Discussion in 'Pyra News' started by EvilDragon, Sep 5, 2017.
@klapse : here, I'll gift to you this "-h", use it well
It is starting to become a dirty word, if you haven't noticed. See what happens when you try to do something nice.
I preordered a Pyra with 4 gigs but honestly I don't care. 2 gigs are perfectly fine. A stable unit is more valuable to me!
I want to voice out that I'm very happy with this project and the Pyra is still the most anticipated tech thingy for me. Its gonna be so much fun : )
Stability is Death
It's quite.... Trashy..
Now that the thread is fully derailed, we as a forum of netizens need to assist ED with the great difficulty detailed when he wrote the first post.
Apparently he's low on sleep, coffee, imagination, time or other, so let's pitch together and assist in the way only 5000 anonymouseish strangers can.
Has anyone seen my Pyra? It was here somewhere in the pile with everyone elses.
Since the plastics guy is behind schedule we milled a batch from aluminium.
Would someone with more creativity than me like to give it their best shot to continue? (That one is from me, Grench. I'm out of ideas, anyone feeling the muse?)
Since the OP talks about cooling, I reckon he should simply have entitled it 'Cool news'
Cool Devices? No, wait. That's erm. That's... Something else.
There's two benefits to higher-than-sustainable-cpu-clocks: better performance and better battery life.
A lot of games still benefit from having the CPU burst through its AI and render code and then go idle as quickly as possible so the GPU can start its own work as soon as possible.
ARM does a lot of testing on power usage and found that there's a sweet-spot where it's more power efficient to have the CPU clocked higher in bursts so it can go to full sleep sooner than run at a lower clock and stay active all the time. In other words: higher clocks (within reason) CAN improve battery life.
This is called "race to sleep" or "race to idle" https://www.quora.com/What-is-race-to-halt-strategy-to-make-a-processor-energy-efficient
As for marketing I've seen actively-cooled 1.7Ghz tablets throttle down to 800Mhz within 15 seconds of 50% CPU usage (only ONE core out of two at 100%.)
They're still marketed as 1.7Ghz.
If the Pyra can manage 1.5Ghz for a few minutes on two cores in full use and then throttle to 1Ghz all on passive cooling it's still much better.
Very few mobile devices can sustain their top clocks for more than a few seconds but it still improves performance and battery life in real world applications.
I'm gonna throw in my 2 cents now as well.
I'm definitely not a novice linux user any more and I have a few tasks that are bound to take up more than 2GB of RAM, but I could also very well outsource these tasks to my laptop.
What I mainly want to do with the Pyra is to have a stable and long lasting companion for my scatterbrain.
I need access to my calendar, notes, cloud storage, an IDE (worst case, Sublime would do the job, but I mainly work with IntelliJ for comfort-reasons).
I need to be able to be called on the Pyra and I need mobile data/GPS to not kill the battery instantly (I can leave them turned off for most of the day)
Ideally, I would be able to run Android apps without having to run a VM or their webpage-equivalents for longer periods of time in the background, ready to be usable within seconds.
Examples: I have an idea that I need to take note of, before I forget it again.
So really my priorities are: (high to low)
Applications that might need>2GB RAM
When the 4GB issues are resolved, I would really like to see is if someone could run a test, showing the differences in battery life on the different models on standby, idle and under load.
If that turns out to be a significant difference (the definition of that is subjective), then I would like to change my preorder to a 2 GB Mobile edition model.
For now, just focus on resolving the issues and please stop bickering about the RAM.
It's not going to get us anywhere.
I can't find it right now, but I once read a quote from EvilDragon:
"The last thing we should do is rush it."
Where has that gone?
People are totally convinced that a new CPU board will happen for sure.
I have serious doubts about it, call me pessimistic. There are a few threads here talking about how many dead ends ED got into when searching for an available SOC. Not to tell that Pyra should sell well in order to be make a new CPU worth.
I preordered a 4GB version mostly for that reason, if this is the only version of Pyra ever to exist, at least I want it to make all of it. That said I would take a 2GB Pyra at anytime.
ED searched for a high-end 32bit ARM CPU years ago.
Things should be a lot easier those days, and we will get a 64bit ARM CPU :^) .
Yeah, I didn't pick up on that until after I posted.
--- Double Post Merged, Sep 9, 2017, Original Post Date: Sep 9, 2017 ---
I would love to see that happen. Or I think I would, not knowledgeable enough. What does the Linux software landscape look like for 64b ARM?
--- Double Post Merged, Sep 9, 2017, Original Post Date: Sep 9, 2017 ---
Power of positive thought?
Curious... with the SoC&RAM&eMMC board, what is the largest usable physical dimension for the SoC while leaving room for the rest?
Yes, there are more factors to consider.
It works well.
I play with Gentoo aarch64 on my Jetson TX2, I compiled xfce, firefox, libreoffice to name a few.
Of course I have to stick with nvdia's kernel until the mainstream is filled enough.
And there's still the proprietary display drivers problem until nouveau handles it.
Regarding the 4GB vs 2GB and battery life.
Depending on the storage device you use, having the extra 2GB act as cache (Linux does this automatically) can help reduce power usage by having the storage devices stay inactive and the CPU can return data immediately rather than run all the code required to read from the storage device and "pollute" the CPU caches which reduces efficiency.
It depends on how much power is required to read from the storage device (CPU, DMA, Storage device workload), how much is avoided, what the consumption difference is between 2GB and 4GB, and how many I/O reads are saved.
Also, keeping well-behaved-when-inactive programs (eg: NOT browsers) in memory, rather than unload/reload them repetitively, can save battery life because of all the CPU work required to load and dynamically link a program.
Additionally, the extra 2GB can be used to mount /tmp from tmpfs to use RAM instead and spare the flash storage. Which I bet is how the Pyra OS will be configured by default.
You can do the same with your browser cache but this means the browser cache will be lost on every reboot (which might be what someone wants anyway)
All of these can offset the higher power drain of the extra RAM.
BTW, anyone knows what the exact mWh difference is between the actual 2GB and 4GB setups when idle and working? If that info was shared I missed it.
Interesting, but the Pyra is closer to a desktop in typical usage than to a mobile phone, hence the usage and demand for power will not necessarily follow the same patterns.
Like with the Pandora battery life, the fact that everyone else is doing it does not imply that the Pyra should do it too. The Pandora could have been marketed with "25 hours of battery life", but for intensive gaming that would not have been true.
Agreed, but I prefer it to be marketed as such instead of pretending that customers are too stupid to understand this nuance. E.g. "Runs at 1Ghz sustainable, with short bursts up to 1.5Ghz" (or whatever the numbers will eventually turn out to be). There is nothing wrong with educating the audience a little instead of spreading misinformation like the rest.
The fact that a processor can run at 1.5Ghz, does not imply that the device itself can too. Cooling capabilities and battery concerns should be considered as well. All I say is, let us just be open about what the Pyra can and cannot do.
Separate names with a comma.