I'm out of ideas for news-titles

[EvilDragon]

Please provide a source to back that
[doublepost=1504719379,1504718804][/doublepost]

The mobile version (which we wanted to use) uses PoP memory like the Pandora and is the OMAP5430.
It only exists as prototypes though, so we had to use the OMAP5432 which is meant for tablets or automotive, yes.

 

[Swordfish II]

The mobile version (which we wanted to use) uses PoP memory like the Pandora and is the OMAP5430.
It only exists as prototypes though, so we had to use the OMAP5432 which is meant for tablets or automotive, yes.

OK so tablets. Still mobile.

Again, my anger is absolutely not at you. At all. I have already fully paid for my Pyra ( final payment was Christmas last year) and I will love every second regardless of spec (watching a show on the TV via Pandora right now). I will likely even get one for the wife.

You are open and honest. If TI said "hey in most use cases it is stable at 1ghz, but possible up to 1.7 depending on cooling," I would say "OK cool"

But they don't, while you do. To me that is super shady of them.

I'll stop cocking this thread up. Just a peeve.

 

[Letalis Sonus]

OK so tablets. Still mobile.

Actually, the OMAP5430 is the one that was advertised as the tablet SoC in their product description. The whole OMAP5 platform is being referenced as a mobile platform, but that includes things like automobile as well. The OMAP family was indeed initiated for smartphones, but TI's core business with ARM has always been the industrial and automotive markets - if they ever make a successor for the OMAP5 it won't be called OMAP anymore, those ties were cut with the OMAP5.

The OMAP5342 wasn't being designed to be used with small Li-Ion batteries, the most obvious hint for that is the lack of support for LPDDR2, which was being used by the OMAP5340 and the whole OMAP4 series. The more power hungry DDR3 RAM of the OMAP5340 does not provide any noteworthy performance benefit, those SoCs offer the same max bandwidth on both interfaces.

 

[Grench]

OMAP 5432 and 5430 I think - otherwise correct.

https://www.engadget.com/2011/02/07/ti-announces-omap-5-two-high-performance-and-two-low-power-core/

"Availability
TI's OMAP 5 platform is expected to sample in the second half of 2011, with devices on the market in the second half of 2012. The OMAP5430 processor is offered in a 14x14mm Package-on-Package (PoP) with LPDDR2 memory support. The OMAP5432 processor is offered in a 17x17mm BGA package with DDR3/DDR3L memory support. "

 

[Kev2442]

I have another idea: All problems have been fixed and Pyras start shipping next month.

Adjusting reality accordingly is left as an exercise to the reader

We know it ain't true, 'cause otherwise it would have been in two months™

Pssst, hey, @Dark Pulse : @Shenmue blew your cover !

 

[bubblebee]

Just read about the LENOVO Thinkstation P320 Tiny on golem.de:

Silently cooking the processor

There is only one problem: The cooling of the Thinkstation is not designed for the permanent but dynamic load in games and has problems with that. The system tries to work as quietly as possible. Therefore, the temperature load on the CPU is high. This is noticeable when the CPU is 53 degrees Celsius when idling - with six monitors connected it is 57 degrees. The fan, on the other hand, is not audible in this state. The system is also very quiet even during light office work and surfing the Internet.

That's very good in itself, but when playing games the small fan turns audibly on, but it still can't cool the processor down enough so that it has to clock down to 800 MHz at regular intervals. The chip is then heated to almost 79 degrees. It follows that in CS: GO, the frame rates go down to 38 fps. It's strange that the CPU is already reducing so much at this temperature.​

Even a big company like LENOVO has difficulties in cooling down their processor in small factor desktop PCs and even though they use a fan.

ED: So please take the 2GB Pyra with the currently established cooling system into production as soon as possible and hand it over to your customers like every established and big manufacturer already would have been done.

Then go ahead and develop a new CPU board with 4/8/16 GB quad or octa core CPU.

It's really that easy.

 

[Dark Pulse]

Can you bring me a Pyra for my Christmas gift this year Santa?

Well, my grandfather on dad's side looked exactly like Santa, so sure, I guess.

There'd better be several hundred dollars worth of cookies and milk though.

Having more RAM than the SoC can process against efficiently does not make it a 'do 100% device'.

Jesus Christ, enough with the "it can't process more than 2 GB of RAM efficiently" already. This was debunked, and you're doing nothing but spread misinformation because you're all butthurt that 4 GB of RAM is holding up THE PRECIOUSSSSSSSSSSS. Every one of us who ordered one of these things has a stake in this venture, not just you, so quit crapping on everyone who opted for the 4 GB version.

As for the non-garbage point of your post, no processor out there can be a "do 100% device." Do you run the sorts of programs that supercomputers run on your desktop? Would those supercomputers run Windows 10?

Different processors are for different market segments. It's why your desktop processor has more power than your phone or tablet and can do more than those, and then servers kick your desktop's ass, and then supercomputers beat the servers.

It's about balancing power, performance, and general stuff like cooling. Supercomputers need air-conditioned rooms. Servers generally need active cooling all the time. Desktops need active cooling but not nearly to the same extent, and most mobile stuff depends on passive cooling for everything, because especially there, battery life is king.

 

[rygD]

I have another idea: All problems have been fixed and Pyras start shipping next month in Two Months.

FTFY

Edit: I guess Kev2442 beat me to it, or something.

