Some more of my ideas


A few things I'd like to see, if possible, would be;

- A proper space bar

- IR port

- Micro HDMI port

- eSATA port

- I know some would be against it but maybe microSD slots instead of full SD. If only for the sake of freeing up more room on the PCB.
 
Agree, RAM speed should clearly be the priority, and with 1G RAM you can run about every Linux distro out there without any issue.
If you just want to stare at desktop and probably run some small programs then it's enough. For example P1 is quite slow device so RAM is not a problem there. 

Nowadays computing power of modern SoCs could allow Pandora Successor to be a real low-power desktop with more performance then fastest Pentium 3 machines so it could (potentially) run quite modern (read: bloated) software like eclipse IDE that needs about 1G RAM just for itself, octave/maxima that are quite small but can use a lot of RAM (especially octave) for intermediate data storage.

So basically for serious desktop usage you need at least 2G RAM nowadays while gaming/light desktop usage (like on most mobile devices) don't really need that.

I don't think that I'll be willing to spend $600-800 for just a linux gaming device that is just a small upgrade over P1 in terms of desktop usage
 
Last edited by a moderator:
RAM speed is the main bottleneck on the current Pandora (and will be even more so if and when the full potential of the DSP gets used). So it makes sense to get the fastest we can get. Especially with a higher resolution display and a faster CPU, I think many applications will be memory-speed-limited.
What are you basing this comment on? Select some programs and benchmark them with clock speed varying. Now do it again with memory bus speed varying. You won't find anything especially useful which varies more with the latter than the former. Nor should you. Something is only a bottleneck if you can perform all of your other processing while that part is busy. That's generally not how it works with Pandora's CPU, when waiting on main RAM it sits doing nothing unless you managed to initiate a software prefetch.

RAM bandwidth has gone up tremendously on SoCs. On (1GHz) Pandora the peak is 200MHz * 2 * 4-byte = 1.6GB/s. On something like Exynos 5250 it's 800MHz * 2 * 8-byte = 12.8GB/s - that's a factor of 8. Single core CPU power, on the other hand, has only gone up by maybe a factor of 3-4. If you make both cores saturate bandwidth in that situation you'll drive the requirements somewhat close, but in practice that doesn't tend to happen. The GPU will be more likely to take advantage of that bandwidth. But if you're not playing the latest games on Android you won't have an awful lot of opportunity for that. And if you are running Android you'll appreciate having more than 1GB of RAM.

Of course, bandwidth is only part of the story. You're more likely to waste time on latency than bandwidth, even though newer SoCs are better at hiding latency than Pandora's. But then is there really a lot of room to spend more to get lower latency RAM on SoCs? The truth it, little of the latency is really spent in the RAM chip, most of it is dictated by the SoC's memory hierarchy/memory controller; the DRAM can offer access times < 20ns but typical access time on the best SoCs tends to be > 100ns. Bottom line, I really don't think you've got a lot to game here by trying to get faster RAM.
 
Last edited by a moderator:
- I know some would be against it but maybe microSD slots instead of full SD. If only for the sake of freeing up more room on the PCB.
It's again gaming vs desktop usage problem. By going microsd route you're reducing maximum capacity and speed (or heavily reducing maximum capacity without reducing speed). Not a big deal for games unless you want to have a huge amount of them installed simultaneously and bad for desktop usage because, well, you often need to haul some random crap around for whatever reason
 
- A proper space bar
agree
- IR port
That damn thing just won't die. If you really need one you could build a small IR emitter and connect it to the ext port.
- eSATA port
To small use cases, if we are lucky, we get a USB3 port, so data-throughput should only be limited by the SoC itself
- I know some would be against it but maybe microSD slots instead of full SD. If only for the sake of freeing up more room on the PCB.
I would like to keep both FullSD slots, but could live with only one, if loosing creates a real benefit. A No No for loosing both. FullSD Cards are too widely in use to be ignored - it is just great if you can just pull the SD Card out of the camera and insert it into Pandora (especially with _wb_'s PandoraImageViewer).
 
