Future of the CPU board?

[Askarus]

Let's be happy with the OMAP for a while then

 

[fahrstuhl]

The problem with the Rockchips isn´t only software. There is only a chinese hardware documentation available (if at all) which makes designing the CPU board nigh impossible.

On the official Rockchip wiki, they have some English language documentation: An 800 page technical reference manual and a zip containing reference schematics. Sadly, the hardware design guide for the RK3399 isn't linked to on the wiki, but it is findable via google. To me it looks as if they do want to be open towards the English speaking open source community. Maybe contacting them can help if more documents are needed?

 

[netlinker]

How about a Ryzen Embedded R1102G?
I'm not sure how to source them tough.
I've seen some embedded motherboard manufacturers on Embedded World this year that used them, namely eepd.

 

[asimov-solensan]

Let's be happy with the OMAP for a while then

Ramen! After following how difficult was developing actual CPU board I preferr to wait and get something noticeably better. And not just battery improvements.

 

[netcat]

Ramen!

 

[Askarus]

Maybe i.mx is a good way to go.
Here is an interview in German with an important NXP person.
They are talking about the future plans of the company and i.mx9 and i.mx10.
Both of them shall use the latest technology and be high quality devices.
The 9-series will be A55 and the 10 Series will be what comes after Cortex A77.
So at least the 9-seriex might be a nice upgrade for us.
But there is no release date, so we could either wait, or make an i.mx8 board to make friends with them already

 

[oskda]

That sounds nice! What GPU does it have?

The best part is Vivante GPU in i.MX8 has OPEN SOURCE drivers. Those i.MX8 SOCs are the most open source SOCs I know in ARM world.

If you look around you can find it is used by computers looking for most open source components. for example:

• Librem 5 (Wikipedia) [A smartphone is a type of pocket personal computer]
• MNT Reform

On other side i.MX8 SoC Production Lifetime is about 10-15 years, because it is very intended for industrial use. so you can assure it will be available for very long time.

 

[oskda]

Honestly? That sounds like a lot of headaches and effort for next to no gain. I just don't see many existing users switching and if the difference is barely there, it wouldn't really get us many additional either.
If OMAP5's EoL is imminent, I suppose it's the next best step, but otherwise? I know there aren't really many good alternatives, but perhaps it's not the time yet? Getting the OMAP5 to work took a while (and is partially still an ongoing effort, iirc), so imo the steps should be significant.

At the end I can only speak for myself, though. i.MX8 doesn't look like an upgrade board I'd buy in the future.

It may be you don't know about all the facts:

1. Less heat is a good point.
2. Less power hungry is a good point.
3. OPEN DRIVERS is a VERY VERY GOOD point, and i.MX8 are the most open drivers SOC I know in ARM world, for that reason it is used by open source minded computers like Librem 5 (smartphone) or MNT Reform (notebook), and others. With other SOC you have to load tons of blobs.
4. Cortex-A53 is Spectre free (no Out-of-Order execution), while Cortex-A15 suffers Spectre bugs.
5. Pyra needs a SoC with long production Lifetime and good documentation, exactly like i.MX8.
6. It can run 64-bit ARM code.
7. As having OPEN DRIVERS it can attract new customers, concerned about this point, like Librem 5 does. In fact it will have more points because Pyra has integrated keyboard, and I think this is a good point for a Linux in your pocket mobile computer.

I think in this (not complete) list are some important steps, some more important than raw performance increase.

 

[Letalis Sonus]

The best part is Vivante GPU in i.MX8 has OPEN SOURCE drivers.

Only the kernel part is actually provided by Vivante, though, the user-land part is a 3rd party driver developed entirely without support by Vivante, they only provide proprietary blobs.

Remember that shiny benchmark results do not represent the actual quality and range of supported features of the driver (*cough* nouveau *cough*)

 

[Kev2442]

I suspect most of these will stay in future i.MXes. With how fast the mobile market goes, it might not be long before their release and it would leave some time to get the most out of the OMAP5.
If it is financially viable, and there aren't many differences between the two, I guess making a few prototypes using the i.MX8M while waiting for the 9 series would show our interest and help familiarize with NXP chips and software.

