Pandora Not Powerful Enough?


EvilDragon said:
I mean, if you draw 2D stuff using OGLES2, how many polygons do you need? 2?

You would usually use 2 triangles per sprite, the more sprites you have on screen at a time (like a ton of particles like modern 2D games use) the more PPS you need.
 
Last edited by a moderator:
WizardStan said:
Except the PSP2 will not have a keyboard.
I doubt it as well, but you never know, mate. SONY might surprise us and have a touchscreen, dual analogue sticks, querty keyboard, and all?
Let's just wait for the official unveiling, shall we.
 
Last edited by a moderator:
Awakening said:
EvilDragon said:
I mean, if you draw 2D stuff using OGLES2, how many polygons do you need? 2?

You would usually use 2 triangles per sprite, the more sprites you have on screen at a time (like a ton of particles like modern 2D games use) the more PPS you need.

Ah, sure, for particle effects :)
Well, I doubt there will be that many games on the Pandora that will need so much particles or sprites at once that it will actually really matter speedwise :)
 
Last edited:
MWeston said:
My only addition to this topic is that I want to share my experiences with these injected molded case samples. Pickle made a very stable version of Quake2 with dual nub support and it is the first time I have truly enjoyed playing a 3D shooter on a handheld. I played all the way though Resistance: Retribution for the PSP and that was a bit awkward with the four buttons to look around. Having the second nub act like a mouse is just so fantastic. Being able to move that cross hair anywhere I want at any speed I want when being attacked from everywhere was wonderful. That's what Pandora is all about and not necessarily trying to show off the fastest clock speed although unofficially, it's doing quite well there too.


Oh you tease!
 
Last edited by a moderator:
EvilDragon said:
Awakening said:
EvilDragon said:
I mean, if you draw 2D stuff using OGLES2, how many polygons do you need? 2?

You would usually use 2 triangles per sprite, the more sprites you have on screen at a time (like a ton of particles like modern 2D games use) the more PPS you need.

Ah, sure, for particle effects :)
Well, I doubt there will be that many games on the Pandora that will need so much particles or sprites at once that it will actually really matter speedwise :)


lol, i'dd like to see a bullet hell shoot 'em up game issue 116k bullets at 60fps on my pandora...
 
Last edited by a moderator:
EvilDragon said:
Awakening said:
EvilDragon said:
I mean, if you draw 2D stuff using OGLES2, how many polygons do you need? 2?

You would usually use 2 triangles per sprite, the more sprites you have on screen at a time (like a ton of particles like modern 2D games use) the more PPS you need.

Ah, sure, for particle effects :)
Well, I doubt there will be that many games on the Pandora that will need so much particles or sprites at once that it will actually really matter speedwise :)

Well, not being an expert on the subject, I would say that pure PPS numbers mean very little as you would be drawing all these little alpha blended images all over the screen. PPS values, as far as I know, are just solid untextured triangles. But in the end 2D or 3D makes very little difference, both are textured triangles these days. The exception would be Flash based games, how I wish the Pandora would support Flash as there are so many indie games made in Flash nowadays.
 
Last edited by a moderator:
EvilDragon said:
Well, the same SoCs running with higher clockrate is nothing special, it will obviously happen sometime. But this doesn't need a completely new design, you could simply make a new batch with those new SoCs if they are pin-compatible.

Regarding those A9-based ones... I doubt they will become mainstream in 2010.
The Cortex A8 has been released approx. 1,5 - 2 years ago. How long did it take for the first mainstream devices to appear?

I wonder about this.. on the surface I would expect adoption to take a while based on how long Cortex-A8 adoption took. But I think Laurent knows things we don't and has a reason to be more optimistic. What I also think is that the commercial landscape has changed a lot since 2008 and things have gotten a lot more competitive in the high end mobile space. We'll probably see Cortex-A9's adopted as soon as possible to climb to the top of phones (fully expect Apple to jump on this with as quickly as they transitioned to 3GS) and to make a more competitive major entry into the netbook game.

I think a lot of the work towards Cortex-A8 and implementing SoCs contributed towards the implementation of A9 and its SoCs. I see the CPU as kind of a stepping stone.

