Analyzing 4GB RAM


Or ask for a number on a scale from 1 (sticky) to 10 (switchy). That would give you more control over the amount of people switching - making it fit batch sizes better - while taking into account their willingness to switch.
Good idea, I'd put myself at 4.
 
Good to hear from ED again. It's not a voting time yet, but I'll join the fun - yeah 2GB is fine for me too.

And since we are discussing web browsing on slow computers...

WARNING: PURELY SUBJECTIVE RANT ABOUT SPEED, LOW SPEC COMPUTERS AND WEB BROWSING BELOW

I use these for web browsing for different purposes. It would be fair to mention that I'm the type of person who don't use > 3 tabs because I get disoriented, but even with only 1 tab open browsing on all three of these except MacBook is far from productive web use for me.

1. Hewlett Packard NX6320 laptop (Core 2 Duo T7600 2.3GHz, 4GB RAM, 1024x768 screen)
I installed Debian on my 2006 Hewlett Packard laptop to see how it fares as a web browsing computer and get used to Debian before Pyra hits the store. I upgraded WiFi to a more recent module for speed and security (AC7260 WiFi module). I found that this machine is not enough for comfy web browsing for me at all. Scrolling is slow, tabbing between tabs is painful, animations are sluggish, page rendering is hit or miss depending on a webpage, but mostly miss. I can actually see how browser redraws pages when I tab through them. Mozilla Thunderbird workflow is counterproductive because of major speed issues as well.

I realize this is your subjective feeling about this. I just wanted to give my own subjective experience with some older hardware. I have a Hewlett Packard Compaq 8710p laptop with a Core 2 Duo (I'm not sure of the specific one since I don't have it with me at the moment, somewhere between a T7100 and a T7500, because that's the range that was offered with them), 4GB of RAM, and a 256MB nVidia Quadro NVS 320M graphics chipset. I tend to use the original wi-fi card in A mode (it's an A+G card; A is 5GHz; G is 2.4GHz; they both max at 54Mbps). I browse with this quite a lot, and it handles 1080p YouTube videos without issue (albeit scaled to 1680x1050, the screen resolution). I don't really notice any problems when I'm browsing with this machine, and I probably have about eight to fifteen tabs open at any given time. This is with Ubuntu Studio 16.04 (this is an Xfce desktop) with software compositing turned off and Compton used for hardware accelerated compositing. I have also used the same model with 2GB of RAM, and there is not much of a difference with the number of tabs I generally might have open. Of course, if I try to run certain games that are newer than the hardware, I'm out of luck, but browsing doesn't seem to tax it much at all. I have also edited 720p video on this machine without noticeable issues with 4GB of RAM (source videos were about an hour or a bit less, and target videos were about four or five minutes each; editor was Kdenlive).

2. Prestigio Visconte V Windows 10 tablet (Intel Atom Z3735F 1.33GHz, 2GB RAM, 1280x800 screen)
Browsing on Prestigio Visconte V Windows 10 tablet is indistinguishable to my Debian HP to me personally except the fact that higher resolution helps a little. Both these computers are very slow for modern web. This tablet is closest I can get to GPD Win which has Z8700 or Z8750 CPU plus 4GB of RAM so you may be probably have a better experience with web browsing than this tablet on a GPD Win.

In my experience machines with 2GB of RAM don't seem to do very well with Windows 10. I'm not sure if that's actually because of the 2GB of RAM limitation or just because machines with that little RAM tend to be underpowered by today's standards. I would expect this machine to run significantly better with a lighter weight Linux on it, barring any compatibility issues. Usually, though, the difference between 2GB of RAM or 4GB while Web browsing is mostly just the number of tabs you can have open, especially if you use Chrome or Chromium (which are memory demanding).

Edit: One thing that I wanted to add is that experiences on x86 hardware do not necessarily seem to translate directly to ARM. It seems like everything is smaller on ARM.
 
Last edited:
We still plan to produce a stationary motherboard

pyradesk.jpg


takemymoney.jpg
 
well it's up to the ED and the dev team anyway, I just wish everyone wouldn't press for a decision so much. maybe I'll be fine with a 2gb unit, but I still have my doubts. @Swordfish II that is good to know but not my usecase. I'd need someone actually running a GAE server, load up some database and webapp, then debugging that as a live-system.
 
I would be fine with switching over to a 2GB unit. It will ensure that the project still chugs on and the potential of grabbing the eyes of would-be purchasers is a massive plus. My use cases for the pyra aren't as heavy as a few select others might be. I plan on using mine for music, minor social sites and light retro gaming [mainly arcade nights since the card lines for IIDX can get a bit long, plus I simply enjoy listening to my own music with STGs and fighting games] so the 2GB is a ok with me and I have absolutely no problems paying for the upgrade if that is possible. I feel that at the extreme would be to use it as a recorder for interviews at tournaments which in general, wouldn't be using much ram just space if I am not careful lol.

I do have an idea of where my preorder number is at on the production scale [somewhere between the first 250 regular preorder] but my focus really is the people that lie outside the first 500. If this switch helps them to get their pyras then I am all for it. This is a community driven device and really without any of you here that are reading this, neither the pandora nor the pyra would have ever been made. This is the device worth fighting for. I hope that once I obtain mine that I can start porting over games I've made for it.
 
