Ram Poll.


Well I'm not even sure I proved that fully, there may be some application that consumes a bit of memory without eating up all the CPU cycles. Just not sure what that would be, I'm open for some ideas, I know some may suggest VMs, but they are a bit of setup and I'm not sure I'll get to any time soon.

As I already mentioned it's pretty trivial with Audacity. A couple basic tests on my 1GHz 512MB Pandora eats up all the available memory.

I recorded a 4 minute sample at the default 44.1 kHz and 24 bit PCM, mono channel. That eats up 31 MB just for the data plus any metadata Audacity stores. Copied the track five times, adding various effects that affect the waveform. Audacity reports no space left. I'm at something like 50 MB in the OS according to the free command; guessing I've used about 256 MB after system overhead. (I tried taking screenshots but none of the repo apps work anymore.)

I've worked on much larger compositions with many more tracks, stereo audio, 96 kHz... more RAM is always useful. I'm just speaking from experience that a 2GB ceiling would definitely be felt in some cases. 4GB much less so. Are there workarounds like working with multiple project files or zRAM? Yes. That doesn't make 4GB useless, and people saying as such are betraying their lack of objectivity.

The same argument can be extended to 8GB, but again it all depends on the costs. Certainly there is diminished return from 4-8 in comparison to 2-4. If it were $4 more, I'd say hell yes. If it's $40 that would probably not be worthwhile. It's all a matter of price point and return.

We don't even know price points and we're getting 2GB anyway, so I don't have a firm stance on the matter.
 
Last edited by a moderator:
The cost of RAM is not just in dollars but also in battery life. Remember, we won't have low-power RAM. If we have exact numbers (dollars and power consumption), we can make the trade-off, but for now, I think 2GB is a good guess.

Sure, more RAM will be useful in some cases, like editing huge audio/video files or browsing with hundreds of tabs. Money and battery life are also useful in some cases.
 
Right, I'm operating under my assumption that the battery life won't be noticeable. I don't use suspend, so I suppose I am biased in that respect. To me the loss of a few days is ok, but if that's a big deal people can argue that side of it.
 
Last edited by a moderator:
The pandora had its ram upjusted in planning, and then a few times over in production. Re-buying the whole thing sucks, but knowingly supporting and having minimum nuisance is good.

But the Pandora had a better CPU before it had better ram. Plus, you wouldn't be buying everything again, just the CPU board.
I maintain my belief that 2GBs is plenty enough for the current specs. 4GBs would only see heavy use with a better CPU/GPU.
I'm still pretty curious about how much RAM could a Pyra-specific high-end 3D game take. But it shouldn't eat more than 1.5GBs without being unoptimized/GPU limited.
 
People don't buy offroad things because they are mostly offroad, they do so for convenience, not getting stuck in something that may or may not have been a problem to them, or even just a fear that it might be in the future.

Its the same thing as a good scheduler, to me those exist, and they deal with that problem.

Sometimes things are badly coded, and buffer whole heaps of junk data till the program its in is eventually closed. I care about it, but i more often care more about what im doing, right now.

I don't want to mentally plan how much memory something uses, not because im not able to, its just inconvenient, im interested in something else.

Its a lesser concern here, but low amounts of built in ram is planned obsolesence, inadvertantly or not. The pandora would for example be fast enough for many more things, for longer, if i could just load its whole OS in ram and have something left over to get things done.
 
Ok, so since 2GB is not enough to a considerable amount of people. 10%+ atm, that leaves us at some choices.

2, 4 and 8 GB are the legible choices.

2 only, with respect to the above, isnt that great.

2 and 4 is ok to everyone, and leaves you with the option of up-selling the 4, and then the same to 8 later.

2, 4 and 8 means people going for 8 will be encumbered with the soft limit at 2, its also sourcing 3 things at once.

4 only Nothing to up-sell to right away, when money is needed the most.

4 and 8 positions it at an exclusive pedestal in terms of image and commitment to future use. I think this may be the best.

All in all i think even though sourcing is cheaper on lesser parts, they are selling the whole device short, and that if the money is there for more expensive things upfront, they pay themselves off in a plethora of ways.
 
Last edited by a moderator:
@comradekingu I didn't vote, I'd rather settle the Memory debate with real metrics, show me how having more than 2GB impacts battery life or potentially degrade performance or not. Prototypes are coming, so it would be a good opportunity to settle these debates. 