 

[oskda]

The problem with the Rockchips isn´t only software. There is only a Chinese hardware documentation available (if at all) which makes designing the CPU board nigh impossible.

Go ahead with i.MX8 (M version I suppose, as the other are more power hungry), as it is really an upgrade in lots of points, and even it is more powerful in multithread as it has 4 Cortex-A53 cores vs 2 Cortex-A15 in OMPA5. I think there are a lot of good points:

1. Less heat.
2. Less battery wasted.
3. OPEN DRIVERS is a VERY VERY GOOD point, and i.MX8 are the most open drivers SOC I know in ARM world, for that reason it is used by open source minded computers like Librem 5 (smartphone) or MNT Reform (notebook), and others. With other SOC you have to load tons of blobs.
4. Cortex-A53 is Spectre free (no Out-of-Order execution), while Cortex-A15 suffers Spectre bugs.
5. Pyra needs a SoC with long production Lifetime and good documentation, exactly like i.MX8.
6. It can run 64-bit ARM code.
7. As having OPEN DRIVERS it can attract new customers, concerned about this point, like Librem 5 does. In fact it will have more points because Pyra has integrated keyboard, and I think this is a good point for a Linux in your pocket mobile computer.
8. Pyra soft and SO would welcome more developers attracted by open source hard drivers, and privacy.
9. Who really wants Chinese doc/support, short lifetime, NDA, incomplete doc, tons of blobs you can't know what they are doing and limiting kernel updates, because they are closed and not updated to support them.

With i.MX8 Pyra not only will be a gaming pocket computer, but also it would enter in selected club of no-closed-no-spyware pocket computers, like Librem 5. I think it can attract even more users. And it would do it with connectivity (as a smartphone) and physical keyboard and mouse controls, something Librem doesn't have: having those controls are excellent for a Linux in your pocket computer (without them you are restricted to less usable only touch soft or virtual keyboard software).

Finally, in x years, in future boards you could use better, more modern, i.MX SOCs, because previous experience with them would make transition easier.

I think those who criticise i.MX SOC are not aware of all points, and if they know them, they could change their point of view.

 

[oskda]

Only the kernel part is actually provided by Vivante, though, the user-land part is a 3rd party driver developed entirely without support by Vivante, they only provide proprietary blobs.

Remember that shiny benchmark results do not represent the actual quality and range of supported features of the driver (*cough* nouveau *cough*)

Librem 5, MNT Reform and others have chosen i.MX8 SOCs because they have the most open drivers and less hidden components (and other good points, of course).

MNT Reform, for example, can operate with ALL OPEN SOURCE, including firmware, except for only 2 small bolbs: RAM (not ARM code) and HDMI output (HDCP blob) if you need HDMI usage (I don't know if they could avoid this if they would use DisplaytPort output).

 

[elvissteinjr]

What is the current state of blobs on the OMAP5 Pyra anyways? SGX, sure. Apparently for Wifi, but firmware, not driver (that's from a 2014 news post though, so dunno if that's current at all). Pyra repo has the SGX blob in clear sight, but no idea if others aren't hiding somewhere as well.
The device can run fine without SGX drivers present, assuming it's just that or close to it, the Pyra is already pretty open.

That open source 3D driver which isn't officially supported needs to be in a usable state too. Is it? These GPUs are already pretty slow, so using a half-baked driver is far from ideal. Priorities, right. But that device with gaming controls should do its best to run games too.
Is being Spectre-free even important these days? With the majority of targets being patched anyways, how many will bother attacking this niche device?
64-bit ARM is neat, but what are the practical advantages for software running on handhelds? Sure, Dolphin runs on this compared to the OMAP5 (also newer graphics APIs), but that performance level doesn't run any games at full speed.


Having the system more open is a plus for sure, but how big of one is it? Upgrading to a slightly more powerful SoC with the full development cost of a new CPU board. It's already fairly dated and doesn't look like a long-term solution at all.
I'm more worried about the financial aspect of this move. Having worked with the i.MX 8 will probably make it easier to implement the next generation of it, sure. But that's still going to be additional development cost of another board in that scenario.

