New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Anyone designing their own high-speed circuitry out here?
Does anyone design high-speed circuitry for their own servers? As well as the cases and racks themselves. I am an electrical engineer and eventually will want to develop custom server motherboards. It doesn't have to be anything complex, maybe someone has designed 8-layer PCBs for SATA3 HDD connection or something of that sort. So far not too easy to find any books or materials that specifically target these types of things, if you have some advice on where to look, that would be incredible.
Comments
Yes, but you need deep ties with China and Taiwan.
You won't be. Building custom server motherboards in small quantity costs so much more than you're imagining. Take a look at kickstarter projects for the teams that build custom hardware for an idea of all the things you're not aware of right now.
Which program do you use for your design work? Altium? KiCAD? What year of school are you in?
Nope, I say you are wrong. Watch me do it.
You come here asking for advice and when someone gives you advice (and asks some followup questions) you immediately shut him down. You're gonna go far kid!
Developing a server motherboard? Testing stability, power consumption, etc. will take up a lot of your time, or are you just developing based on the public version?
Ktalap Computers Corp, Model 1:
Y'all are shitting on me only because you can't imagine yourself doing something like this in your life. And you are right, YOU can't.
On the other hand, I CAN. Go grind motherfuckers.
How long would it take to design a sata controller?
Emm... high-speed circuitry and cases and racks don't really go together.
One is electronic-related, there other is basically, mechanical-related. You don't really need much electrical knowledge to design racks, but you need to understand ergonomics and mechanical considerations for rack design. You should know that, as an electrical engineer, which by the way, is very different from an electronics engineer. The fact that you talk about designing boards for connection doesn't show that know much about electronics design.
Multi-layer circuit boards. Probably doable these days if you outsource to PCBway or JLCPCB.
If you're looking at custom build motherboards, I guess the question is why are you even thinking of doing this. There are many motherboard manufacturers out there. Most of the design starts off based on the reference chipset design from the chipset manufacturer, e.g. Intel.
If you're really keen on electronics design, prtobably start off with something what davide suggested, a 555 blinker circuit
Thank you for this load of crap you just commented. Fuck you, and I will use this as my fuel for my victory, I will listen to what you have wrote here on repeat, forcing rage outside of myself, to have more power for my victory that is destiny and not a question.
You will keep hating on me and I will come back better than ever. Go grind motherfucker.
What the hell just happened here? Ktalap, dude, relax. You have succumbed to trolls.
The main question here: do you understand that custom circuitry for such tasks as servers is insanely difficult to produce? Here people do not know anything beyond low end vps or mere javascript.
So be forgiving, a lot of people visit grinder regularly, so no need to remind them to grind more.
Peace man!
I never designed it on my own, but I ever worked on FPGA project to create triple digit Gbps NIC with WAF integration in it. afaik there are no easily available books beyond the basic knowledge, because they contain intellectual property by the company who made it. If you're interested I'd suggest you to actually work with those kind of manufacturing company, otherwise create a better search term to find your reference.
Thanks. I was farming some of them messages of hate. I ran out of things that induce anger into me. I save them and go over them, again and again when I think about not doing anything. So yeah, the trolls did a free work for me here, I would say around 2 hours of thinking, that's like 15$. Got a medium warning for that, but it's well worth it.
I don't. Simple reason: Even a 1 GHz scope doesn't cut it nowadays with any serious processor and/or PCIe or (not super old SATA etc.). And good luck finding a logic analyzer with 96+ channels and > 1 GHz for less than the price of a nice house. Translation: the TME cost alone is in Mercedes Benz (and beyond) territory.
And why would I? I mean it's not that a say, 500 - 800 MHz power CPU or even an old 68K or NS 32x32 or ... is chicken shit and boring. Quite the contrary, one can get astonishing performance out of those older processors and quite a few of them are not less but actually more interesting than X86. And anyway hardware people tend to put demanding stuff in FPGAs nowadays (often and increasingly in the X86 world).
Besides, > 1 GHz and more than 4 layer boards are outside of most SME (and certainly hobby) electronics reach anyway. Also: why? After all one can get even quite powerful processors (plus some periphery) boards relatively cheap. So, why heavily invest in "serious" lab TME unless one intends to develop for mid to large scale series production?
If I need e.g. a board for SATA 3 (your example) I simply look around in the market. Hell, they'll even throw cheap boards with not only SATA 3 but a > 1 GHz processor too at me.
Anyway, I don't think that LET is the right place for your topic. Maybe try eevblog.
FWIW I use - and quite like - diptrace.
Thank you man. In case you have a business and let's say make 50k$ a month, it may make a lot of sense. Especially if you want to develop something really, really dense and powerful. I don't know, 1U server with 4 EPIC 64 core CPUs and 50 NVME storage disks and a couple of 5000W PSUs to that and maybe a single GPU on top of that, it sounds absolutely ridiculous, but I am pretty sure it's possible to do that.
