Power, Memory and Schematics


Please don't even think about sending out a 1.2 GHz version, please.

@Grench(and others who want some thing fast): We know that 1.5GHz is possible(eval boards), thus everything less indicates a problem(sofar a unknown one). Maybe even a problem that can happen with 1.2GHz later on, after some aging.
 
D32_bobjob said:
Please don't even think about sending out a 1.2 GHz version, please.

@Grench(and others who want some thing fast): We know that 1.5GHz is possible(eval boards), thus everything less indicates a problem(sofar a unknown one). Maybe even a problem that can happen with 1.2GHz later on, after some aging.
Click to expand...

rygD said:
1.7GHz or death.
Click to expand...

jonlad1 said:
Yes yes, please find the fix before releasing :)
Click to expand...

If a fix were easy or the actual problem and solution known that would be a logical demand. However, with the knowledge that the cause is currently unknown and the time that it will take to invent a solution if one is even possible without resulting in a complete re-design of the SoC module - or even possible within the constraints of the system... There is a lot of if going on there. A potential outcome could be that the 1.5GHz simply isn't possible to attain inside of the development budget and time frame. Hope it works out, but it isn't the end of the world if it doesn't.

Regardless, there is a working (albeit 80% of ideal clock speeds) solution. If those parts that are already created can be distributed to developers for getting the Pyra specific software underway, I don't see any reason not to get those parts sent out. For those units slated to go to the prototype pre-orders, if there are parts available after getting them to devs, @EvilDragon can simply give them a choice to take one from the 1.2GHz prototypes or continue to wait.

Frankly I'm pretty thrilled that they have working hardware - even if there is this one caveat.
 
Yes, Grench. I don't know how hard/expensive/likely it is to get 1,5 GHz for the production run. But it seems obvious to me that developers should get 1,2 GHz protoptypes ASAP. Even if eventually they can be replaced with 1,5 GHz prototypes (or production units). I mean if a developer gets us usable phone software for Debian (in the Pyra), can't we really each of the other 900 of us pay one more EUR to get her a fixed CPU board ? I'm exagerating a little but you get the idea.
I guess since it was advertised as 1,5 GHz prototype recipients could be asked, but even if they insist on 1,5 GHz I'd send a 1,2 GHz at least temporarily and see later if there's a way to upgrade those prototypes based on whatever it takes (donations, prototype owner contributions, I don't know).

it is a waste to keep 29 1,3 GHz CPU boards unused while chasing for the Holy ultimate 0,3 GHz missing. Get those in the hands of developers first and sort it out later what to do.

This does not mean not trying to fix what's possible before the production run.It's just not delaying work that could happen in parallel to that.
 
EvilDragon said:
The reason is most probably a bad decoupling (not enough capacitors).
We WILL do another production run with quite a lot of improvements in that area, so that should fix the 1,5GHz speed.
Click to expand...
Can you not prototype that fix by splitting a trace and bridging it with a through hole capacitor? I guess that'd be easiest on the main board, rather than the CPU board, but it shouldn't really matter where you put the capacitor I think as long as it's downstream of the power management chip.
 
@Grench, I wasn't completely serious when I said that. 1.7GHz was officially reduced by TI (I may be wrong), and I don't want the Pyra team to die, from sleep deprivation or something else, because of that. I want the Pyra to be made, and if they think they can get it to 1.5GHz they should keep at it. It is better to do it right than to do it quick.
 
EvilDragon said:
The reason is most probably a bad decoupling (not enough capacitors).
We WILL do another production run with quite a lot of improvements in that area, so that should fix the 1,5GHz speed.
Click to expand...
Sounds great, do whe have enough bare cpu platines to produce another run ??
When i remember Correcly it takes 2 Months (tm) for a new run of bare cpu boards..
 
EvilDragon said:
The reason is most probably a bad decoupling (not enough capacitors).
We WILL do another production run with quite a lot of improvements in that area, so that should fix the 1,5GHz speed.
Click to expand...
Is there any possibility to overclock the SoC by adding few voltage like overclock PC CPUs? If so, is it possible to refer some overclock ideas to just make the SoC runs stable enough at 1.5Ghz (not overclock it)?
 
levi said:
Can you not prototype that fix by splitting a trace and bridging it with a through hole capacitor? I guess that'd be easiest on the main board, rather than the CPU board, but it shouldn't really matter where you put the capacitor I think as long as it's downstream of the power management chip.
Click to expand...
No, that's not possible, as all relevant traces run on an inside layer and it's actually pretty important where the capacitors sit.
 
matzesu said:
Sounds great, do whe have enough bare cpu platines to produce another run ??
When i remember Correcly it takes 2 Months (tm) for a new run of bare cpu boards..
Click to expand...
It can't be done with the current CPU boards, we need new PCBs.
 
Ted said:
Is there any possibility to overclock the SoC by adding few voltage like overclock PC CPUs? If so, is it possible to refer some overclock ideas to just make the SoC runs stable enough at 1.5Ghz (not overclock it)?
Click to expand...
One of the first things I tried when testing, but doesn't have any effect.
 
matzesu said:
I ditnt meant the CPU Boards (Populatet) i meant the bare platines, whitout the Parts,
The Population should be done fast, but to make the platines needs some weeks..
Click to expand...

Traces need to be changed and vias (through-holes) as well as pads for various capacitors added. That requires new bare PCBs.
 
Well, i would also take a 1,2 ghz unit, as im not realy wants to make that demanding tasks, like Running a Windows VM,
I just wants to play my Pandora Emus (from Atari 2600 to N64/DS/ PSX) plus Dreamcast and maybe PSP..
Also OpenMorrowind, and Doom 3..
But thats it, i ditnt want more..
So if you dotnt know where to put the 1,2 ghz Unit, i would take one...
the only thing is, that the rest should work like on the other Units, and it should run all Pyra Games..
But im no programmer, so it want help much if im take on of them..
And because im Number 27 or 28 in PreePreeorder, it want take long until i get my 1,5 ghz unit..
 
matzesu said:
Well, i would also take a 1,2 ghz unit, as im not realy wants to make that demanding tasks, like Running a Windows VM,
I just wants to play my Pandora Emus (from Atari 2600 to N64/DS/ PSX) plus Dreamcast and maybe PSP..
Also OpenMorrowind, and Doom 3..
But thats it, i ditnt want more..
So if you dotnt know where to put the 1,2 ghz Unit, i would take one...
the only thing is, that the rest should work like on the other Units, and it should run all Pyra Games..
But im no programmer, so it want help much if im take on of them..
And because im Number 27 or 28 in PreePreeorder, it want take long until i get my 1,5 ghz unit..
Click to expand...
Can Doom 3 run on the Pyra? Would love to see that on video.
 
You must log in or register to reply here.

Similar threads

EvilDragon
Schematics and packages
2 3 4
Replies
60
Views
19K
spud42
spud42
EvilDragon
Stop the sirring noise!
6 7 8
Replies
141
Views
73K
Grench
Grench
EvilDragon
Finished! (Well... the schematics, that is)
8 9 10
Replies
192
Views
45K
Silent-Hunter
Silent-Hunter
EvilDragon
Analyzing 4GB RAM
16 17 18
Replies
359
Views
157K
javaJake
javaJake
EvilDragon
I'm out of ideas for news-titles
15 16 17
Replies
323
Views
132K
TDT
T
Back
Top