 
 
Last edited by a moderator:
According to the datasheets  on comparable chips its all different sorts of draws, based on what you are doing. And the voltage isnt constant either i think. We know what kind of usage there is on a 2GB total system. I suppose the easiest way would be to give it a run on a higher configuration.

Some of the dev-boards have 1GB, some have 2GB, we could see what difference it makes there.

Personally i would gladly halve the batterylife to have twice the memory, and we know its less grim than that, so its within the scope of doubling 2 to 4 for me.
 
Last edited by a moderator:
Future proofing really isn't a valid argument here.   The CPU board is replaceable.   If 2GB is more practical now but 4gb becomes a more practical option in the future, then ED can start releasing CPU boards with 4GB of ram, and could probably trade in 2GB boards to be resold at a discount as he did with the "Classic" Pandora.

-Neelix
 
 
@comradekingu I didn't vote, I'd rather settle the Memory debate with real metrics, show me how having more than 2GB impacts battery life or potentially degrade performance or not. Prototypes are coming, so it would be a good opportunity to settle these debates.

I agree.  If anyone really feels strongly about any of this I think we should wait for some testing with complete systems.  I am pretty sure that if ED would even consider adding more he would wait until after he has seen where things stand with 2GB.
 
Last edited by a moderator:
I've worked on much larger compositions with many more tracks, stereo audio, 96 kHz... more RAM is always useful. I'm just speaking from experience ...
We don't even know price points and we're getting 2GB anyway, so I don't have a firm stance on the matter.

While I agree with your last sentence, I think the real problem in this discussion is that people assume they can apply the desktop usage paradigms they aquired directly onto the Pyra because it promises the same user interface and, for those that already used a Pandora, a boost in terms of raw cpu power. But the Pyra won't be able to fullfill these expectations. Most examples I have read so far to promote more then 2 GB leave the CPU (and GPU) power completly out of the picture, what use is more RAM if I have to wait half an our till the process that will eventually max out the Pyras RAM is finished, because the CPU was never meant for such usage scenarios. Almost the same for your Audacity example (does this thing keep everything in tha RAM? I wonder how much RAM most use on their desktop if they want to work with files that are around 2 to 3 hours long), the RAM may be the limit here, but how feasable is it that you would really use it for a task you have in mind (more tracks, stereo audio, etc.).
Just because one can connect the Pyra to a computer screen and use it with a mouse and a external keyboard does not change (one of) the main design goal of the system: beeing a portable device.

Its a lesser concern here, but low amounts of built in ram is planned obsolesence, inadvertantly or not. The pandora would for example be fast enough for many more things, for longer, if i could just load its whole OS in ram and have something left over to get things done.

In the same vain you could argue that the SoC ED (had to) choose was planned obsolence too. Other current SoCs would open up a lot more usage scenarios just with their "more" in raw CPU-Power. And what exactly would be the benefit of having the whole OS in the RAM ?

On a side note:
CPU-Board- Upgrade seems to be a thing a lot of people have on their mind (which is rather odd given that the device isn't even out yet), but I have doubts that we will see something like that in the forseable future, or maybe even ever.
 
Last edited by a moderator:
Future proofing really isn't a valid argument here.   The CPU board is replaceable.   If 2GB is more practical now but 4gb becomes a more practical option in the future, then ED can start releasing CPU boards with 4GB of ram, and could probably trade in 2GB boards to be resold at a discount as he did with the "Classic" Pandora.

-Neelix
 

It's about as valid an argument as "buy a new cpu board" is a rebuttal.
 
While I agree with your last sentence, I think the real problem in this discussion is that people assume they can apply the desktop usage paradigms they aquired directly onto the Pyra because it promises the same user interface and, for those that already used a Pandora, a boost in terms of raw cpu power. But the Pyra won't be able to fullfill these expectations. Most examples I have read so far to promote more then 2 GB leave the CPU (and GPU) power completly out of the picture, what use is more RAM if I have to wait half an our till the process that will eventually max out the Pyras RAM is finished, because the CPU was never meant for such usage scenarios. Almost the same for your Audacity example (does this thing keep everything in tha RAM? I wonder how much RAM most use on their desktop if they want to work with files that are around 2 to 3 hours long), the RAM may be the limit here, but how feasable is it that you would really use it for a task you have in mind (more tracks, stereo audio, etc.).
Just because one can connect the Pyra to a computer screen and use it with a mouse and a external keyboard does not change (one of) the main design goal of the system: beeing a portable device.

