Back to normal!

[Silent-Hunter]

Question, what are the consequences of disabling self-refresh on the RAM? Does the CPU have to tell it to refresh every time or something?

 

[TrashyMG]

Notaz and Tony are working hard on it... relax.

 

[elw3]

Nice color, id like it a tad lighter tho. Like the color nintendo used for the NDS light.

 

[AVahne]

Don't forget to taste the plastic! After they're safe to lick of course, but yes it's very important that the Pyra is not only durable, and feels good in the hand, but also feels good in the mouth as well.

 

[dfkman]

Well, the fact we're back to discussing colors is a good sign, if nothing else . Good work all, and I hope we find the proper fix soon!

 

[Silent-Hunter]

Notaz and Tony are working hard on it... relax.

I'm fine, just curious is all.

 

[Nintendo]

You mean like this?:

Linked from the thread with all of the colours.

That's (too) hot pink!

Maybe, something more gentle like translucent baby pink?

Or maybe...Sakura baby pink?

 

[troy]

Question, what are the consequences of disabling self-refresh on the RAM? Does the CPU have to tell it to refresh every time or something?

From what I understand the main consequence of disabling self-refresh is power consumption. I can't say for certain how much, so that's something ED will have to test. Assuming we still have decently long battery life, it's probably a compromise worth making at this point in the project.

Further reading suggests that it's probably going to only affect power usage in idle. So regular use might not be affected.

 

[Silent-Hunter]

From what I understand the main consequence of disabling self-refresh is power consumption. I can't say for certain how much, so that's something ED will have to test. Assuming we still have decently long battery life, it's probably a compromise worth making at this point in the project.

Further reading suggests that it's probably going to only affect power usage in idle. So regular use might not be affected.

Awesome. According to Askarus, the Pyra already lasts something like 10 hours of normal use. So that probably won't change much, just how long it lasts when it's idle and not being used.

 

[Askarus]

Awesome. According to Askarus, the Pyra already lasts something like 10 hours of normal use. So that probably won't change much, just how long it lasts when it's idle and not being used.

I'm more concerned about idle and power saving.
That's a real issue.
The only options I have right now is to either turn off the Pyra or turn off the display. It's annoying if I forget to turn it off and it's empty the other day.

What's the downside of turning off the DLL instead of the Power refresh.

 

[CommanderB]

I'm more concerned about idle and power saving.
That's a real issue.
The only options I have right now is to either turn off the Pyra or turn off the display. It's annoying if I forget to turn it off and it's empty the other day.

What's the downside of turning off the DLL instead of the Power refresh.

I think that the self refresh is just a small part of the power saving. The dynamic CPU frequency adjustment and the dimming of the display should contribute much more. I do not know, if low level stuff like Smart Reflex is implemented in the current pyra kernel yet. The other stuff should be possible to enable and to implement so I would not worry about that

Gesendet von meinem SM-N9005 mit Tapatalk

 

[BeMoreViking]

Hmm. Sounds like mixed news on the mailing list. Good that we think we're narrowing it down, bad that we're going to have to choose the lesser of two undesirable options to make it work. This makes me wonder if it isn't an issue that's internal to the hardware. I suppose the chances of having TI look to determine if there's a hardware issue with the design of their boards are probably slim. (Or at least advise on the selection of a workaround.)

Regarding refresh (and not knowing much about it) I was wondering if anybody can tell me if this Wikipedia article is relevant to the discussion. Link: https://en.m.wikipedia.org/wiki/Memory_refresh

If so, that makes it sound like you need the memory refresh to avoid data loss. But from what I've read on the mailing list, it sounds like they could also be talking about something else.

Anyways, I'm glad we're making progress on debugging this. I'm curious to see how this all shakes out.

 

[CommanderB]

Hmm. Sounds like mixed news on the mailing list. Good that we think we're narrowing it down, bad that we're going to have to choose the lesser of two undesirable options to make it work. This makes me wonder if it isn't an issue that's internal to the hardware. I suppose the chances of having TI look to determine if there's a hardware issue with the design of their boards are probably slim. (Or at least advise on the selection of a workaround.)

Regarding refresh (and not knowing much about it) I was wondering if anybody can tell me if this Wikipedia article is relevant to the discussion. Link: https://en.m.wikipedia.org/wiki/Memory_refresh

If so, that makes it sound like you need the memory refresh to avoid data loss. But from what I've read on the mailing list, it sounds like they could also be talking about something else.

Anyways, I'm glad we're making progress on debugging this. I'm curious to see how this all shakes out.

