... we're still alive!


My current system (gentoo):
du -c /usr : 21450896 total
du -c /usr /bin /boot /etc /lib32 /lib64 /libx32 /sbin : 23169820 total
Main offenders are compiled kernel source (6.5 GB) and portage/distfiles (package cache, 4.7 GB) so 16 GB should be enough but without kernel compilation on e-MMC and aggressive package caching.
That means that 32GB should be enough for one system for all usage scenarios excluding putting user data to e-MMC and 16GB is good enough for everyday tasks using default OS (debian) but there should be some limits on package caching.
Mind that for performance reasons any flash based storage should have some free space (10 to 30%).
 
Its the same on every eMMC and SSD-review - spec-sheet ive ever seen. Certainly there is nothing to lose from going up.

Here is a breakdown into Random writes/reads and sequential read/writes

http://www.anandtech.com/show/7790/...-2014-finally-a-good-emmc-solution-for-mobile

Sequential writes are double.

Some things scale with amount of chips or type of controller, and the amount of memory it is paired with, again going back to the controller.

See http://elinux.org/images/4/41/Linux_on_eMMC-_Optimizing_For_Performance.pdf for some overview.

And no, putting things on SD just isn't the same. Because finding one that isn't dog slow on IOPS, is almost impossible.

Because it is more cumbersome, and difficult (here is the normal user argument), and if you fall into the trap of thinking its the same, slows down things, all of the time. Because bigger eMMCs are faster.

SD-cards are unreliable
, wear down, have requirements for filesystems to work (eFAT for example) and cost retail price, whereas upgrading the eMMC costs bulk price.

For everyone that keeps harping on power use, using both eMMC _and_ SD is more power than one or the other.

Additionally, if you end up having to use SD for anything non-hotpluggable, you have one less slot open. Big deal for photographers. (People with money, that there are a lot of, that the pyra makes an insane amount of sense for. Photogear is heavy enough by itself, and having to lug a laptop gets oooooooold. Still you are looking at just 1 SD slot, or adapters that break and are cumbersome.)

Hopefully there will be a good and updated alternative to the OMAP5 in the future. Bonus for having an incentive to do so.
(edit: debunked, both are on the moduleboard)

Edit: Programs aren't getting smaller. Bigger eMMC is future proof for a future that many of us are already in.
 
Last edited:
I'm currently sitting in front of an Odroid board with Debian installed on it which I use as a PC (except for gaming, thats when I use my larger PC). That should be pretty similar to the Pyra OS-wise. The OS is installed on an 8GB micro SD Card. The OS itself is only 2.6GB. But that's just because i took care not to install huge and unnecessary software packages (I bet half of those 2.6GB is iceweasel aka firefox & dependencies). Anyway, the SD card is constantly filled up with all the things I put in /home. And with Debian you are always just one 'apt-get install texlive-full' away from filling all the 8GB :p which is actually a really useful thing to be able to do to be able to use the Pyra for more serious work.
So, 8GB would be enough for basic usage, but I would really prefer 16GB. And 3 Euro more isn't that much either (That's exactly one Döner Kebab less that I could buy, so its even the healthier option :D ).
 
Okay, so I have a debian install that is a couple years old. I've installed everything I need and then some on it. On my / filesystem, I'm currently using 17 GB.

diskusage.png


As you can probably see, the vast majority is actual installed programs and development tools. Some stuff in /tmp according to my normal usage, and some custom compiled tools. From this, I suspect that I personally would just barely get by with a 16 GB / considering my current usage pattern. 16 GB is acceptable, but anything below that is a no-go for me.

EDIT: And this setup has already required me to move the bigger Enterprise™ tools like Quartus/Xilinx ISE/Lattice Diamond/MatLab to squashFS images on my /home, but it'd be nice to not have to do that with stuff I use all the time that I'd prefer to load fast.
 
Last edited:
Sorry if this will be a longer rant, but comradekingu's message has a multitude of problems:
Its the same on every eMMC and SSD-review - spec-sheet ive ever seen. Certainly there is nothing to lose from going up.
Yes, if it's the same generation, most producers give bigger eMMC an extra speed bump. But ED already said that no matter how much faster it is, the OMAP5 just supports eMMC 4.5, so the extra speed of 32GB eMMC 5.1 in comparison to 16GB, also 5.1, will most probably not be usable anyways.

And no, putting things on SD just isn't the same.

Because it is more cumbersome, and difficult (here is the normal user argument), and slower. SD-cards are unreliable, wear down, have requirements for filesystems to work (eFAT for example) and cost retail price, whereas upgrading the eMMC costs bulk price.
SD cards are technologically VERY similar to eMMC, so calling them unreliable and praising more eMMC space does not make that much sense. The biggest difference is the connection, and if you really wanted to make them similarly "reliable" you could even solder it, seems very much like you want more irremovable memory. They are also not that slow nowadays, still slower than eMMC, correct, but for the majority of data (emulated games, music, movies) it should absolutely suffice. And I personally think it's a good thing to leave it to the user how much extra memory space he or she needs, most will pay less than if huge eMMC would have to be included in every Pyra.