"X years" for the Pyra project can mean half a decade. I still believe in the project since I've held the device in my hands multiple times and trust the people behind the project. What I don't believe in is fast turn-around times. We've seen how hardware development is not smooth sailing countless of times. These steps should be big enough to warrant time and money spent in it.

Open-source is cool, but for me it also needs to be practical. I don't know how many people would buy a 2017 SoC device in, say 2022, solely for it being most open source if perhaps much better and cheaper options exist which come close enough to that ideal for them. I just don't think it's countless people. The Pyra will likely stay unique in its feature set of course, but many may not prioritize having all of its things in one device.

Just my opinion, though. Lots of it is speculation indeed and I don't get to make any final calls anyways. Still wanted to state it.

 

[levi]

The only thing I'll add is that I think there may be some value in practising making a CPU board with another CPU variety in it. And of course, to really learn what kind of CPUs can run in that thermal environment, you'd need some prototypes to be sure. And some of those GPU capabilities in some of the rightmost iMX8 chips on that pdf (from iMX8M skipping iMX8Mmini) look interesting, while the OMAP5 is limited to ogl-es2.0, but I guess we'll get that with almost any newer chip provided they actually have a GPU in them.

 

[Silent-Hunter]

Looking at the Ryzen Embedded R1000 series, is 12 to 25W too high of a TDP for Pyra?

Post automatically merged: May 25, 2020

Oops, I was looking at the wrong chip. The R1102G is 6W.

Post automatically merged: May 25, 2020

The R1305G is 8 to 10W TDP. Is that too much?

 

[daveshah]

.

Looking at the Ryzen Embedded R1000 series, is 12 to 25W too high of a TDP for Pyra?

Post automatically merged: May 25, 2020

Oops, I was looking at the wrong chip. The R1102G is 6W.

Post automatically merged: May 25, 2020

The R1305G is 8 to 10W TDP. Is that too much?

Almost certainly too high even with significant downclocking, not only for thermal reasons but also battery life. Looks like the FP5 package size is 25mm x 35mm so unlikely to fit the Pyra CPU board form factor either.

 

[Djoga'Ro]

I have high hopes for Risc-V wiping the floor with x86 and ARM in a few years in many (to-me-important) aspects. If beforehand the Pyra gets an upgrade that does away with lots of problems, what incentive would be left to switch again? I don't want my Pyra to be the only device, with which ARM keeps a foot in my door.
Of course, I still may be very wrong about the future of Risc-V.

 

[Phlyra]

I have high hopes for Risc-V wiping the floor with x86 and ARM in a few years in many (to-me-important) aspects. If beforehand the Pyra gets an upgrade that does away with lots of problems, what incentive would be left to switch again? I don't want my Pyra to be the only device, with which ARM keeps a foot in my door.
Of course, I still may be very wrong about the future of Risc-V.

What’s wrong with ARM? (Genuinely curious)

 

[netlinker]

I have high hopes for Risc-V wiping the floor with x86 and ARM in a few years in many (to-me-important) aspects. If beforehand the Pyra gets an upgrade that does away with lots of problems, what incentive would be left to switch again? I don't want my Pyra to be the only device, with which ARM keeps a foot in my door.
Of course, I still may be very wrong about the future of Risc-V.

That would me my preferred choice, too. AMD is nice driver and performance wise recently, but as already stated, it has a big package and 6W is already a lot.
The only thing necessary for a RISC-V is an open source graphics card, as the latest cores are already pretty powerful (maybe not as powerful as recent arm, but I would take that if I get a RISC-V based product). Any volunteers to write a hdl model? I would probably try to join in .

 

[daveshah]

I would love to see an open source graphics card, even if it is just on an FPGA, but the problem with any RISC-V SoC is the cost of a tapeout on the kind of process a Pyra SoC needs - many millions for a proper mask set, hundreds of thousands just for an MPW, not to mention software and engineering time.

Most of the current RISC-V work is either smaller (microcontroller level, or control processor in a larger ASIC) or bigger (massively multicore datacenter/AI type stuff) than a Pyra needs. The only project targeting this space is Libre RISC-V and whether it turns out to be vapourware or not remains to be seen.