RAM speed is the main bottleneck on the current Pandora (and will be even more so if and when the full potential of the DSP gets used). So it makes sense to get the fastest we can get. Especially with a higher resolution display and a faster CPU, I think many applications will be memory-speed-limited.
What are you basing this comment on? Select some programs and benchmark them with clock speed varying. Now do it again with memory bus speed varying. You won't find anything especially useful which varies more with the latter than the former. Nor should you. Something is only a bottleneck if you can perform all of your other processing while that part is busy. That's generally not how it works with Pandora's CPU, when waiting on main RAM it sits doing nothing unless you managed to initiate a software prefetch.

RAM bandwidth has gone up tremendously on SoCs. On (1GHz) Pandora the peak is 200MHz * 2 * 4-byte = 1.6GB/s. On something like Exynos 5250 it's 800MHz * 2 * 8-byte = 12.8GB/s - that's a factor of 8. Single core CPU power, on the other hand, has only gone up by maybe a factor of 3-4. If you make both cores saturate bandwidth in that situation you'll drive the requirements somewhat close, but in practice that doesn't tend to happen. The GPU will be more likely to take advantage of that bandwidth. But if you're not playing the latest games on Android you won't have an awful lot of opportunity for that. And if you are running Android you'll appreciate having more than 1GB of RAM.

Of course, bandwidth is only part of the story. You're more likely to waste time on latency than bandwidth, even though newer SoCs are better at hiding latency than Pandora's. But then is there really a lot of room to spend more to get lower latency RAM on SoCs? The truth it, little of the latency is really spent in the RAM chip, most of it is dictated by the SoC's memory hierarchy/memory controller; the DRAM can offer access times < 20ns but typical access time on the best SoCs tends to be > 100ns. Bottom line, I really don't think you've got a lot to game here by trying to get faster RAM.
I'm just saying, if the choice is between 4GB of LPDDR1 and 2GB of LPDDR3, I would prefer the latter. Or if the choice is 2GB of LPDDR1 or 2GB of LPDDR3 and paying $10 more, I would also prefer the latter.

Of course for any component in the system there will be applications that have that component as its bottleneck. Most will be limited by the CPU (and/or its cache) or GPU, which is of course why the choice of SoC is so important. I'm just saying that the next important component is probably the RAM. It's also somewhat important in terms of power consumption, because RAM is one of those few things you can never switch off.
 
- A proper space bar
agree
- IR port
That damn thing just won't die. If you really need one you could build a small IR emitter and connect it to the ext port.
- eSATA port
To small use cases, if we are lucky, we get a USB3 port, so data-throughput should only be limited by the SoC itself
- I know some would be against it but maybe microSD slots instead of full SD. If only for the sake of freeing up more room on the PCB.
I would like to keep both FullSD slots, but could live with only one, if loosing creates a real benefit. A No No for loosing both. FullSD Cards are too widely in use to be ignored - it is just great if you can just pull the SD Card out of the camera and insert it into Pandora (especially with _wb_'s PandoraImageViewer).
eSATA is realistic though, apart from the limited connector space. Most SoC's nowadays have both eSATA and PCI-E, even ARM ones.
 
I'm quite happy with the keyboard layout and the number of keys of the P1. The main things that are missing are dedicated TAB and DELETE keys. The space bar is fine like it is now, it does not need to be bigger imo.

So two more keys would be nice (for TAB and DEL), but imo it's not needed/desired to add much more than that - adding a full extra row of keys would decrease the size of the buttons by too much I think.
 
The chance of the P2 supporting GLES 3.0 is about 40:1.... so that's something

So let's assume it's GLES 2 as usual.. with something more powerful
 
Last edited by a moderator:
I'm quite happy with the keyboard layout and the number of keys of the P1. The main things that are missing are dedicated TAB and DELETE keys. The space bar is fine like it is now, it does not need to be bigger imo.
Agreed about the space bar and adding a dedicated Tab button would be good. Personally I'd also like a dedicated forward slash button too - with that and Tab, command completion would be straightforward.