Weird, I inherited an old dual-core Pentium E2200 (conroe - basically a core microarchitecture chip from the time after Pentium 4 but before the Core i5 etc. branding), and that has a max RAM on the motherboard of 4GB but currently has the 2GB it originally came with. Its main purpose is to be a build machine, and I've built FLIF on it already, but I have browsed the web on it particularly when I was setting it all up, and never ran low on RAM. Right now I'm not doing much on it - just playing a few AV files to generate entropy for key generation, and it's using 173MiB RAM + 88MiB shared and the vast majority of the rest of it is being used for buffering.

I did price up what it would cost to buy two 2GB DIMMs for it, but my experience thus far is that it's not necessary for that machine. Now games with lots of assets can benefit from more RAM, but in terms of ARM Linux gaming, only a couple of the games ptitseb's ported need more than the 512MiB of RAM my Pandora has and require a swap file.
 
Thanks for the update!
Why don`t you try 8GB? That might be easier. :)
OK, not funny. I am totally fine with 2GB. Maybe you find an option for 4GBs in the future, would be great. So far, my impression of this community is like they all would like to do do their best to help you with this issues. I think there are many qualified people outside waiting to do the best to find a way.
 
First: It doesn't make sense posting here or via email, PM, etc. to switch to a 2GB RAM version yet.
We're still evaluating, and based on the tests from yesterday (where I was running a 4GB RAM Pyra for more than 24 hours until I shut it down), it could be the change to the PCB is minimal and then there's no need to produce 2GB RAM versions.
IF we really want to do a 2GB RAM production run first, you'll receive an email from us :)

As promised, here are some more updates.


Regarding the Wifi antenna:
The current antenna is single-diversity (as the orginal chip didn't support true diversity).
The new one adds diversity, which improves the reception for portable devices A LOT!

This is one of the improvements.
The second one is to match the impedance of the PCB to that one of the antenna. The antenna needs a connection of 50 Ohm, whereas measurements of our current design shows it's 60 Ohm - so there's not much difference.
The change that needs to be done to the Mainboard is minimal. Nikolaus is already working on that, will produce a new dummy board and then this is also set.

And about the cases:
This is still the most annoying thing.
FormAction is still "waiting" on the new material (the one they received had some issue) and claim to have implemented all fixed to the case yet (but we've heard that).

As mentioned, we have a backup company (with which I've been working a couple of times already to produce SNES, N64, NES and MegaDrive cartridge shells) and they have a plastic/carbon-mix material that is VERY robust (supports up to 1.5t weight).

While they don't have our moulds yet, they will send samples of the material to me on Friday (as they have a different product that's currently being produced with that material).

We'll evaluate the material and we're ready to move the moulds and do the production there (in case FormAction doesn't suddenly wake up).

I wonder why it's always the case...? It also was a major headache for the Pandora.
Seems like it's the most complex thing in the world.
 
There's always a case with the case.

I have a feeling that the moment you'll get serious about switching to another company, FormAction will magically produce the cases.
 
a plastic/carbon-mix material that is VERY robust (supports up to 1.5t weight)
I appreciate you're being informative a lot - I really do, but this is barely useful information. Candyfloss can support a load of 1.5 t - you just spread out the candyfloss and the load :)
 
I appreciate you're being informative a lot - I really do, but this is barely useful information. Candyfloss can support a load of 1.5 t - you just spread out the candyfloss and the load :)
Jeez, talk about being picky with your words. Does he need to specify the area of the weight as well as the amount of surface in contact in both inches and millimeters too?

Let's put it this way: There's probably nothing you could do, short of running it over, that'd break a case made out of that material. As long as the overall amount of pressure put on it doesn't exceed 1.5 tons, it will withstand it, and it's essentially overkill for day-to-day use.

Got it? Good. That's all that matters.
 
to all those winging about ram
all I have to say is: swapon!

srsly: has anyone benchmarked a loaded browser running off an sd used as vram?
 
Jeez, talk about being picky with your words. Does he need to specify the area of the weight as well as the amount of surface in contact in both inches and millimeters too?

Let's put it this way: There's probably nothing you could do, short of running it over, that'd break a case made out of that material. As long as the overall amount of pressure put on it doesn't exceed 1.5 tons, it will withstand it, and it's essentially overkill for day-to-day use.

Got it? Good. That's all that matters.
If it was regarding a certain formula of concrete, I'd know the standard procedure to measure its weight carrying capacity woould be to make a cube 1 dm³, put pressure on it and see when it breaks.
Since here it's about plastics, I have no idea, if the same standards apply - or if it was a standard procedure measurement at all. So maybe they made some testing with hollow structures, which I'd find way more impressive. But maybe you're right and it makes no difference. Like all that matters is, that your bank account says something positive.
 
To the people with swappiness >5: Remember that it will also be a phone in the future, and you might want to run Android side-by-side(ARChon/AndroVM)... which will need 1GB to work ok-ish, but moar is better... (I am sure there will be a native Linux data over GPS solution too) - disclaimer: Data is Based on fiction.
 
Back
Top