ARM vs. X86. Lets vote

To make things clear now: Which Architecture do you want for a Pandora Successor (assuming both are


  • Total voters
    104

I find the smiley you used after the question was a perfect summary of the situation with MIPS.
 
My personal preference is for ARM, if comparing like with like in terms of overall CPU / GPU power / performance per Watt / price / availability etc.

However my understanding is that in reality there is no current Gen (i.e. A15) ARM SOC that is available to a project like P2. There is a strong possibility that a current gen X86 SOC (Bay Trail) would be available to a project like P2.

If the only available current generation SOC available was x86 based then, with some major regrets, I'd rather see P2 with an x86 based SOC.
 
Given the fact that cutting-edge big.LITTLE A15/A7 SoCs seem to be hard to obtain for a small niche product, maybe a good strategy would be to use a decent A9-based SoC like the Rockchip RK3188 for now. It would still be a major upgrade compared to the current Pandora, and it's probably relatively easy and cheap to get at the moment. Then later on it can be upgraded (relatively painlessly, like with the 1GHz Pandora) to something in the upcoming RK32xx series (announced for 2014Q2), which will be A12-based.
 
Let's open another thread !
Maybe this one about planning out how we're going to run support for Windows XP, Vista, 7 and 8 all simultaniously. Sice its not been decided that Windows will be the main OS on the P2 (because we all know Windows is the best OS ever) we need to ensure all users recieve equal support.

When if there were 4 seperate forum sites? One for each OS version so as not to favor any one user group
 
Given the fact that cutting-edge big.LITTLE A15/A7 SoCs seem to be hard to obtain for a small niche product, maybe a good strategy would be to use a decent A9-based SoC like the Rockchip RK3188 for now. It would still be a major upgrade compared to the current Pandora, and it's probably relatively easy and cheap to get at the moment. Then later on it can be upgraded (relatively painlessly, like with the 1GHz Pandora) to something in the upcoming RK32xx series (announced for 2014Q2), which will be A12-based.
So you're basically saying "hey dont buy the initial run because better ones are comming"??
 
Given the fact that cutting-edge big.LITTLE A15/A7 SoCs seem to be hard to obtain for a small niche product, maybe a good strategy would be to use a decent A9-based SoC like the Rockchip RK3188 for now. It would still be a major upgrade compared to the current Pandora, and it's probably relatively easy and cheap to get at the moment. Then later on it can be upgraded (relatively painlessly, like with the 1GHz Pandora) to something in the upcoming RK32xx series (announced for 2014Q2), which will be A12-based.
So you're basically saying "hey dont buy the initial run because better ones are comming"??
If you like waiting for 3 years...
 
Given the fact that cutting-edge big.LITTLE A15/A7 SoCs seem to be hard to obtain for a small niche product, maybe a good strategy would be to use a decent A9-based SoC like the Rockchip RK3188 for now.
If it was a choice of Rockchip RK3188 Vs Bay Trail I'd go for Bay Trail. This is due to gut feel that a premium priced P2 based on a SOC that could probably be bought in a $40 PC stick would make P2 a rather hard sell.

It would still be a major upgrade compared to the current Pandora, and it's probably relatively easy and cheap to get at the moment. Then later on it can be upgraded (relatively painlessly, like with the 1GHz Pandora) to something in the upcoming RK32xx series (announced for 2014Q2), which will be A12-based.
The decision of when P2 should come to market will of course have a highly significant impact on SOC choice, who know what will actually be available SOC wise in 6 months time?
 
Last edited by a moderator:
I want my Pandora 2 to Transform into a robot. We can call it: Retrowave. He can have working SD cards that transform into robots and other creatures.

To get an idea of what I meant by the SD cards, take a look at this USB example:

ku-medium.jpg


-God Ginrai
 
...


Now seriously, 


I am pretty sure that A15/A7 L/b SoC can be obtained from Media Tek pretty easy.


...
What makes you sure of this?
Because there are plenty of low volume devices powered by their SoCs, and the DEV team should really consider them as a factor, if they insist to put Big.Little SoC in the next Pandora.

for sure Media Tek will agree a lot easier to provide the SoC than Samsung.  :)
 
Last edited by a moderator:
Given the fact that cutting-edge big.LITTLE A15/A7 SoCs seem to be hard to obtain for a small niche product, maybe a good strategy would be to use a decent A9-based SoC like the Rockchip RK3188 for now.
If it was a choice of Rockchip RK3188 Vs Bay Trail I'd go for Bay Trail. This is due to gut feel that a premium priced P2 based on a SOC that could probably be bought in a $40 PC stick would make P2 a rather hard sell.
Well yeah... but maybe still not as hard a sell as an x86 based device that is 1 generation behind, I would think?

