The Future of Pyra's CPU


I totally agree with you @levi, I myself am enjoying older hardware much more, trying to put together three retro PCs (one circa 86-96 with a Pentium 200 MMX, one 96-02 with Voodoo3, and one for beyond 2002 running XP) while barely touching any current AAA games.
As I don't watch TV, I'm waiting for the day nice big 4K HDR monitors become affordable. I'd like to keep my screens mostly OS-free, thank you.
 
Thanks for the info. I figured as much as the $250 Ultra96 board needs a 8V 3A PSU.
Automotive-grade chips that target a certain ASIL level tend to use ancient manufacturing nodes in order to achieve the strict reliability requirements - 28nm is currently still state of the art. UltraScale+ is certified for ASIL-C, with such beefy CPU cores a high TDP is practically unavoidable. ARM64 is especially affected, because there is no official lock-step design, yet.
 
Was fun to watch him try to get games optimized for a voodoo 3 run on a matrox.
Which one ? It has "LGR" written all over it but I don't think I've missed any of his videos since I've been folowing him.
If there is enough interest, maybe we should open a retrocomputing topic on the boards, or is there one already ? I don't really want to derail this thread any further :)
 
  • Like
Reactions: rSl
Wait, I've seen that one. But it's not Voodoo3 games running on a Matrox, it's Voodoo Graphics with a Matrox as the 2D card. I guess I read your comment wrong, I shouldn't lurk late at night :mad:
 
Automotive-grade chips that target a certain ASIL level tend to use ancient manufacturing nodes in order to achieve the strict reliability requirements - 28nm is currently still state of the art. UltraScale+ is certified for ASIL-C, with such beefy CPU cores a high TDP is practically unavoidable. ARM64 is especially affected, because there is no official lock-step design, yet.
Interesting to know! Xilinx says their UltraScale+ are in TSMC 16nm FinFET process, but you can still have power hungry chips in 16nm.
 
I'm a tech enthusiast, but not like that. This laptop I'm typing on now is 12 years old, and similarly my video server is of a very similar vintage (at least according to the CPU date; it was second hand). I like being able to do modern stuff on old kit, and I never buy things just because they're the newest shiny. I buy infrequently and try to buy the best when I do so that it'll last me. The Pyra was something of an exception to that for me, but I enjoyed the journey with the Pandora once I actually got one in my mitts and my Pandora was a bit tired by the end. The Pyra should be closer to the ideal of something the lasts for me.

Agreed.

If I absolutely need Windows I pull out a 2009 viliv n5, otherwise it's a cheap couple year old chromebook, a raspberry pi 4 dev desktop that replaced my 3b+, or an original eeepc 701 with a minimized debian 10 build.

I have my original wii for gamecube, wii, and some n64 emulation.

I just bought a vita but it's still old tech with a robust homebrew scene.

I do buy new tech like the pinephone, but it's usually for a specific purpose and I'm not going to mess with it till the keyboard is out.
 
There is a drawback though. I checked how much power the old Pentium draws compared to the system I bought after (that I still have)... and because the CPU is more idle during normal usage... it was something like 20 kwh less (been a while, don't remember exactly)... so if you use your machine often, that adds up. Of course, the one that draws the most power is the screen itself, and going from CRT to LCD did not make a difference (partly because the screen is bigger even if the led technology (the background light behind the liquid cristal) should use less.
Also, we need to keep using old tech, here in NL, we have so much moisure in the air, that a PC degrades/oxidates if not used (where it gets hot and blows out all the moisture).
 
Also, we need to keep using old tech, here in NL, we have so much moisure in the air, that a PC degrades/oxidates if not used (where it gets hot and blows out all the moisture).
It's a long time since I met an iron PC, I think it was an old Epson 8086 based machine. Most if not all of my newer machines use aluminium for the casings. Copper will degrade to some oxide or other if let to weather naturally, but most boards are conformally coated in some way, so that's rarely an issue in practice. Probably a bigger killer is dust, dust added to oil or grease makes a very stiff sludge that resists any turning platters, which could be bad news for floppy disk drives (hard disc spindles are better protected against ingress I believe).

