcome aboard!

Thank you! Welcome aboard!

Great to hear! Thank you!

Thank you! And indeed, we just had a meeting yesterday around this very topic, how to make it modular, so that we can sell a "base unit" then upsell modules (we were talking about thread, but applies to anything that can be plugged into an M.2 or mPCIe slot).

Of course! Sister owns an aluminum foundry, so that’s actually very easy for me! :)

Worldwide shipping is planned since the start, I see no reason why it wouln't be possible. 🙌

Thank you! There's plenty of content coming, both basic and more intermediate! Welcome on board!

YES! Finally someone who wants the same things as myself. Except for the case. But we'll talk about that. :D

If I focus on an x86 platform, I'm competing with PCs (to a degree), of which there are thousands (both models and manufacturers). On the other hand, nobody made a router that's ARM-based, open-source, well documented and looks good. 😅

Challenge accepted. 💪

Yep. In fact, I'm announing two versions in tomorrow's video: one in CNC-milled alu, and the other in sheet metal enclosure.

Thanks! Already visited a sheet metal workshop!

Hvala! Happy new year to you as well!

Yes, considering three options: fit it in a shelf, make rack ears or make a special mount. No decision yet, but will update here once we explore further.

Hvala!

Thank you very much!

Now where would be any fun in that? 🤣

@@tomazzaman You have a good point there :)

Check my latest video and be pleasantly suprised:)

In that case, make sure to watch tomorrow's video ;)

In that case, you'll be happy to learn that we plan two versions. CNC body for those of us who value that kind of look and sheet metal for those who'd rather save a buck. :)

Yep, exploring a bunch of options around the case.

Honestly, I'm not trying to beat R86S G2. Or any other device, for that matter. I want to make something great that people will be involved in the creation of, hopefully pass on as much knowledge as possible in the process, offer great support once it's done and sell enough to be able to build an ecosystem of other products around it. Ambitious, I know. 😅

Have you gotten to test it with any CPU-intense things like connection tracking or WireGuard? I'm curious what kind of throughput you get with NAT also.

@@john_elliott I am not even sure how I would go about running throughput tests. If I do a speedtest_net it shows up 913up/932 down. Fast_com shows 1gbps down / 910 up. Although fast used iip6 so not my NAT. But that's just my 1 gbps down / 950 up pppoe fiber internet. I only have 10gbps devices on my local VLAN and everything on the other VLANs runs at 1 gbps. I do know it is noticeably faster than the Deskpro with a 2x 10gbpc card it replaced. I mainly did it because the R86S uses less power so the UPS downstairs struggles less during an outage and also the smaller size helped as well since my landlord would not appreciate me bolting a second rack to the wall downstairs.

Agreed. Watch tomorrow's video ;)

There’s so many “votes” about this that I know absolutely must consider it :)

Yep, that seems to be the general consensus. Working on it.

Thank you! Will test the chosen CPU thoroughly before any final decision is made!

Why not sheet aluminium?

​@@DanSloteacomparing like for like sheet stock, steel is cheaper, but also steel is stronger so you could use a thinner sheet too. Light weight isn't really a concern for something like this that's fixed to a rack / desk.

@@carneeki I see your point. It's just that he likes aluminium too much. But milling the case from a solid block is pointless, unless the case is also a heat sink. I'd be happy with a 3d printed case any day.

@@DanSlotea machining the enclosure from solid is silly expensive. But if his heart is set on machining it, there's plenty of plastics that might do the job too. On the topic of plastics, you make a great point about 3d printing... Open source the files and have the user print an enclosure. It even design around a Hammond enclosure (or a similar off the shelf enclosure). Maybe even design the board to microatx or miniitx fork factors so an off the shelf media pc case could do the job :)

Thanks for the feedback - it's more valuable thank you think (regarding the final design). Will share more soon!

@@tomazzaman you're welcome, I'm glad I was able to help.

Thank you!

Thank you!