Pyra's new cpu board soc.

[133794m3r]

I am wondering why the pyra would use anything other than what hardkernel/odroid is using for their SOC. I know that the replacement board talked about was talked about for it's linux support but all odroid socs have to have linux support out of the box. They also support the SOC for at least 4yrs and plan on being able to sell it for that long so there won't be an issue with sourcing the parts. Yes they use a chinese SOC maker but the biggest issues with using linux seems to be an a non-issue. They are still updating their board c1 board that came out in 2014 with the latest ubuntu including drivers and there are folks on the forums making sure that upstream debian works.

My main concerns with the propopsed soc for the new cpuboard is that for any sort of applications that aren't quad core the cpu perf is lower due to the a53 cores. Plus the GPU at it's stock speed is _slower_ than the current omap. Yes it'll be more power efficient but that's about it. I can't imagine that the rockchip(or whatever their new one is) part will consume much more power than the omap5 due to it using much newer lithography.

 

[Letalis Sonus]

They need to be available in small quantities and without weird license requirements as well - and the manufacturer still has to provide you with support. Most will only sell you tens of thousands of chips at once, and some may even only sell you chips when you're contracting certain certified companies for your board design (IIRC Nvidia was doing that).

the propopsed soc

"The" proposed SoC does not exist, yet, there are only ideas being thrown around. There is no work going on whatsoever on any new mezzanine board until the devices are actually being shipped to the public - and enough money was being made that there is actually money that can be thrown at it, there are depts to be paid off after all.

For all we know it might as well end up to be a new type of Zen based embedded SoC - AFAIK ED does have some contacts at AMD and the right people showed interest, they just have to scale their chips down another notch. River Hawk was announced to go as low as 4W TDP and might be of interest next year. As much as ARM is being preferred, the available choices are very limited and have a lot of drawbacks.

 

[levi]

These Socs are certainly worth keeping an eye on especially if they can be sourced easily.

You say the graphics should be more efficient, but is this actually true? It's not just about them using less power, what counts is them using less power doing the same thing.

I wonder if there's some relatively easy way of benching some of these SBCs in terms of graphics so that we know what they're capable of and at what wattage.

If it turns out these do have a genuinely more efficient GPU (and it's not impossible - the OMAP5 is built on 22nm lith after all, plus is fairly old) then these are definitely worth watching and if they can produce a SoC with equivalent or even better GPU and CPU performance than our old OMAP SoC in the future then they're definitely interesting.

Another concern is whether the SoC has all of the inputs and outputs we need. To a large extent we can multiplex that up a bit, but I feel supporting eSATA going forward might be unlikely.

FWIW I also have an issue with the title of this thread. Could someone change 'Pyra' (it's superfluous; we're in the pyra section already) for 'Proposed', or something?

 

[asimov-solensan]

My experience with an odroid board is that you are stuck with wathever kernels are the drivers are available for.

Ok, hardkernel promises support. This means that you may have a new kernel in two years.

That's what you get with chinese soc with "out of the box linux support". My U3 is still being used but thar was a noticeable flaw.

 

[levi]

I think that's what we would have got with the OMAP 5 SoC if Nikolaus et al. hadn't picked it up and tried it with newer kernels then patched them when they fell over.