As I know TI could in theory provide support for the memory but the project is too small and so the support would too expensive. As ED said, there are other companies available, who are specialized in fixing RAM issues, but again the price tag for the service is the problem.

In the mean time Tony suggests on the mailing list, that the issue is hardware related. But let's not speculate and wait for official statements.

Gesendet von meinem SM-N9005 mit Tapatalk

 

[levi]

Regarding refresh (and not knowing much about it) I was wondering if anybody can tell me if this Wikipedia article is relevant to the discussion. Link: https://en.m.wikipedia.org/wiki/Memory_refresh

If so, that makes it sound like you need the memory refresh to avoid data loss. But from what I've read on the mailing list, it sounds like they could also be talking about something else.

Yes, as far as I grok the term, that's what they're talking about. Given manufacturers can't guarantee the consistency of data after a few milliseconds, I'd expect a computer that wasn't actually refreshing it's ram at all wouldn't survive a boot, especially if there's a bootloader that waits a second or two before choosing an image to load, so the system must still be refreshing its RAM somehow. It may even be more performant to refresh the RAM using the kernel to do it rather than letting the RAM chips do it themselves, because then you can tune how often it happens, and it's possible the CPU triggering a refresh of a few rows will use less additional battery charge compared to what the RAM would use doing the same, or maybe it's something the MMU does itself.

Edit: On the mailing list today, Tony reckons it'll drain the battery faster with it disabled. We'll have to see what that actually means in the tests that Notaz and ED are running, I guess.

 

[kuru]

This says DLL helps with faster RAM clock rates when conditions for transfer like temperature and voltage change

 

[Silent-Hunter]

It sounds like the 4GB version uses less power than the 2GB when under load. That's odd.

 

[levi]

Other news from the mailing list:

ED Wrote:
Notaz Wrote:

> - shoulder buttons are barely usable (tend to get stuck pressed)

The case is not a final one. I'll tweak the buttons together with the
new company once we've switched the molds.

Sounds like the clear cases made at the original plastics firm in Athens might not be usable as real replacement cases even if they are sold on later. The original intent was to sell those to case modders and it sounds to me like that plan might be falling by the wayside to the mission to get a usable case.

 

[AndiTheBest]

If the ram-refresh needs to be done by the cpu finally, we may could use one of the M4 cores to save power.

 

[crazyhorse]

Looks like Notaz has some issues with his unit, maybe its an older prototype -

Hi,

I think my latest unit is supposed to have all fixes, so I was
surprised to still have some of the fallowing issues. I though my old
unit was faulty, but now it looks some of the problems may be actual
design issues.

- the LCD is flickering at many brightness levels (same as older
unit). It depends on lighting conditions how much it disturbs, but I
can always see it and just can't stand it. There is no flickering on
full brightness, and it's almost invisible on the lowest levels, but
all others are unusable for me. FWIW I don't see flickering on any
brightness level on my pandoras.
- the wifi is extremely slow on my unit, tens of kB/s while the phone
does at least 1MB/s on the same router on the n (5GHz) band. On a
different b/g router it does 200-300kB/s and at least works (older
unit doesn't work at all there).
- the speakers make a disturbing loud crack while DE is loading (same
as older unit). No idea what's going on there, trying to play a boot
sound? I've turned down and muted everything in audio mixers and it
still does it. Ended up disabling everything audio related in dts just
to get rid of this.
- shoulder buttons are barely usable (tend to get stuck pressed)
- there are periodic sata errors in dmesg ("exception Emask 0x10 SAct
0x0 SErr 0x4040000 action 0xe frozen"), don't know what it means, and
I don't have anything connected
- still no signs of life from OTG port, but it's charging at least
- lid switch isn't working (old unit too)
- touchscreen is unusable on the edges. With dead nubs it makes some
UI elements inoperable
- (broken board?) nubs are never responding. On older board, they die
after some time (devices on the bus stop responding)
- (broken board?) noticeably draining the battery while off (older
unit doesn't do this)

confirmed fixes:
- USB squeek is gone
- write protect works on SD2 now
- wifi now works on b/g, but very slow on b/g/n
- keypad is a lot better

 

[Djoga'Ro]

If the ram-refresh needs to be done by the cpu finally, we may could use one of the M4 cores to save power.

I'd say, for that the RAM chips must be buggy first. Or a compatibility issue in hardware. Otherwise one should ever strive for utilizing the beneficial features given.
[doublepost=1539590892,1539590776][/doublepost]@crazyhorse Didn't you follow that conversation any further?