But I do agree that if you go back to something like a Pentium III, it has fewer useful sleep states, so they idle hotter. Since Pentium IV though it's been better, and gains in that sense are going to be a lot smaller. You also have to consider the use of resources and energy used to make new casings and chip dopants. The sort of open cast mining that digs up the ores to make this stuff makes life harder to anyone living nearby let alone the impact on wildlife.
 
Back to the topic of future ARM cores that might be part of the Pyra's ongoing journey (but in all likelihood, not the first replacement CPU), ARM yesterday announced that big A- cores designed from 2022 and on won't have 32-bit modes:

It shouldn't be a problem for us, because Linux can and has just been recompiled for 64-bit ARM cores. It's more of a problem for RISC OS which last time I looked at least still contained quite a lot of hand crafted ARM-32 code. I've never had the same feeling that ARM-64 code is special the way I did when I read about the first ARM code in the late 80s, but I don't know how much that's me. It doesn't feel as orthogonal as it used to, but perhaps that doesn't matter for compilers these days.
 
The curent Omap 5 CPU Board dos have this Adhesive Heat Spreader Tape on it, this is something that lets me concerning about a future mandatory CPU Update: Lets say the Omap 5 is out of stock, so all new Pyra will have the new CPU Board, so the Debian have to recompile to the Newer Hardware..
So will Omap 5 Pyrae be supported even when there is a new CPU Board? I know that some software want run on the OMAP 5, but i think more on Software that only needs the Power of the Omap 5..

So: When i have to upgrade, i would have to unscrew the Botton Part which is not that big issue for me, as i dit this on my Pandora many times bevor, but then i have to loosen the Adhesive Heat Spreader, and carefully pull the OMAP 5 CPU Board Out, and put the new CPU Board in, and have to stick the new Adhesive Heat Spreader in, i changed a Graphics Card once on my Old Computer, and the Fan, which wasnt that complicadet, but the Pyra may be a bit more divicult..
 
The curent Omap 5 CPU Board dos have this Adhesive Heat Spreader Tape on it, this is something that lets me concerning about a future mandatory CPU Update:
So: When i have to upgrade, i would have to unscrew the Botton Part which is not that big issue for me, as i dit this on my Pandora many times bevor, but then i have to loosen the Adhesive Heat Spreader, and carefully pull the OMAP 5 CPU Board Out, and put the new CPU Board in, and have to stick the new Adhesive Heat Spreader in, i changed a Graphics Card once on my Old Computer, and the Fan, which wasnt that complicadet, but the Pyra may be a bit more divicult..

I'm not sure why the Heat spreader tape is a concern, likely a new CPU would need a different solution, just a case of applying it when swapping over.

Lets say the Omap 5 is out of stock, so all new Pyra will have the new CPU Board, so the Debian have to recompile to the Newer Hardware..
It may not necassarily be the case that we have to recompile everything with a new CPU, but it's kind of a non-issue as everything is opensource and a recompile away if need be.
 
Whit the Tape, my concern was that i have to peel it of in order to remove the CPU Board, on the Picture on EvilDragons Twitter its looks like the tape is just on the CPU Board, so it might be not that though than i think, i think that whit the Heat, the CPU made, it will be quite though to peel the tape so im worry that i also pull some of the SMD Chips from the Board..

Im allready damaged my Pandora inside, so i dont want this to happen whit the Pyra, also because i wants to go full Pyra, which means it will be my Notebook Replacement, it have to run after the CPU Change..
 
I think the tape you're looking at it not the heat conductive tape, it's an insulator to stop the CPU heating up the battery compartment. I'm not sure what the solution is to attach the CPU to the heat sink block that lives next to one of the USB type-A ports, but perhaps when you get a new board it might already be attached to it's specified radiator.
 