Still, if Pandora is released before these A9 devices surface - and it'd damn well better - it'll still be in a comparable speed grade. You won't be able to say that it's significantly outclassed by anything currently on the market. OMAP3530 is still the best choice OP could have made, as it was 2 years ago.

EvilDragon said:
Are you sure about that?
I checked the technical data of the new iPhone, it can do 667MHz, has 32kB L1 Cache and NO L2 Cache (compared to the OMAP 3530 which has 112 KB (DSP) and 32 KB (ARM Cortex-A8) L1 Cache and 256 KB (ARM Cortex-A8) and 96 KB (DSP) L2 Cache).
I couldn't find out if the iPhone 3 GS has a DSP, but the Pandora does have one.
The GPU used in the iPhone is the PowerVR MBX-Lite which is a bit slower than the MBX. The Pandora uses the PowerVR SGX which is superior (and even fully backwards compatible to the POWERVR MBX).

So I wonder why the iPhone should be more powerful...?

I'm not sure that you read Anandtech's breakdown very well.. iPhone 3GS probably has 256KB of L2 like the Pandora, and it has an SGX 535 which is more powerful than Pandora. Don't confuse it with the older iPhone 3G, that's a last generation device.

Yeah, the Pandora has a C64x DSP, which makes it more balanced than other Cortex-A8 SoCs currently out. However, I consider this a bit of a wildcard until someone actually uses it for anything, in particular on the Pandora itself. I'm sure someone eventually will, possibly even something substantially useful, but it won't have nearly as widespread benefit as having another Cortex-A8 core or a faster one.
 
Last edited by a moderator:
IF the only factor in 2D games was PPS, and SGX can do 14M of them, and each sprite would take two triangles would this mean 7 million sprites on the screen at once? :p

I kid, I kid. It's a nice thought though :p

I agree with ED on realism in games, which is pretty much why I have a wii and not XO or PS3. The more demanding (possibly because of said realism) ones I do want to play I want to play on a PC anyway.

On topic, the pandora is plenty powerful for anything I want it for. If the controls are on the spot, I'm going to be very, very, very happy when it arrives :)
 
B-ZaR said:
IF the only factor in 2D games was PPS, and SGX can do 14M of them, and each sprite would take two triangles would this mean 7 million sprites on the screen at once? :p

I kid, I kid. It's a nice thought though :p

PPS is mostly meaningless for 2d style rendering which depends mostly on fillrate ... to give you an idea ... I was able to pump about 2500 alpha-blended 32x32 sprites on iPhone 3gs ( which is slightly more powerful than pandora as it contains SGX535) while running at ~30 fps.
The performance was entirely tied to fillrate as all quads were being rendered with a single batch ( and were being transformed on the CPU side)
 
Last edited by a moderator:
Exophase said:
I wonder about this.. on the surface I would expect adoption to take a while based on how long Cortex-A8 adoption took. But I think Laurent knows things we don't and has a reason to be more optimistic. What I also think is that the commercial landscape has changed a lot since 2008 and things have gotten a lot more competitive in the high end mobile space. We'll probably see Cortex-A9's adopted as soon as possible to climb to the top of phones (fully expect Apple to jump on this with as quickly as they transitioned to 3GS) and to make a more competitive major entry into the netbook game.

I think a lot of the work towards Cortex-A8 and implementing SoCs contributed towards the implementation of A9 and its SoCs. I see the CPU as kind of a stepping stone.
If I knew anything specific about release dates I wouldn't have written anything about it :) My line of thought is exactly the same as yours: A8 paved the way for A9 and now ARM is being taken more seriously than it was even a year ago.

So I think we'll see products using A9 before end of this year, but that's just what I hope will happen, it's not based on any insider info.

Still, if Pandora is released before these A9 devices surface - and it'd damn well better - it'll still be in a comparable speed grade. You won't be able to say that it's significantly outclassed by anything currently on the market. OMAP3530 is still the best choice OP could have made, as it was 2 years ago.
That's what I think too. OTOH I also think that now OMAP3 isn't amazing anymore, while I thought it was 2 years ago.

@ED: Sorry, I didn't mean you were a fanboy, that was just the feeling I got reading the whole thread not your particular message.
 
