If you could spec from scratch, the ideal, low-cost router for LEDE, what would those specs look like today?
I understand, that for this platform, a new LEDE build might be required.
Example (following the Archer C7 specs):
CPU: Qualcomm Atheros, e.g. QCA9563 + QCA9880 <-- AC1750
More than 32MB RAM
More than 16MB flash
Switch: Qualcomm Atheros
Would you substitute QCOM for another, even lower cost vendor (e.g. Mediatek)?
I understand Realtek is hit or miss and Broadcom is a NO.
Would you go with an integrated SoC, i.e. all-in-one?
Long-term idea is if we could put a group order together. What would that look like?
What are the best components based on the past LEDE dev experience?
Imagine building the perfect LEDE router. It's a quite intriguing thought exercise and maybe one we could take some action on, i.e. create a high volume (need >1k units to get started), white-box-based, reliable stream of LEDE-ideal hardware at an affordable cost. This would also help focus LEDE/OpenWRT development efforts on this open hardware platform and the hardware would get warranty and support.
For a router, a quad core x86 based CPU with AES-NI, 4Gigs of RAM and 32 Gig SSD, with 4 intel NICs, fanless.
Then separately 3 or 4 access points with standard PoE and 3x3 mu-mimo with high quality upstream drivers. I don't know what that'd be whether it's qualcomm or whatever, main thing is to be rock solid, dual band, and have adjustable power output.
@jwoods As you can see from my posts, I'm not trying to sell anything, rather I'm trying hard to find a low-cost, LEDE-supported hardware platform that I can buy or custom order at scale with h/w support. The answer to this, I think, can benefit the whole community.
@diakelan I'm concerned with x86-based CPUs since at the low end they require active cooling and make noise and at the high end, they get expensive and only some type of UI can justify their existence. Think a MIPS-based architecture is the balanced answer to this.
Not to rain in on your parade, but I have a feeling that MIPS is slowly being phased out for ARM or x86, so depending on long you want your platform to be available MIPS might not be the best bet. Note I am not talking about today or tomorrow, but in say 10 years I would be cautious about MIPS' availability...
It's already the case that MIPS architectures and even ARM aren't fast enough for connections people have today. I have ATT gigabit fiber, my sister in Portland has a fiber connection from Centurylink, gigabit is available but I think she chose 100mbit. Family members I know have access to 200mbit downstream connections over DOCSIS networks, and over the next 5 years or so, the number of people with high speed connections who need good quality bufferbloat free bandwidth shaping at symmetric rates upwards of 200Mbit/s is going to explode. x86 will handle this, MIPS available today, forget it, some of the high end ARM will handle it. But basically the days of consumer routers are I think nearly over. You will see already that among the enthusiast types who would want OpenWRT/LEDE the number who are moving to a dedicated x86 router + several wifi access points is ever increasing. The current high popularity of Ubiquiti and pfSense and the like is testament to this. Look at the popular tech press and you'll see lots of articles about how "moving to enterprise access points solved my home wifi problems" and the like. I think the writing is on the wall so to speak.
I run PfSense on a Qotom mini-PC and also run LEDE on PC Engines.
I understand the benefits associated with x86 and fully agree with you.
However, even in the US with high ARPUs, these prices are not easily justified, they're for hobbyists.
If you really need the impact, you need to get from the Roadster to the Model 3.
That's the spirit of the discussion.
In my mind, LEDE and OpenWRT, with their following, should be aligning around 2-3 configurations at any point in time of open hardware.
High-end can be an x86-based one, mid one can be an ARM-based and low one a MIPS-based.
Maybe keeping open tabs across two OEM vendors to maintain a low-cost balance.
My question here was mostly around the lower end one, the "Model 3".
What would that look like today if we could massively order one?
Are there crystallized views on that or do we need to jump from flower to flower?