It would seem to fit Eben's willingness to step in where "the items being built didn't get it right".
There is a feature that the RPT keyboards have that others don't. The cable isn't permanently attached. It's just a USB-A to USB-microB cable. The built in hub is something you can find elsewhere, but in this case it fits well with the rather short cable on the official mouse.
I dare say there will be a buyer of anything RPF make.I'd buy one.
I'm familiar with that. Indeed, I've used that argument myself as a reason why some things are unlikely ever to be put forth as RPT products. However, it fails when you consider the CM, CM3, CM3+. What is the educational use of those products?
Much as I--personally--use USB booting and have never used netbooting, I wouldn't dream of disputing the priorities. If I get to anxious for USB boot to be released, I can always do what I did before it existed and use the simple workaround of putting /boot on an SD card and root on the mass storage device.Hence why Netbooting Pi4 rightly has a higher priority than USB boot.
And there is likely the answer to "Why the CM* boards." What I think matters is whether or not enough SATA adapters would sell. I suspect not enough to make a viable market (but there is no harm in asking) since WD never made a go of it. Or...it might just be that not enough people knew about them and/or their flaws were just too obvious and a "got it right" would take off. For that matter, even many Pi users appear to be unaware of the existence of the CMs to start with.What we users outside education want to see is very much secondary.
However, the licensees want to sell lots of Pi's and don't care who too
Those licensees sell lots to the industrial type markets.
Could not buy any from local suppliers in OZsince WD never made a go of it. Or...it might just be that not enough people knew about them
Yes, and a major proportion of the profits from those non-educational sales go to funding educational projects.
So we might see RISC-V SoC PCBs one day? er, kidding, a littleSo it is worth RPT (not RPF) producing and licensing hardware aimed at other markets.
One obvious idea I had was to build a data center of Pis. With 64 Pi4s you can build a nice compute cloud with 256GB of RAM & 256 cores. Probably for $2-4K. But a lot of the hardware on each board will be wasted. Which leads me to thinking that I'd want three kinds of boards:W. H. Heydt wrote: ↑Wed Jul 17, 2019 7:57 pmTherefore, I think that once the RPT launches a CM4, I would dearly love to see them build a *proper* CM4 SATA Adapter board. It should have at least 2 USB ports (USB 2.0 would be fine), Ethernet jack, and one HDMI output would be sufficient. It might be interesting to use the single PCIe lane from the SoC to be the mass storage interface, either as SATA or as single lane M.2. The common connectors--USB, HDMI--could be micro for all of me. It would be nice to retain one other feature from the WD board...an actual power button.
Dream on. The mods will be wielding the thread-lock key soon.Bakul Shah wrote: ↑Sat Jul 20, 2019 12:32 pmOne obvious idea I had was to build a data center of Pis. With 64 Pi4s you can build a nice compute cloud with 256GB of RAM & 256 cores. Probably for $2-4K. But a lot of the hardware on each board will be wasted. Which leads me to thinking that I'd want three kinds of boards:
The idea is to have a set of building blocks that can be put together in many different ways, from a small two board design with more IO than a pi4, to making a cluster of any number of Pis. But for the latter you also need a high speed connection fabric with scalable bandwidth and PCIe will probably do quite well. For scaling up you also need to worry about other issues like supplying enough power, cooling/airflow, etc.
- CPU board: 2711 + memory + PCIe bridge + some GPIO
- IO board: USB3, M.2, ...
- Backplane: N slots with a multi host PCIe bridge
Hahahahahahahaha. "Next to no R&D". You have no idea how much pain it took to get those right!!
Some of the folks in the CM forum were practically drooling over the prospect getting access to a PCIe lane. Unfortunately, some of them seemed to think that there were multiple PCIe lanes (if only...).
Luck has nothing to do with it. When I start a thread like this, I put it here deliberately.Lucky this is in the off topic posts.
There are PCIe expansion switch devices: 1 upstream, 2 or more downstream ports. This can be put on CM board as I suggested or on the backplane or an IO module board. CM4’s edge connector layout would not be compatible to older CMs so there is an opportunity to think through and come up with a design that can last for a few more Pi generations and allow for new uses. I’m sure Eben Upton & co are well aware of this.