For everyone that keeps harping on power use, using both eMMC _and_ SD is more power than one or the other.
No matter what kind of memory, it will definitly be just a minor factor when compared to the SoC or the display. But I think we agree on that from what I gathered by the way you wrote that line.

Hopefully there will be a good and updated alternative to the OMAP5 in the future.
And here is your biggest counterargument: The OMAP5, which is just able to use eMMC 4.5 and gets 5.1, will be on the same removeable PCB as the eMMC. So as long as this one uses the maximum speed of the SoC, it doesn't matter how much faster it could theoretically be because it's going to be replaced by whatever ED will deem to fit to the next SoC.


It seems as if you disliked the possibility for every user to install as much memory as one likes and want to have it fixed and non-replaceable on an open-source system just because it would fit you better. But please remember there are other users who like to have options.

Edit: Damn, ninja'd by comradekingu's edit. Oh well, I'll let it stay this way for now.
 
Where did you get the "probably not" from?

If speed goes up within the same I/O class/spec, why should the effects that increase speed be rendered totally void just because its backwards compatible? That doesn't add up.

The GPD WIN has 64GB on eMMC 4.51, that is the competition.

A 32GB SD card costs the same 10€ that I estimated going from 16 to 32GB eMMC would cost. It is a worse solution. And the cost isnt taking into account you have to pay shipping on a SD card. Assuming you dont buy it from ED.

Consumers dont magically become experts on memory technology just because there exists one or two golden SD-cards, that you have to know bin-numbers to make sure you get.

Also no, you cant wish or look away from how things actually go together:

www.bunniestudios.com/blog/?p=3554
www.bunniestudios.com/blog/?page_id=1022
http://www.howtogeek.com/196541/emmc-vs.-ssd-not-all-solid-state-storage-is-equal/

Fair point about upgrading the moduleboard. All 5.1 special-effects go to waste once its populated with an OMAP5.

It seems as if you disliked the possibility for every user to install as much memory as one likes and want to have it fixed and non-replaceable on an open-source system just because it would fit you better. But please remember there are other users who like to have options.

So you are equating a fixed, and OS-holding partition with just "memory" and then say because I'd like to increase the eMMC, for reasons you dont adress, that i somehow am against options. ...?

Would you get more options from going to 2GB eMMC? No, you would get less. You would save some €'s but it would hamper _the whole device_. Just like most other increases in space on the I/O level. But also in operation. And also you enable more possibilities by going up, that means more options.

Sweetspot means something. If you read my posts you wouldnt have to make up arguments like "because it would fit me better" If you read my older posts you would know I absolutely support modularity. Because it means less returns, more choice, and longer product-life.

I wouldnt be so much about choice if it didnt make so much sense, as in 10€ to pay for twice the storage, potentially a more responsive system, is a no-brainer. The only question is who is going to pay for it. Rarely did "open-source", the meaningless term, miss the point as much.
 
Last edited:
Fantastic update ED! The keyboard looks fantastic, I also noticed you raised the action buttons like you said you would :) I think 16gb emmc would be fantastic but 8 is also acceptable.
 
For the eMMC, either 8 or 16GB will work for me. It is nice to have some wiggle room, so if you want to go with 16, you might as well. That also allows us to play around a bit more with other OSs, and there might be some that are a bit more bloated. If I am working with a VM I will have it on an SD card of one sort or another (possibly micro, as well as whatever OS I am using that day).

The keymat does look like it might be better for typing like that. The really great thing are those gaming buttons. I think that will help, and it looks interesting too (Nintendo has done some things that looked like that, plus probably other companies).
 
Where did you get the "probably not" from?
I didn't want to go through multiple datasheets, so I just assumed that memory specifications from 2012 are slower than those from 2015, even when comparing the faster of the first ones and the midrange speed memory of the second. But if you want to go through these specifications and compare the classes of multiple generations by multiple producers, you're free to do so and correct me if I'm wrong, with the proof linked. I've honestly got better things to do on a sunday...

Consumers dont magically become experts on memory technology just because there exists one or two golden SD-cards, that you have to know bin-numbers to make sure you get.

Also no, you cant wish or look away from how things actually go together:

www.bunniestudios.com/blog/?p=3554
www.bunniestudios.com/blog/?page_id=1022
http://www.howtogeek.com/196541/emmc-vs.-ssd-not-all-solid-state-storage-is-equal/
Dude, really? These may be interesting links, but can't you just recap what you're trying to tell me instead of giving me pages over pages of articles and a video that's an hour long? I'm not an expert on memory technology, I just read a few datasheets once in a while.
 