The world of x86 is even worse in this respect, isn't it?

And eventually the P2 will be a generation late again?

I mean, at least the way I've noticed this stuff works in the x86 side seems to be: Latest intel corewhatever comes out, 1 month after the common as muck electronics shops give the prevous gen netbooks and laptops away free with flat TVs... happens less with ARM based devices.

But then again I don't follow "the scene" that much these days, for example, my most powerful desktop PC is a Pentium III and it still works nicely... this is just an impression gained from casually reading the ads every third saturday morning...

However, I think in the PC side people have slowly stopped staring at GHz alone and are starting to also count cores. And I think something similar is going on with ARM... Which is kind of sad when you consider the role of GPU etc. these days. Most people bying x86 iron need a quarter of the CPU they get and thrice the GPU they get...

I guess a curse of an x86 device is also that the usual PC suite of benchmarks can be run and thus it will be very clearly and precisely put into the performance chart bars among it's brothers... This is how Apple managed to convince the world PPC is still better just before the transition to Intel: There was no good direct comparison, there was always a "yes, but..."-argument going on.

In the very optimal world I'd even trade a millimiter or two of thickness to have a truly swappable/upgradable SOC in P2 for future proofing them, but that would not do any good, because in the end they never are swappable, even if a new revision comes out...

I agree with _wb_ fully on this.

And I remind that as already discussed in other threads, the P2 might not need as much CPU as it needs love from the devs...

I also love the attitude of Pandora devs, it has so much in common with the demoscene of days long gone by. Basically what they've done is:

"Oh, so the Pandora can't do it, because it's 'too slow' eh? Eh? Well you just watch us buddy!"

As long as that attitude is maintained I don't care if ED decides to run the P2 on dog turds...
 
I still find this type of irrational fear a combination of tragic and funny.

x86 sucks! Long live ARM!

I don't think Pandora would have been able to reach 10+ gameplay hour with an x86 SoC. Lately, Intel is making great progress on the low power race, but x86 CPUs have a lot of bizarre instructions and "features" (like internal instruction translation from CISC to pseudo RISC) that will always drain a lot of power.

Sure you can install Windows + PC software on an x86 CPU, but I don't want Windows at all, if I have to sacrifice battery life!

x86 has lasted way too long. I'm taking each opportunity I can trying to bury it. E.g.: the last laptop I bought was an ARM Chromebook (I thrown away ChromeOS and installed Arch Linux, and now it's a nice machine).

Abandon "Wintel" and embrace Linux+ARM  ;)
Your 10+ gameplay question - ANY of the modern mobile SoC examples should be able to do the same clock cycles as the existing Pandora on less energy if allowed (or forced) to throttle back to the Pandora's level.  Your fears of loosing, "reach 10+ gameplay hour with an X86 SoC" is even more unfounded than anyone stating that it will - there isn't enough information yet, but everything found so far (to my knowledge) has been positive/favorable for the Z3770.

Nobody is talking about pre-installing, forcing or even encouraging Windows installs - regardless of SoC.  The Z3770 makes it theoretically possible for some people, if inclined, to hack an install - of course they could do the same with Mac OS X if they want.  Freedom includes allowing others to do things that you personally find distasteful.

"Abandon "Wintel" and embrace Linux+ARM"

Recently ARM SoCs have become more closed with proprietary drivers and restrictive NDAs to sign.

Intel's recently released Z3770 (The X86 in question) appears to be fully open FOSS.

Hypothesis: Linux + Z3770 >= Linux+ARM.  Can it be proven one way or the other through facts instead of rhetoric?

We can always found defaults, but anyway ARM > x86.

And btw, x86_64 is better, because it was made by AMD.
See the above hypothesis.  Please prove it wrong using facts, not irrational rhetoric.
 
Last edited by a moderator:
...

Now seriously, 

I am pretty sure that A15/A7 L/b SoC can be obtained from Media Tek pretty easy.

...
What makes you sure of this?
Because there are plenty of low volume devices powered by their SoCs, and the DEV team should really consider them as a factor, if they insist to put Big.Little SoC in the next Pandora.

for sure Media Tek will agree a lot easier to provide the SoC than Samsung.  :)
It's been stated in pretty much every topic that has discussed possible SoCs that Media Tek is not an option because they have no Linux support.

-God Ginrai
 
I don't think Pandora would have been able to reach 10+ gameplay hour with an x86 SoC. Lately, Intel is making great progress on the low power race, but x86 CPUs have a lot of bizarre instructions and "features" (like internal instruction translation from CISC to pseudo RISC) that will always drain a lot of power.
That's been proven wrong over and over again, ask any prominent CPU designer today and they'll tell you that for something targeting mobile device level perf and perf/W the "x86 tax" is minor and easily outweighed by higher quality manufacturing and design. What made a lot of sense in the 80s to early 90s has basically turned into wishful propaganda today.

