What Is The Clock Frequency Of The Pandora ?

[icurafu]

This is very interesting, thanks squidge. Imagine what you'll get when you start using the SGX.

I'd love to see some more results as you port other application. Even some youtube videos would be awesome.

I'd also love to see what performance (and power consumption) you get if you underclock to 100MHz.

You could also ask PowerVR for the source for the OpenGL ES 2.0 demos on the PowerVR webpage.

http://www.imgtec.com/Downloads/DocumentDo...Demos/Index.asp

 

[Joshwaan2k]

So Pandora's cpu is running at 500MHZ and it's getting 138FPS in a NeoGeo game. This is with sound disabled ?

That's pretty damn good mate. Squidge I'm very much looking forward to your Snes emu mate.

I can see it now, pretty much all Snes games running 100% with 2xSAI filter .

Can't wait for this handheld

 

[StealthBagel]

GBA has a good order of magnitude more CPU power than anything else that's emulated on GP2X with the exception of PS1 and Jaguar, both of which are probably not that far off due to being gimped in the fast memory department.

Or Amiga, or any PC system emulated in DOSbox, and higher-end arcade hardware (NeoGeo, high-clocked 68K/Z80 pairs in many arcade platforms) should about break even with it, or exceed it.... I am not referring to raw capability alone, but also the amount of host CPU cycles used trying to emulate the hardware in question.... also if I'm not mistaken, the GBA's ~16 MHz ARM7TDMI is badly bottlenecked in the manner it accesses the bus, reducing its effectiveness quite substantially.... The GBA's video hardware is impressive, but its CPU is by no means a regular powerhouse.

And.. I hate to burst your bubble, but you're not going to see anyone using hardware acceleration to emulate SNES or GBA's video subsystems while retaining any kind of accuracy.

If you say so. I'll take your word for it since you're an emulator programmer and I'm not, but I find it pretty hard to believe hardware blitting of the different graphical layers and implementation of transparency could not be accelerated in such a fashion, at the very least. It might be less accurate than drawing everything completely in software.... however such a thing can be switched on or off for the bigger sticklers for maximum accuracy. Most people here seem concerned with speed over 100% accuracy, since this is a handheld we're talking about, and portable emulators are usually a great deal less accurate than their PC counterparts.

 

[Exophase]

Or Amiga,

How about no, that's a 7.16MHz 68k for you, less than 1/10th the power of the CPU in the GBA.

or any PC system emulated in DOSbox

Any PC system, lozlz. I guarantee you that not a single person is emulating a PC system with a CPU anywhere near GBA's at an acceptable framerate, that'd have to be like a 33MHz 486 or something. Most anyone using DOSBox on GP2X (if anyone does) is emulating old 8088's that are tons weaker than GBA's CPU.

, and higher-end arcade hardware (NeoGeo, high-clocked 68K/Z80 pairs in many arcade platforms) should about break even with it, or exceed it....

Oh man, you really have no concept of how these old CPUs performed do you. Okay, let me break it down for you, both of them get in the range of 0.08 to 0.12 or so MIPS per MHz, while GBA's ARM7TDMI will get closer to 0.8 or more MIPS per MHz. Maybe you don't believe me, but I've seen GBA games pull exactly this, and Lordus has given me numbers that Genesis games pull (Genesis has that exact combination) because we've both been heavily profiling CPUs. 68K + Z80, please, don't joke with me. If that were even close to true then none of the emulators for those platforms would run well at all because they're all using interpreters.

Oh, and the ARM ISA is more sophisticated than 68k's, don't even mention z80's. That's just ancient by its standards.

I am not referring to raw capability alone, but also the amount of host CPU cycles used trying to emulate the hardware in question....

Let's see, opcodes that are generally 8 or 16bit in length vs 32bit instructions with dense encoding, hmmm, having to handle conditional execution on every opcode, a barrel shifter on every ALU op2, being able to modify the program counter with almost any instruction, those things must all be easy to emulate. That or you just don't know the first thing about emulation, you probably don't even know the first thing about these CPUs to begin with :rolleyes:

Granted, GBA games often spend a lot of time in Thumb mode which is simple, but games can and DO spend a lot of cycles in ARM. Go play Asterix 3D and tell me that your Neo Geo can do that.

also if I'm not mistaken, the GBA's ~16 MHz ARM7TDMI is badly bottlenecked in the manner it accesses the bus, reducing its effectiveness quite substantially....

