Tests, tests, and more tests.


ED posted a status line from one of the Pyra that crashed. It was barely warm when it went down, so it doesn't look like a thermal issue. That's not the only issue heat sinks could cause though, the metal might be shorting something out or creating a capacitance with something, I guess.
 
I'm pretty sure half of the people who pre-ordered would be fine with 2GB RAM.
I consider it a good possibility to produce a 2GB batch and sell those.
A lot of people would be happy to do that and get a special thank you card or some little gift in exchange for the 4GB RAM.
No gift needed. I would be fine with 2, as that was always a stated possibility. I don't expect a refund for the cost difference, and might be able to pay more to cover the cost of the 4GB hype I helped start.
 
I was always under the impression that 640K...uhm 2GB RAM is enough (for my taste). Simply because it is cheaper than 4GB. :D I don't know the "official" specs of the OMAP5 and it's limitations or "safe-to-work-within" conditions. But as long as 2GB work, well fine. And as long as ED don't buy 4GB single modules it may be possible to just use one 2GB Module or 2 1GB or 4x 512MB or whatever.

Not that simple.
Two memory controllers.
Each controller's RAM must be loaded in pairs - limit 2.
Only loading one controller cuts the RAM throughput by half.
 
like most of the people here, i would be fine with a 2G-pyra too

Gesendet von meinem Aquaris M5 mit Tapatalk
 
Personally I would like to see the 4G version. But, as the 2G version is already working, I can't really understand why @EvilDragon isn't already selling it. Get some devices out, get some money rolling in, and then continue working on the next release, which may very well support 4G.
 
I can't really understand why @EvilDragon isn't already selling it

Relatively very few people, pre-ordered the 2GB version, and realistically the 4GB version hadn't really held up anything seriously in the past as there were other things to do.

I suspect that this issue may end up being something simple to fix, but you're right, I suspect that the 2GB version would sell well. Depends what's more economically viable in the longer term.
 
Personally I would like to see the 4G version. But, as the 2G version is already working, I can't really understand why @EvilDragon isn't already selling it. Get some devices out, get some money rolling in, and then continue working on the next release, which may very well support 4G.

Very simple: Because he already has the 4GB RAM modules purchased.
So not using them would be a huge loss.
and now buying the 2GB Chips in addition would cost a lot money, too. And as he mentioned, there are not unlimited funds available.

When there is no money for the 2GB Version left there is always the option to do a Kickstarter for the first Batch with 2GB RAM.
And whoever preordered will get a discount for what he already payed when he supports the Kickstarter.
 
NB : I am basing the following on logic and general experience rather than any expertise or knowledge when it comes to getting an OS to play nicely with hardware. Forgive me if any of the below is obviously wrong / ridiculous to anyone who does know what they are talking about

Yep, tests are running, but the 4GB RAM units are still freezing for unknown reason, yet they survived a memtest by both notaz and tmlind for a couple of hours without any crash... so what's the difference here?

Also what is the difference with the above findings and the findings here where the following was noted : "4GB RAM CPU Boards are working! Finally, after a long, long search for the reason, Tony Lindgren found the real issue for 4GB RAM crashes." .... "Nikolaus ran a 4GB Board without heatsink and 1,5GHz and did an apt-get upgrade which installed 400 new packages... without any issues. After a few hours, he shut it down again. Without any crash or overheating!"

Have there been any hardware / OS revisions between the units tested in the 'Light at the end' news post and those tested recently? Seems to me that they key may lie in this area.

Does the fact the latest boards survived the MEM test mean that the timings are OK, i.e. does the mem test use the same Ram timings as the OS?

Were any 2GB units produced using the latest board revision? If yes then do they work OK?

If only 4GB units were produced with the latest board revision is it possible to disable / remove 2GB of Ram from a small sample so tests can be done on a 2GB version? Would it be possible to disable 1GB and check a 3GB version?

Are there any current devices / dev boards that use the Pyra SOC and run with 4GB of Ram?

How easy would it be to get Android or another OS running on a Pyra prototype? If this could be done then would this be useful for stability testing of the RAM or would the expectation be for similar lock ups / crashes?

Worst case would be to switch to 2GB RAM, but that would be a loss of around 80k EUR... I really hope that's not necessary, as I don't have unlimited funds. But it would be an option for the first CPU board

Can you tell us a bit more about this, i.e. would the switch to 2GB mean that the boards would have to be revised to use 4 X 512MB or would they use 2 X 1GB (or 4 X 512MB) with the current hardware revision?

I'm sorry that this is not what you would've liked to hear here (believe me, I'm feeling worse than you), but I'm sticking to my promise that you the journey will be an honest one - no lies, no hidings.
Here's hoping you feel with me and continue to support me instead of getting angry.

100% understood, the honesty and open nature of what would normally be way behind the scenes is hugely appreciated.

It seems to me that you should set a time limit to solve the 4GB issue and if that limit is exceeded then you should reach out to the pre orderers and enquire whether they will be happy with a 2GB version. I bet most would.