 

[Grench]

Jesus Christ,

Sorry, wrong guy, but I can pass along a message if you want.

enough with the "it can't process more than 2 GB of RAM efficiently" already.

https://pyra-handheld.com/boards/th...stions-ahead-of-preorders.77046/#post-1369486

This was debunked, and you're doing nothing but spread misinformation

Not really. No in use examples exist that extend the @TrashyMG 2GB RAM on OMAP example. Processor has problem keeping up with a mixed active apps load prior to running out of RAM. If you have a counter example, please present it.

because you're all butthurt that 4 GB of RAM is holding up THE PRECIOUSSSSSSSSSSS.

Thank you for your concern. My butt is doing fine. The 4GB of RAM was an extraneous spec on the OMAP5. Those who understood this are fewer in number than those chasing specs and those who try to use windows usage models that don't apply well.

Every one of us who ordered one of these things has a stake in this venture, not just you, so quit crapping on everyone who opted for the 4 GB version.

I do have a bigger stake than most, but not a controlling one by any stretch. People ordered the most expensive one considering it to be the 'best'. That is natural.

As for the non-garbage point of your post, no processor out there can be a "do 100% device." Do you run the sorts of programs that supercomputers run on your desktop? Would those supercomputers run Windows 10?

Not quite, but close. Algorithms are the same, but I can't keep up on the permutations they can. I run Debian Stretch on all 4 of my workststions, and it can run a supercomputer. In both cases the RAM size is appropriate to the processing capabilities though.

Different processors are for different market segments. It's why your desktop processor has more power than your phone or tablet and can do more than those, and then servers kick your desktop's ass, and then supercomputers beat the servers.

My desktops have nothing to fear from servers of their same age, though I often treat them as servers, they're technically desktop workstations (HP Professional Workstation towers Xeon & Opteron, hardware RAID, etc). Server class hardware stuff.

It's about balancing power, performance, and general stuff like cooling. Supercomputers need air-conditioned rooms. Servers generally need active cooling all the time. Desktops need active cooling but not nearly to the same extent, and most mobile stuff depends on passive cooling for everything, because especially there, battery life is king.

OK, so you get the concept of scaling cooling to the total heat generated by a system. I can work a mixed analogy from there.
Consider the OMAP5's processing capability as being your system heat. 2 GB of RAM (cooling in analogy) is enough to saturate and cool it. The other two GB of RAM (cooling in analogy) never leave the firetruck. But, the fire engine, suspension, tank, and fuel capacity needs are all higher cost just to haul around all that cooling that generally never gets used.

Now, there can be cases where it makes sense. Where RAM is loaded but then accessed infrequently. A VM sitting idle for example.

 

[Djoga'Ro]

@Dark Pulse , it's no use, @Grench seems to think after processing 2GB the OMAP5 is worn down and needs to be replaced in terms of max. service life or something.

 

[Bosbeetle]

https://www.pyra-handheld.com/pipermail/kernel/2017-September/002018.html

maybe Im to early but

 

[slaeshjag]

https://www.pyra-handheld.com/pipermail/kernel/2017-September/002018.html

maybe Im to early but

I've lost count on how many times we've gotten a fix for the 4GB instability. Let's at least wait until the fix has been tested on more than one board.

 

[Wrath Of Khan]

Screw it lets just go for 8 gig...

 

[PCXT]

8GB is a bit for virtualization, scientific computation or high-quality games. What can be virtualized on this CPU? Smaller Pyra? Better with chroot. I don't think about running FEM simulation (or even GA optimization) on this CPU. High-quality games... On this platform, are there? There will be lots of emulators of course, but I doubt they require so much RAM. 2GB is nice, 4GB is totally sufficient and leaves a big space for OS's buffers to make things run faster.

If there are separate settings for RAM with this stability issue and it's a total enigma, can these tests be automated and just "mechanically" optimized? No, not brute-force but e.g. evolutionary algorithm taking non-crashing time as fitness function? I made genetic algorithm reverse engineer me a serial display protocol.

 

[JDTAY]

I see some new messages by Tony on the kernel mailing list.

Did he fix all the things yet? >_>

 

[Grench]

@Dark Pulse , it's no use, @Grench seems to think after processing 2GB the OMAP5 is worn down and needs to be replaced in terms of max. service life or something.

Not at all, but once saturated with commands dealing with objects in RAM, adding more RAM and more things to do isn't going to make it do any more work faster.

Clearly there are others here who think you can make an MX-5 (Pyra) haul the weight of a semi truck (Desktop computer or server) if you simply add more tires.

 

[levi]

I dare say that ptitseb and other super-porters will be able to produce projects that need swap on a 2GB Pyra somewhere down the line. In that case, having more RAM would improve speed versus using swap. But personally I think it'll be a while until that happens, by which time I'll probably have a new CPU board and will have bought the 4GB version of that if it looks to be heading that way.

If it turns out video editing is practical on the Pyra, then 4GB might be a bonus there right away, but that's not one of my use cases personally.

 

[FBnil]

maybe Im to early but

 

[Autian]

I personally like to use tmpfs on ram, especially if I'm testing or using file copying scripts, just to minimize storage wear or to get things faster done. Otherwise the free ram could be used by programs or the system itself to cache stuff (already mentioned by someone, as well as the tmpfs thing).

 

[Djoga'Ro]

@Grench You don't need lots of things to do to justify more RAM. You could have a tiny programm crunching huge loads of data and depending on the task at hand data might need to be touched more than once - having it still sit in RAM speed things up a lot then. As long as reading from and writing to RAM is magnitudes faster than with storage and the CPU can go through data faster than storage can provide and consume, more RAM can be justified.
I don't know if I'll really need 4GB or not. Better safe than sorry is my crede here (for myself). And far less am I able to decide, if 4GB are useful to others or not.