You're wrong, again. That's only true if the game isn't programmed to strongly utilize the 32KB of zero wait-state IWRAM on the GBA, which almost every game ever released for GBA does, some much more than others. Ignore the MIPS figures dwelch reported in his benchmark, he was running the ARM code in ROM which is not a realistic setup.

The GBA's video hardware is impressive, but its CPU is by no means a regular powerhouse.

Look, you don't know what you're talking about. Go look at the library of emulators available for GBA, you will be absolutely shocked. Comparing GBA's CPU to SNES's is a disgrace, it's in a totally different league. There's an SNES emulator for GBA that actually runs some games at good speed.

QUOTE
If you say so. I'll take your word for it since you're an emulator programmer and I'm not, but I find it pretty hard to believe hardware blitting of the different graphical layers and implementation of transparency could not be accelerated in such a fashion, at the very least.


Funny, that doesn't sound like taking my word for it at all

QUOTE
It might be less accurate than drawing everything completely in software.... however such a thing can be switched on or off for the bigger sticklers for maximum accuracy. Most people here seem concerned with speed over 100% accuracy, since this is a handheld we're talking about, and portable emulators are usually a great deal less accurate than their PC counterparts.


Actually, since SquidgeSNES gets great speed if you turn off transparencies and raster effects and whatever and yet so many people still complain (because I guess they want an accurate gaming experience, unthinkable) I would say you're way off the mark.

And there you go again saying portable emulators are less accurate than PC ones, which is just something you've made up. The only time this might apply is for DS emulators using hardware acceleration (which is much more feasible than doing it on a 3D accelerator because lo and behold, it has very traditional styled 2D acceleration)

Most emulators for handhelds are PORTS you know.

EDIT: Come to think of it.. and I hate to pull a craigix on this one, but some of the stuff you're saying sounds a LOT like what Epicenter went on about (XScale being 50% speed and not a "true ARM", handheld emulators being inherently inaccurate, insisting that hardware acceleration be used in emulating 2D consoles).. if I find out it's you again I'm gonna be really peeved

 

[StealthBagel]

When did I say the SNES was more powerful than the GBA? That's insane. You've got a knack for putting words in my mouth.

There's a line between wanting 'accuracy' like the MAME crowd does which can't even be witnessed by anyone not using a debugger, and accuracy like 'not turning off critical PPU features'.
I mentioned already how most portable MD emulators have mediocre YM2612 emulation compared to their PC counterparts because that is a very CPU intensive thing to emulate.... that is just one
example but there's a lot more than that. Portable emulators usually take shortcuts or disable feautres the PC version could support but the hadheld in question would not. If they were ports that is....
which I am well aware of but the ones written for very limited platforms usually aren't.

Last I checked you guys decided Mister B or whatever his name is was this epicenter you hate so much. I don't know if this guy raped your dog or something but maybe it's time to move on if you and one of the big players in development for this handheld are calling out random people who might otherwise be customers due to suspicion they're some forum member they don't like.

I guess you're going to have to close the forum off from registration and start banning people left and right because we definitely can't have people who disagree with exophase here. Once you work on a psx emulator you're automatically god himself.

 

[Exophase]

When did I say the SNES was more powerful than the GBA? That's insane. You've got a knack for putting words in my mouth.

"like SNES and GBA which are graphically very intensive without packing a lot of CPU power...."

Directly implied that they were of similar capability.

There's a line between wanting 'accuracy' like the MAME crowd does which can't even be witnessed by anyone not using a debugger, and accuracy like 'not turning off critical PPU features'.

Which is exactly what would be necessary in order to emulate w/hardware acceleration.