I wouldn't call it ridiculous but I'd ask whether it's sensible. You see, even EPYC processors have limits, incl. the number of PCIe lanes. 50 NVMes need 200 lanes (but not "a couple of 5000W PSUs") plus there still is some other stuff that needs lanes as well.
In other words: There are solid reasons that the usual suspects produce what they produce.
Besides you might be better serving your interests designing something that (a) the large common players do not produce/offer/sell, and (b) that's actually and reasonably within reach (at least in terms of budget).
Their strength, after all, is "a million or two TME costs? No problem, when it leads to some million devices sold", while your highly likely is "a 'weird' niche thing? Great for me!".
Nah, I am pretty sure it will make sense. It's quite sensible especially if you want to fit some 60 HDD disks into a single rack, and if you have 4 epyc cpu's, you will have 128*4 PCIe lanes. A lot of the components you may buy, just like you said. But part of it you will make yourself.
It's just like with tuning cars, you want to make something exceptional, unbelievable and incredibly powerful, you will have to do some custom work. Even though you will start with some basic engine of your choice.
I would like you to know, and not guess, that my stance is "to change the world and defy limits", that should be clear. Whenever people try to put limits on things is the exact reason why so many are mediocre, I refuse and will not bend to anyone's opinion, even if they are a world expert in whatever. Spacex's rockets were said to be impossible, but look at them now. Thank you again.
We know more than low end VPS.
We know HTML.
We wrote an entire thesis in HTML in college.
We also know how to make popcorn 🍿
That's just not true. Plenty of talent here.
Cope harder kid.
Everybody thinks that it would cost so much more in small quantities but we do it. Not a lot of motherboards but we design and build our own KVMs and environmental sensors. (plus a few other things)
You can get the boards stamped in China for a few cents each in quantities of 20-30. Our remote sensors, temp, humidity, pressure. Cost us about $3 each to build. Our KVMoIP modules we've developed cost us about $13... and that's with and expensive buck converter. We build them 10-20 at a time.
No, he is clearly the troll. In the first post, you could assume he was first year student who was just extremely naive. But the other responses clearly show there's no valid intent by the question and its purely trolling or mental illness.
He has never designed anything let alone produced anything. I doubt he's even a student.
Yes, your basic sensors can be made as JLCPCB for dollars, it's designing that that takes the time and effort. But surely you know JLCPCB isn't equipped to make server motherboards due to part availability and complexity.
Unfortunately AnandTech went down, someone over there could give you solid pointers/startups, ( if you can disclose/share what your exact expertise, objectives, etc, etc ).
Quite a few opensource options are available in various categories.
But, opensource circuitry are not high-speed or not-able to handle well.
Another type is mixed type, (which you indicated), mixing some proprietaries & opensource & your own custom components.
there are compact form boards/models available from existing established mfr (manufacturers), asm (assemblers).
those can be used in custom big size boxes or in 1U or 2U boxes, and also have additional components packed in.
air-flow-cooling or liquid-cooling or electronic-cooling etc need to be thoroughly tested+developed when compacting.
cheaper sata-3 also available, (but may be not opensource).
so basically, work+improve on cooling .. on those compact boards ... as its related with density, flow-rate, etc, etc.
But if thats not-satisfactory,
and you still want to really develop+work+research on your own interface or router etc devices, etc ... get into touch with Chinese, etc mfr/asm ...
one degree in something is not-enough, depends on what university, professor, etc.
sometime even two degrees are not-enough.
not-enough for extreme level or extreme high-speed stuff.
multiple people from multiple different category of technology, etc need to work/collaborate on such (extreme level or high-speed related) project.
but low-to-moderate speed or standard speed (which is changing every year) or slightly-higher-speed than standard ... device ... is possible by single person or 2, with longer time.
so you may begin with opensource or chinese devices ... with target to gear/shift toward/into high-speed devices/solutions.
not only talent, but money $ is also another big factor, that changes various project R&D activities.
I don't know who JLCPCB is so I have no idea what their capabilities are.
Finally, some "get it done" mindset. I think developing the rest of the parts is not complex as long as you keep motherboards as they are, and you just buy them. And other components like remote control boards are not worth designing, and they are boring too, just buy one and be done with it. It's mostly the PCIe gen5 and SATA3 connections you are designing for, extensions for disks, custom cooling and custom cases.
When it comes to motherboards it might really be expensive, but I am pretty sure there is some superior way no one knows about, that's not only much cheaper but also faster in production. You can do anything, we live in non-deterministic world.
Some people out here have nothing in their lives and go around pumping shit into the world, it's sad, sad to see how much they are missing out on. But hey, without that my achievements wouldn't be sweeter. Say that I will fail, say I won't be able to do it, bet against me.
Stay there, don't do nothing, better for me. When someone needs help, I will surely offer them a hand, today is not the day, and you are not the person.
use your real account @Otus9051
I think no one likes me the second time around either. LET sucks, will go hang around with EE nerds. And by the way, I WILL build a motherboard, eventually.