Last edited by a moderator:
Exophase said:
I'm not sure that you read Anandtech's breakdown very well.. iPhone 3GS probably has 256KB of L2 like the Pandora, and it has an SGX 535 which is more powerful than Pandora. Don't confuse it with the older iPhone 3G, that's a last generation device.

Oh damn, you're right. They start with the old one on that article. I didn't expect that :)

Yeah, the Pandora has a C64x DSP, which makes it more balanced than other Cortex-A8 SoCs currently out. However, I consider this a bit of a wildcard until someone actually uses it for anything, in particular on the Pandora itself. I'm sure someone eventually will, possibly even something substantially useful, but it won't have nearly as widespread benefit as having another Cortex-A8 core or a faster one.

So iPhone 3 GS and Pandora do basically have the same horsepower. iPhone has a bit better SGX while the Pandora has a DSP. It all depends on the apps.
I still doubt many homebrew / independent games will make full use of the SGX, so the end user will probably not notice any difference :)
 
Last edited:
Interesting. I mentioned, that the "3D" specs of the Pandora-Chipst are pretty clear and listed. But what exactly are the "2D-Specs" of such a Chip? Is there actualy a measurement for 2D Power? x-Pixel per second or x-sprites per second?
I remember the C64 manual and it offered the C64 with an amazing skill to generate ONE Hardware-Sprite (usualy the Cursor AFAIK) :D

And I believe the Pandora IS Powerful enough. I think you can pull out a lot of cool stuff, if the Devs have learned how to use every bit of the OMAP Chipset. :D I mean, look what late games pulled out of old Consoles, SNES with Donkey Kong Country for Example or "Turrican" onto the C64. It pulled more out of the Commodore than ever was inside! If Devs can manage to make good usage of the OMAP Chip(s) and handle timing-problems and all the small starting issues then the Pandora can be a powerful Device for many years. In thePast, Devs needed only 5MHz or less to make Outstanding Games. Imagine, what you could do with 500MHz if you learn to optimize your Stuff. ;)

notaz said:
EvilDragon said:
Errm, the WIZ and Dingoo can do this, and the Pandora is about 10x faster...
You should be more careful with such claims, it depends on the task. Picodrive, for example, runs ~3x faster than good old GP2X with some overclock (Pandora is on stock speed). Currently 32x emulation doesn't even reach 3x, probably because of lots of cache misses it causes.
Cool, so 1MHz onto the Pandora is at least as powerful than 1MHz onto the GP2X? :) I was able to play most Megadrive Games with picodrive underclocked, 166MHz, for Sonic with best Sound and all Features enabled 200MHz. When it also runs at 200Mhz onto the Pandora it would be really underclocked and save battery Time. VERY nice. ^_^
 
Last edited by a moderator:
EvilDragon said:
Exophase said:
I'm not sure that you read Anandtech's breakdown very well.. iPhone 3GS probably has 256KB of L2 like the Pandora, and it has an SGX 535 which is more powerful than Pandora. Don't confuse it with the older iPhone 3G, that's a last generation device.

Oh damn, you're right. They start with the old one on that article. I didn't expect that :)

Yeah, the Pandora has a C64x DSP, which makes it more balanced than other Cortex-A8 SoCs currently out. However, I consider this a bit of a wildcard until someone actually uses it for anything, in particular on the Pandora itself. I'm sure someone eventually will, possibly even something substantially useful, but it won't have nearly as widespread benefit as having another Cortex-A8 core or a faster one.

So iPhone 3 GS and Pandora do basically have the same horsepower. iPhone has a bit better SGX while the Pandora has a DSP. It all depends on the apps.
I still doubt many homebrew / independent games will make full use of the SGX, so the end user will probably not notice any difference :)

the sgx 535 is clocked at about 200mhz compared to the 530's 100mhz. That isn't a bit more powerful, it's twice as powerful. Oh well :(
 
Last edited by a moderator:
you are quite right, the Pandora is not powerful enough, now go away and buy something more powerful :)
 
fusion_power said:
Cool, so 1MHz onto the Pandora is at least as powerful than 1MHz onto the GP2X? :) I was able to play most Megadrive Games with picodrive underclocked, 166MHz, for Sonic with best Sound and all Features enabled 200MHz. When it also runs at 200Mhz onto the Pandora it would be really underclocked and save battery Time. VERY nice. ^_^

I think usually clock for clock the Pandora will perform a little to a lot better than GP2X. This is mostly because of the L2 cache, but it's also helped by the dual integer pipelines, branch prediction, and 64-bit interface to dcache.