Great points. Yes, I do think the Pyra would be more than capable of handling the Audacity tasks I mentioned given my limited layman's knowledge. Applying effects to the waveforms took about a minute apiece on the Pandora; I would guess the Pyra will accomplish similar performance on higher frequency stereo tracks. It's all in memory to keep playback and seeks responsive. It's really not all that CPU-demanding to master and mix audio as a general matter, which was the thrust of my example.

However, I hold no illusions that using a handheld device for audio mastering isn't very, very niche. That certainly factors into whether the 4GB matters. There are other examples so we have to lump them together and ask if it's significant. Maybe not.

Others mention the board upgrade, also a great point. I just don't want to lean on that for what might make more sense with the base unit.
 
 
Last edited by a moderator:
There will be tasks that the Pyra is not suited for.  All machines have their capabilities - and their limitations.

HP sells a Linux pre-loaded (Ubuntu or RedHat) desktop workstation that can be configured with up to 36 processor cores and 512 GB of RAM and 8 SSD drives and 2 PCIe SSD cards and 3 PCIe graphics cards that is roughly the size of the old desktop PC towers that everyone over 30 years old remembers.  That kind of hardware exists because there are purposes that are not suitable for a laptop or server - no matter how robust the laptop is or how quiet the server is.

Each device should play to it's strengths.  The Pyra's greatest strength is it's portability.  Portability includes and is very much dependent on battery life.  There are lots of computer platforms out there that can perform the same processing tasks, run the same software and emulators, write to the same media, run more or less the same OS, etc.  None of those 'complete' solutions will fit in your pocket and have all-day battery life.

Although I wish we could have LPDDR, we don't.  The memory in the Pyra is going to be a very significant energy hog.  How significant is directly related to the size of the memory bank.

So - if you have a data processing task that really does require more than 2GB of RAM, shuttle the data to a computer -somewhere-, take it over with some form of virtual desktop software and process it.  The Pyra is a lot of things.  It is -not- a full-on high-powered professional desktop workstation.  It should not be expected to do those tasks well, or at all.
 
http://www8.hp.com/us/en/campaigns/workstations/z840.html
If only I had enough money that I could waste it one things like that.

I think size is a lot more important than battery life for the Pyra.  With frequent use, which will vary by person, I am guessing that a Pyra will need to be charged somewhere between daily and weekly.  Standby time will be important to some people, however I think what can be done with it is more important.  How long the battery will last while it is being used is more significant than how long the battery lasts when it isn't being used.  If some want it to last longer they could always have it hibernate between uses.  Not sure how long they would need to wait for everything to get running again.

A big part of things like the Pyra, and their users, is that they get pushed into doing things they aren't intended to do.  I agree with a lot of what you are saying, but some of us like to try things, especially those we "can't" do.  This isn't RAM specific, however having additional RAM may help some of us in our pointless and stupid fun (although some seem to want to do serious work, but work is also a lot of fun sometimes).  

I get where you are coming from, but I don't think any of us can really understand where everyone else is coming from.  We each have our own ideas of what we want to do.  Part of the idea of the Pandora and Pyra are already trying to do more with less (mostly size).  Some people have a clear idea of what they plan to do, others would like hardware that will suit what they might need, and others don't care.  We all won't get what we want, and even if it doesn't make sense, or reduces our battery life, we still want it.

What I really don't understand is if battery life is so critical why are so few saying they would be ok with 1GB?  It seems that should be enough for most tasks, so shouldn't that be what some people are asking for?  Why does 2GB instead of 1 make any more sense than 4 or even 8GB?  If it is because someone might use it, that possibility is present with more RAM as well.
 
We really need some numbers in this conversation. It's kind of pointless to argue about this since I'm sure everyone has their limits either way. People who want more RAM have an upper limit at what cost they want it. People who think it's not necessary have a lower limit at which they have nothing against it. Then there's the gray area between where this conversation and cost-benefit analyses belong.
 
It's easy for me:
Cost: Don't care
Battery: Important
Too less RAM than needed: Most important
 
Back
Top