Most SD-cards are garbage.
They are very often (a sizable amount of the total raw market) scrap sold off as a real product, even the genuine parts. And there are fakes in every distribution-channel. IOPS is for a bad one, which is 90++% of all, in the low 100s, when a modern SSD has 80.000. Meaning their use is limited to low-level photography or something like that (sequential read and write from buffer). The good ones are sold with the same partnumbers as the bad ones, so its near impossible to know what you are getting without testing it yourself afterwards. And having done that you still dont know if you got a bad one. Due to cheap manufacturing, fakes, and the general cheating, like supplying testers with golden or rigged samples. They cheat on wear-leveling. Isolation of bad sectors. And just about every other place you can cut corners. Re-jigging them is manual slave-labour. The SD-cards that arent garbage, are very seldom cheap.
Even though i am a huge supporter of modularity and choice, SD-cards for OS-level use, when the user has to decide, is far less cost-effective.
If you rely on users putting partitions on SD-cards you lose regular users. And overall, the performance will not only just suffer the ideal case, it will for the most part be swing-and-miss.

If you make the OS-partition 16GB that is right at the current pain-treshold for a regular Linux|Gnu install for people who use
Those are the advanced users. It is the bread and butter. Especially they are the types of users that may want to do something fancy, for which you need even more.

Edit: Keep it simple. Regular users dont use that much space, and are dependent on there being no hackery. If we look beyond ourselves, there are lots of different users who may want a pyra.
Blind users (keyboard and sound), photographers (dual fullsize SD), Big or difficult to drive headphones audio listeners (will pay anything), journalists (can edit and send in the field (the field is not a table you can sit down on with your laptop, nor is a phone a tool for the job)

Think about how many industry jobs virtualize some company image, or just want a thin-client.
Selling lots of pyras+spares all of a sudden makes 1000 units seem like a small number.
 
Last edited:
I hate using SD cards, I've had too many crap out on me. I would like enough eMMC storage to allow me to load my programs on as well. I wouldn't go for any less than 32GB.

Also, in the event that someone puts a dual boot OS together (ie with Android as the second OS) , the extra room will come in handy for it to run on the eMMC rather than uSD.
 
Last edited:
HARDDISK_crashers_v2.jpg
Can't do much tech work. But here is an image to cheer you on...

edit: do'h: "faith" instead of "fate". Half asleep... sorry.
 
Last edited:
What sells the pyra for me is the promise of being a handheld laptop.

There are 3-4 users in this thread alone stating 16GB isn't enough for them now, or that will run out of space in the near future.
Are these not exactly the users we want? Are we not prepared to pay or donate to have the dream device?

16-32GB is chromebook level, that is a toy in my book. I don't use web-apps, nor do a lot of people. In the price-segment, laptops rule.
 
Last edited:
We need to find a balance between price and size, and I think 16GB is enough (as I doubt many will run multiple virtualboxes off the eMMC... besides, you can put it on the normal SD Card then).
The speed must not be ignored, between SD cards it varies like night and day. It will affect boot time, browsing and general responsiveness of the device. Slow storage is especially visible with apt-get, not sure if you care about that though.
 
If I have one pet peeve (don't worry I have a few), it's running out of memory on my fancy new device. Too often have I gotten a nice new shiny android/iPod/SSD computer, only to run out of space to put all my crap a month later. Like the pyra, many of these devices had expandable memory. But that's still a headache. Now I have to move apps around, and I like to install a lot of apps. I understand with the whole open system, doing this will be much easier but the less symlinks and configfile edits in my life the better.

I hear people saying that bigger storage is better, faster, stronger because it allows things like wear leveling and magic (It sounds reasonable and I believe them). The most important thing is future proofing, if we're saying "maybe 16gb would be nice" now, I say bite the bullet and go for 32gb. I believe it would be worth the upfront cost to have a device with some actual storage built in instead of what my iPod touch had back in 2007.

Granted, I'm not the one who has to come up with the money to fund the upfront change. But, as a user, I would love to see at least 32gb for my device that's going to cost 500€. Yes, we could get by with 16gb, we could get by with 4gb and then just stick everything else on external storage. But then why have internal in the first place? For me, 32gb is essential also because I want to keep some data on the device itself, maybe some pictures, a movie or two, ROMS, Music and etc so I don't have to worry if I grabbed the right SD card.

More Space is More Space and remains valuable as long as you use it. I know I will have no trouble using 32gb of space.
---edit---
Registered Millennial
 
Would you get more options from going to 2GB eMMC? No, you would get less. You would save some €'s but it would hamper _the whole device_. Just like most other increases in space on the I/O level. But also in operation. And also you enable more possibilities by going up, that means more options.
Not everyone has a lot of cash to burn on a handheld, and bigger eMMCs can get quite expensive, therefore leading to the base unit costing a lot. A user with very limited money would just be able to either get the base unit with 256GB or nothing, while with a 16GB eMMC, he'd have the option to spend that extra money or not and could even (within limits) choose how fast this extra memory needs to be.