However, it is possible for something to perform worse clock for clock on Cortex-A8 than ARM9. It has some stalls that ARM9 doesn't and some increased penalties. For instance, a typical interpretive emulation loop that has an indirect branch from the same location every cycle will take 13 cycles most of the time on the Cortex-A8 where it would have taken 3 on the ARM9. Some ASM code that was written for ARM9 might not schedule well on Cortex-A8 and might have new shift and address generation stalls.
 
Last edited by a moderator:
EvilDragon said:
I know a lot of people do say "The more realistic games are, the better they are" - which is something I completely disagree.

There's a charm to the non-realism of games that IMHO helps with the whole idea of "getting away from it all". We don't play games to relive reality - we can do that actually IN reality and save money on batteries and power supplies. A large part of many games is learning the "rules" of the universe they work in, so that you can maximise your chances of winning. Those rules may be getting more and more sophisitcated, but to actually approach relity too closely is dangerous IMHO as it risks losing that simple element of games - the "learning how it works" stage (and let's face it, some of the better games you're learning how it works right up to the last level).

I think overall people confuse "sophistication" with "realism". It usually isn't more realistic games we want per se, but more sophisticated ones. Or just prettier ones ;) Ones where the universe of the game is enticing enough to draw us in. Many old games had that "otherworldliness" that provided escapism, and personally the newer games tha tI enjoy also have a large dash of escapist unreality - even if they are graphically and audibly more complex than their earlier counterparts.

Laurent said:
Something bothers me in this discussion: there is too much fanboy'ism and also too much anti-Pandora talk. As always the truth probably lies in the middle :)

This is the only part of your post that puzzled me - it's quite common, after all, for any debate to have polarised positions.

MWeston said:
Having the second nub act like a mouse is just so fantastic. Being able to move that cross hair anywhere I want at any speed I want when being attacked from everywhere was wonderful. That's what Pandora is all about and not necessarily trying to show off the fastest clock speed although unofficially, it's doing quite well there too.

If the nubs work THAT well I guess I'll be chuffed - I've been finding using gamepad joysticks a little lkmiting in terms of speed and accuracy compared to a keyboard and mouse, but if Pandora does betetr that'd be nice.

One concern though - if YOU give "unofficial" comments on the Pandora's performance, where on Earth do we go for OFFICIAL information? :lol:
 
Last edited by a moderator:
Monk said:
Laurent said:
Something bothers me in this discussion: there is too much fanboy'ism and also too much anti-Pandora talk. As always the truth probably lies in the middle :)

This is the only part of your post that puzzled me - it's quite common, after all, for any debate to have polarised positions.
Well I think I was in the middle, and I would hope that most people could reach that position, though it'd be less fun :lol:
 
Last edited by a moderator:
This is a bit off track but why did the OP team not choose the 3550 instead of the 3530 (if that's correct) originally? Was it not available or added complexities to the design?

Thanks.

Edit: Or of course cost?
 
shunun said:
Obviously everyone has different homour, besides, pandora and DNF do have similarities.. They have both taken a long time, DNF is not vapourware, and nor is the pandora. So where's your rebut now?? next time just shut up.

Actually DNF is canceled and Open Pandora is in manufacturing, so at the moment the comparison isn't really valid.

I think the original poster isn't taking into consideration the target audience of the Open Pandora. OP will not be a mainstream device. It will not have tons of commercial games. It will probably not sell more than 50,000 units or so. None of the large commercial developers will develop for it. It does have the ability to run some retro games via emulators, but it is not meant to be a "replace every other game system out there with this one machine that fits in your pocket" type of thing. It will not replace your iPhone/smartphone, PSP, console system, laptop, desktop, GPS, or any of a host of other electronic gadgets out there.

It IS a small linux machine that fits in your pocket, has a nice keyboard for the form factor, and has great game controls. It is for indie and home-brew game development. It is for tinkering. It is an intended replacement/competition for the GP32X/Wiz/Dingoo.

Cheers,
Michael
 
Last edited by a moderator:
Back
Top