There was also another encampment with louder drums, a mob, and pitchforks claiming that anything less than 4GB was a fail

That's not how I remember things (perhaps I missed it elsewhere on the boards), in this thread from 2015 (before pre orders were taken) there was a request that tests were done to illustrate the actual benefits (if any) comparing 4GB with 2GB + Swap. With the benefit of hindsight it would have been massively helpful to the project if said tests were performed in 2015 and the result was 'sorry the 4GB version is unstable so that's not an option in the foreseeable future'. I fully understand that, hindsight is 20/20, ED does not have a magic money tree and such tests would have been expensive and perhaps seen as unnecessary as there is no logical reason for 4GB to cause any problems.

I don't want to rehash old arguments and now is the time for the community to pull together, if you think the above linked thread represented a pitchfork wielding mob then so be it, I'll agree to differ. What would be good for morale and community cohesion (I hope) would be for pre orderers to let ED know if they would be happy with a 2GB unit if the time limit for fixing the 4GB version was exceeded and the choice was 2GB Pyra or no Pyra. My suspicion is the vast majority would be happy with the 2GB version
 
Outcomes regarding the 4GB boards, ordered by preference descending (my take):
1) current problem fixed; getting a 4GB Pyra
2) current problem fixed; getting a 2GB Pyra
3) current problem not fixed; getting a 2GB Pyra
4) current problem not fixed; getting a 4GB Pyra

I've gotten a little confused. Is the current software, which causes those problems on 4GB boards, running well on 2GB boards? If I'm not mistaken, that haven't been tried. ?
 
@Djoga'Ro It's likely some timing configuration that needs to be implemented in the boot manager. While I'm not privy on all the details, but what I gathered from my skimming of the emails and IRC, I guess there were more than one type of 4GB memory modules used in different versions of the prototypes. So what worked for one type doesn't seem to work for what was bought for mass production.
 
Disclaimer: I'm also uninformed and largely basing this on reading the kernel mailing list, but I've noticed a few things that might be pertinent.
Also what is the difference with the above findings and the findings here where the following was noted : "4GB RAM CPU Boards are working! Finally, after a long, long search for the reason, Tony Lindgren found the real issue for 4GB RAM crashes." .... "Nikolaus ran a 4GB Board without heatsink and 1,5GHz and did an apt-get upgrade which installed 400 new packages... without any issues. After a few hours, he shut it down again. Without any crash or overheating!"

Have there been any hardware / OS revisions between the units tested in the 'Light at the end' news post and those tested recently? Seems to me that they key may lie in this area.

Good spot! I think that was before the simulator results came in and they built new CPU board which are supposed to be more reliable. They *are* fully reliable is you only run the memtester on them and don't boot a linux kernel. They've also changed the heatsink solution since then, and perhaps other bits, but it's disappointing that the headline change of the simulated improvements don't appear to have improved anything, but hopefully it's not that and something else has changed and broken it. Also, different users seem to find different kernel versions are stable, which is a peculiar observation.

Does the fact the latest boards survived the MEM test mean that the timings are OK, i.e. does the mem test use the same Ram timings as the OS?

You might say that, but I couldn't possibly comment.

ian-richardson-house-of-cards-bbc-1990.jpg

Were any 2GB units produced using the latest board revision? If yes then do they work OK?

I don't know that, so I'll quote this without comment.

If only 4GB units were produced with the latest board revision is it possible to disable / remove 2GB of Ram from a small sample so tests can be done on a 2GB version? Would it be possible to disable 1GB and check a 3GB version?

I feel it's still a little early to try such crazy ideas to be honest.

Are there any current devices / dev boards that use the Pyra SOC and run with 4GB of Ram?

You can check this yourself. I don't think there are any consumer products that use this CPU with that much RAM, so the dev boards are a known set you can read up on.[/spoiler]
 
Not that simple.
Two memory controllers.
Each controller's RAM must be loaded in pairs - limit 2.
Only loading one controller cuts the RAM throughput by half.
They never make it easy for us, right? ^^""

Actually, ED said from the very beginning that it is not sure if the 4GB version can be made, and initially, such a disclaimer was even on the preorder shop page.
Then I can't imagine he already buyed the Modules for the 4GB Version.
Very simple: Because he already has the 4GB RAM modules purchased.
See above.

I have no clue about all these hardware stuff, may it be possible to build the 4GB version but somehow disable parts of the RAM so that it runs stable? Maybe at 3.5GB or 3.99GB or Whatever? So the RAM chips would be not lost if they are really already purchased.
 
@fusion_power - He has already bought a consider amount of 4GB ram for a mass production run, likely not for the full 1000+ or so units, but a good amount that sets him in a bad spot. He stated this is IRC, my scrollback doesn't go far enough back to grab his statement.
 
Would it be possible to block out half the ram of the 4GB versions to make them run stable with 2GB? I don't know much about hardware, but if that would work then maybe it could be a solution?
 
Would it be possible to block out half the ram of the 4GB versions to make them run stable with 2GB? I don't know much about hardware, but if that would work then maybe it could be a solution?
If it could work, you would enjoy half memory bandwidth speeds.
 
Back
Top