Sweetspot means something. If you read my posts you wouldnt have to make up arguments like "because it would fit me better" If you read my older posts you would know I absolutely support modularity. Because it means less returns, more choice, and longer product-life.
I especially get that you can be quite the elitist and would pay 1000€ for a device with a single-wide keyboard. While that might give you more options, most people couldn't even afford to pay that much for a mobile device and would only have one option: Not to buy it.

I wouldnt be so much about choice if it didnt make so much sense, as in 10€ to pay for twice the storage, potentially a more responsive system, is a no-brainer. The only question is who is going to pay for it. Rarely did "open-source", the meaningless term, miss the point as much.
I have to admit, 10€ more for a doubled internal memory would be cool, eventhough I don't believe in it being faster when paired with the OMAP5, not before I see hard proof for that. But for the extra space alone this should be worth the money.

Most SD-cards are garbage.
(Spoiler)
If you rely on users putting partitions on SD-cards you lose regular users. And overall, the performance will not only just suffer the ideal case, it will for the most part be swing-and-miss.
That's actually quite interesting, eventhough I think people nowadays understand that buying SD cards from eBay is most times a scam. I've actually not had too many problems with branded SD cards from official distributors, except for partitioning with Windows and some speed restrains I knew about beforehand. Also, Kingston SD cards generally suck.

Rarely did "open-source", the meaningless term, miss the point as much.
You're absolutely correct, I didn't think too much about it before writing. Sorry about that, it normally annoys me, too, if people just throw these terms around.

Edit: Keep it simple. Regular users dont use that much space, and are dependent on there being no hackery. If we look beyond ourselves, there are lots of different users who may want a pyra.
Blind users (keyboard and sound), photographers (dual fullsize SD), Big or difficult to drive headphones audio listeners (will pay anything), journalists (can edit and send in the field (the field is not a table you can sit down on with your laptop, nor is a phone a tool for the job)

Think about how many industry jobs virtualize some company image, or just want a thin-client.
Selling lots of pyras+spares all of a sudden makes 1000 units seem like a small number.
While I can't say anything about blind users and journalists, audiophiles don't tend to buy a mini laptop just for audio, but probably go for dedicated mobile audioplayers and don't want to pay for USB ports, a GPU, HDMI, a dual-core CPU and a big display they wouldn't use for their purpose while photographers don't want to pay 500€ for a device that doesn't run Photoshop and Lightroom, but is just good for dumping their .raw versions. And that's a problem in most industries: They often depend on software that doesn't work in Linux and if it does, it's not compiled for ARM.

Let's face it, this is a niche device for hobbyists. Yes, we can get a lot of different things on it, but if you're a professional of any field you don't usually use such a multi-purpose device. On these forums, it has often been compared to a swiss army knife that's not perfect in any field, but a good mobile multi-purpose device, and I like this comparison and want such a device and hope to use it for my job. But I don't specifically buy it for that.
 
As I mentioned, the keymat is working fine, but the hard edges of the keys makes gliding over the keyboard with your thumbs a bit uncomfortable.
That's why they will be rounded a bit.
Does this also include the ABXY buttons? They still look quite sharp-edged there.

Onto the hardware infos, or, to be exact, the eMMC:
4GB eMMC cost between 3 - 4 EUR, 8GB eMMC also cost between 3 - 4 EUR (so it's a no-brainer to NOT use 4GB).
The 16GB one is between 6 - 7 EUR.
Is this eMMC faster than the Pandora NAND? I care more about speed than size, because the Pandora NAND was pretty slow imho. Same for the Pandora SSD, no matter how fast a SSD card was in theory, on the Pandora they were all quite similar (slow).

BTW: Using the right nub click as left mouse button and the left nub click as right mouse buttons works nice and is pretty comfortable!
The Pyra Nubs can click? Wow, I somehow missed that detail completely. o_O That feature indeed gives opportunities, no longer crappy and unprecise analog-nub moving for mouse clicking! :D
 
It will affect boot time, browsing and general responsiveness of the device. Slow storage is especially visible with apt-get, not sure if you care about that though.
I'm actually even more worried about reliability. Last time I heard, as write amplification goes up because there's too little free space (rule of thumb: < 20%-ish free), traditional NAND at its advanced lithography processes of today (< 20nm ?) actually can have realistic reliability problems (significantly low write endurance) (which of course gets even worse if it's MLC or even TLC).

I don't know how relevant this would actually be for "regular" desktop use, though, maybe there's something on Anandtech about this...
 
Back
Top