  • Like
Reactions: rSl
You're in a dark maze. You can just make out exits to the North and East. A new compute module arrives and blocks your path!


This is worth posting because of the new form factor. No longer do the compute modules need a bulky SODIMM connector to be attached to another board, what they've gone for looks rather similar to the connectors used already to connect the Pyra CPU board to it's parent board, although the spacing it entirely different, so would require a redesign.

In terms of grunt, it's based around quad 1.5GHz 'little' 64-bit processors, so less efficient than the A15 cores in the Pyra although with four of them it might be worth playing with. It also offers OpenGLES3.x via a binary blob and various video decode options that might be worth considering. And of course, assuming this form factor isn't a flash in the pan, it could be interesting when the CM5 comes around, assuming that's an actual step ahead of what we already have in our Pyras.

TLDR; have a video:
 
Yeah, though I wouldn't say it's worth it just for a Pi4 level of performance. Sure it's got some more graphics ability, but I wouldn't say it's enough to make it a saleable upgrade. What is important, however, or hopefully will be is that if they stick with this form factor for the Pi5 compute module and on, it might be worth playing around with this Pi4 computer module to be ready for that eventuality.
 
But isn't the RPi4 SOC (BCN2711) compromised by needing blobs to boot ? And how hard is the SOC to obtain anyway ? Or do you mean using the whole compute module ?

Anyway, I don't see how the compute module would help. Even if thermally and physically could fit (no idea), one would still need to adapt the connector to plug it to the mainboard, no ?
Is that feasible ? cheap?

Rpi4 CM alone is 40x55x5-7 mm That'd be wider than the Pyra CPU board ? Would there be place for connectors ?

I don't see it
 
But isn't the RPi4 SOC (BCN2711) compromised by needing blobs to boot ? And how hard is the SOC to obtain anyway ? Or do you mean using the whole compute module ?

Anyway, I don't see how the compute module would help. Even if thermally and physically could fit (no idea), one would still need to adapt the connector to plug it to the mainboard, no ?
Is that feasible ? cheap?

Rpi4 CM alone is 40x55x5-7 mm That'd be wider than the Pyra CPU board ? Would there be place for connectors ?

I don't see it
I think he means using the whole compute module, yes it would be bigger than the Pyra CPU board... still it would likely need an entirely new main board to work.
 
I don't have the dimensions of the Pyra CPU board, but the entire Pyra is 139mm wide according to wikipedia. I doubt the CPU board currently specified is smaller than 55mm although I couldn't be sure.

You'd still need a main board which has all the extra modules and connectors on. Modify that board to fit the placement of the connectors on the Pi4CM and hopefully future more applicable upgrades.

Yes, it needs binary blobs, but then so does the OMAP5 based CPU board; not to boot, but for graphics at the very least. But it simplified development costs considerably, and as documented in the video I linked should also reduce compliance costs. That fact you can't even boot the board without the use of blobs might limit the audience for these devices slightly, but the RPi has a much bigger community than we have here, so I'd suggest a hell of a lot of people don't actually care about that so much. Plus if this became known as a use of the RPi CM we might get more hits from the RPi forum and promotion from RPi sites.

I dunno personally. I do quite like the work that hns and daveshah and others have done to bring this project to fruition, but if we could be prearmed for the RPiCM5 and could make units available within weeks of the new CM being released to retail, it could make ED a lot of money, I tend to think.

Edit: Here's a photo of a dummy board. It looks to me to be about three times as wide as it is tall, and about half of the main PCB's height. That resolved out at 132x43mm, so the Pi4CM is taller by about 17mm, so would likely need to run over the top of the SD card cages, but there seems to be no likelihood of the CM board being any wider than the current CPU board.

Edit2: Or maybe spin the CM by 90 degrees if it's really taller than it is wide.
 
Last edited:
Back
Top