I'd also like to see the buttons moved a little away from the hinge. The current number keys are always a distraction when I need to type a number or square/curly brackets.


But other than that I don't see how the current layout can be improved on much.


I can't say I'm that bothered about the lack of an operational Del key. 'x' works just fine in vim ;)
 
I would like to see dedicated Tab and Escape Buttons.

Predefined Shortcuts for enabling/disabling Wlan/Bluetooth would be also nice.
 
Well, since you know I'm actively working on the successor yet, I think I can share some of my ideas.

Basically, it's to keep what's good and fix what can be fixed :)

What's good:

...

What needs to be fixed:

... 

What could be improved:

LCD: Better refresh rate, higher resolution (ideally 720p, but that's hard to get), size increased to 5"

Keyboard: Slightly improved layout, better feel, probably backlight

Wireless: Improved Wifi, but maybe also optional 3G/4G would be neat.

Speakers: We already have samples of much better speakers. :)

SoC: A more recent one, of course. And yes... we have the confirmation for one now, but I won't tell you :)

So, that's my basic idea - what do you think about that? :)
Great start!

You're missing one from, "What's good".

Audiophile grade Headphone Audio sound. The successor needs to be -at least as good- as the existing Pandora on the headphone audio front. Better, if that is even possible. That is one area where the existing Pandora sets a very high bar.

Screen: 5", or even 5.5" in 16:19 would work with a skinny hinge. 1080p is becoming pretty ubiquitous. Although 720p would be just fine in that size, 1080p may be better from a multi-year device run perspective.

Keyboard: I actually -like- the current Pandora's keyboard. It works great for thumb-typing. Backlight would be great - but needs to be dimable?

Wireless: 802.11 b/g/n with 2.4Ghz and 5Ghz on the n side please. My phone does 5Ghz 802.11n and it gets me 5X the throughput of the 2.4Ghz 802.11n band.

Better speakers is good and all - but not worth giving up screen real-estate for. Anyone expecting big booming bass from a pocketable device should go take an audio engineering course.

Tilt/accel/GPS/rumble - if you can add these for under $25, go for it. They're particularly handy under Android.

Stylus: Keep an on-board stylus please.

Full size SDXC slots = Thank you!!! I do think that they could be moved to the left &/or right sides though if you need or want to. This could free up the front edge of the device for a radiator.

SoC: I'm still holding out hope for the Intel z3770. I realize you're more likely to go ARM, but I can dream right?
 
I'd like a proper backwards-compatible USB port.
Yes, I'd hope that's a given. I have always assumed it was an oversight that the extra chip to support USB1.1 wasn't included on the Pandora board, since they didn't realise that when the chip supported USB2, that meant hi-speed only. But it's worth knowing for sure that we won't have to use a hub when plugging in a joystick or a mouse, and won't have to reboot if we ever forget on the new machine.
 
Last edited by a moderator:
I'd like a proper backwards-compatible USB port.
Yes, I'd hope that's a given. I have always assumed it was an oversight that the extra chip to support USB1.1 wasn't included on the Pandora board, since they didn't realise that when the chip supported USB2, that meant hi-speed only. But it's worth knowing for sure that we won't have to use a hub when plugging in a joystick or a mouse, and won't have to reboot if we ever forget on the new machine.
From my assumptions of what I read on the old forum about the design it was a conscious choice to save space on the PCB.


Not saying that the successor would be limited like this..
 
Last edited by a moderator:
You're missing one from, "What's good".
Audiophile grade Headphone Audio sound. The successor needs to be -at least as good- as the existing Pandora on the headphone audio front. Better, if that is even possible. That is one area where the existing Pandora sets a very high bar.
+1 for same or better audio wizardry, the pandora is the best mobile audio player ever.
 
Back
Top