We can always found defaults, but anyway ARM > x86.
I don't know what's with you, you cry foul when a thread doesn't meet your quality standards but continue to make post after post like this >_> This doesn't help the level of common knowledge about this topic here, which is currently full of misconceptions.

And btw, x86_64 is better, because it was made by AMD.
Your Intel hate is clearly clouding your judgement. x86-64 was a pretty pedestrian and by the books change. x86-64 is better because it's newer (AVX and AVX2 make a lot of fundamental improvements too, and those aren't AMD). It wouldn't have made sense in 1985 when Intel made their last big mode change. Intel had little real problem going along with it. AMD didn't bring any unique magic to the table.

Well yeah... but maybe still not as hard a sell as an x86 based device that is 1 generation behind, I would think?

The world of x86 is even worse in this respect, isn't it?

And eventually the P2 will be a generation late again?
What are you talking about? What's a generation behind about Bay Trail? It has best in class CPU performance before you even consider how much better its perf/W is. In terms of manufacturing capability TSMC isn't going to be able to offer a comparable process until 2015 at the earliest, while Intel is probably releasing Atoms on their next manufacturing process only next year.. Their release schedule has gotten a lot more aggressive and I don't expect it to slow down.

Whatever you're describing about Core iWhatever processors doesn't have a relation to their mobile SoCs, really nothing about them does except that they're x86 (which is less of a critical aspect than many people are treating it as)

I guess a curse of an x86 device is also that the usual PC suite of benchmarks can be run and thus it will be very clearly and precisely put into the performance chart bars among it's brothers... This is how Apple managed to convince the world PPC is still better just before the transition to Intel: There was no good direct comparison, there was always a "yes, but..."-argument going on.
Apple wasn't convincing anyone except their diehard fans who would have believed anything they told them.

I agree with _wb_ fully on this.

And I remind that as already discussed in other threads, the P2 might not need as much CPU as it needs love from the devs...
Why do you think it won't get as much love as Pandora 1 got? Or do you not think Pandora 1 got enough love?

For some reason people keep ignoring efficiency (except when they come in to say that we must use ARM because it'll always be more efficient, which is nonsense - for this space anyway). You can have "good enough" performance but will you have "good enough" power consumption?
 
I prefer ARM, because it's currently what we have and it's my preference for the successor, but if an x86 processor is the best choice and has comparable power consumption as an ARM equivalent and is more available then sure why not look at that as a possibility.

Personally, ARM was something new and exciting to me it's why I even got a Pandora in the first place, but would a x86 successor stop me from buying it? No it wouldn't.
 
I think we should shove in some FPGA's and then we can tell it when to be ARM and when to be an x86 ;)
 
What are you talking about? What's a generation behind about Bay Trail?
No, I meant Bay Trail itself will be old news soon, if Intel is successfull, because...

TSMC isn't going to be able to offer a comparable process until 2015 at the earliest, while Intel is probably releasing Atoms on their next manufacturing process only next year.. Their release schedule has gotten a lot more aggressive and I don't expect it to slow down.
...Intel keeps releasing new stuff so fast? How often would a motherboard etc. need a revision for the P2 to not be a generation behind "others"? At which time sales hurt again, because people seem (once again, purely my perception, not a fact) more aware of x86 performance difference than ARM performance difference?

Well, yes, I know, it's definitely the lesser evil to not have quite the latest and greatest all the time, when compared to being "just slow to begin with"... My point was simply: I don't really think an x86 P2 would be any easier to sell than an ARM one because of this aspect.

Why do you think it won't get as much love as Pandora 1 got? Or do you not think Pandora 1 got enough love?
OK, I'll rephrase concentrating on fact, not ideals:

If, for reasons of availability or whatever, the P2 has to make do with an older SoC instead of the latest and greatest stuff, it probably isn't a huge disaster, because so far the devs have continually shown us that less can be more than enough.

And in that case it is better to just choose a "safe", possibly even "cheap" SoC which gives a sufficient upgrade compared to the current Pandora and has potential for being upgradeable later on.

As for my ideals, they are simple: When a device garners attention from devs that push that device to its limits, I end up liking the device. And in this respect the Pandora has done incredibly well as far as I've seen so far (soon I might know better, when my Pandora finally arrives via UPS...). I don't believe that has happened to an x86 device since the mid 90's... they grow old too fast. But I realize this thinking is just me being a dinosaur at heart :p .
 
Back
Top