I mentioned already how most portable MD emulators have mediocre YM2612 emulation compared to their PC counterparts because that is a very CPU intensive thing to emulate.... that is just one
example but there's a lot more than that.

When are you going to stop talking about YM2612 emulation being inaccurate when it's already become obvious you haven't played picodrive in a bazillion years? You're the only person in the entire forums saying this, it's based on nothing whatsoever, how about giving it a rest.

Portable emulators usually take shortcuts or disable feautres the PC version could support but the hadheld in question would not. If they were ports that is.... which I am well aware of but the ones written for very limited platforms usually aren't.

No, most of them don't when it comes to accuracy, and if they DO (like SquidgeSNES) they let you turn it back on so um.. yeah, I think for the most part emulators don't even have that many options that affect accuracy anyway.

Most emulators for GP2X and PSP are ports, just look at the libraries yourself, for crying out loud.

Last I checked you guys decided Mister B or whatever his name is was this epicenter you hate so much.

No, that was just craigix who only thought that because he had fits with Pandora, now you have lots of particularly specific similarities. And you joined at just the right time too, fascinating!

I don't know if this guy raped your dog or something but maybe it's time to move on if you and one of the big players in development for this handheld are calling out random people who might otherwise be customers due to suspicion they're some forum member they don't like.

Haha what, are you threatening me with not buying Pandora? I'm not your customer, jackass, I have nothing to do with Pandora. I'm just saying you put forth a lot of the same asinine arguments he has. Maybe you're someone else, but you sure share a lot in common which doesn't go over that well for you!

I guess you're going to have to close the forum off from registration and start banning people left and right because we definitely can't have people who disagree with exophase here. Once you work on a psx emulator you're automatically god himself.

Hey, I didn't work on a PSX emulator, I coded a GBA emulator. Which is why I know so much more about the platform than you do.

I swear, the stupid things that come out of your mouth, could you make any more of a fool of yourself?

 

[HIVexploder]

You guys want to knock it off already? we're horribly off topic and now I don't know how many mega cycles the pandora has on its flux capacitor.

 

[StealthBagel]

The YM2612 thing... well everyone knows DaveC is famous for going on tirades about PC-Engine/turbografx16 audio emulation accuracy so I'm not the only one, but this is clearly not a common trait for people who use a lot of emulators to insist on near-100% hardware accurate audio. I'm picky, that's how I am, and it's how I'll always be about some things.... oh well. If no one else is gonna say it, I'll say it. That's why this is a forum...... where would we be if no one siad anything but the same shit over and over.

I'm not going to pick through all your arguments because I really don't have the time. You've made some good points and some really bad and irrelevant ones, but you're clearly just as set in stone and immovable on a lot of points as I am, but I will admit when I am wrong. I don't think you're capable of that, though. If you are I never witnessed it....

so yeah, think whatever you want about what you "think" I meant by any argument or if you think I'm the rebirth of some troll or whatever gets you off, man. I'll be busy with other things.

 

[Exophase]

The YM2612 thing... well everyone knows DaveC is famous for going on tirades about PC-Engine/turbografx16 audio emulation accuracy so I'm not the only one,

Did you realize that PC-Engine doesn't have a YM2612??? Actually DaveC has said that gp2xengine has excellent audio quality but it's a slow emulator (NOT because of the sound, because it's just a slow emulator overall), these other emulators don't have deliberate accuracy compromises, they're just done by people who either didn't feel like emulating everything or didn't know how to! I've seen a bunch of half done PC-Engine emulators for PC that have terrible audio too, does it have anything to do with portables.. no!

but this is clearly not a common trait for people who use a lot of emulators to insist on near-100% hardware accurate audio. I'm picky, that's how I am, and it's how I'll always be about some things.... oh well. If no one else is gonna say it, I'll say it. That's why this is a forum...... where would we be if no one siad anything but the same shit over and over.

You're not picky, you're just ignorant, I don't think you've even used any emulators on GP2X in a long, long time... It's not about being picky about problems, that's DaveC, and as much as we all get annoyed by his persistence from time to time I've got to say that it's very easy to understand where he's coming from. You, on the other hand, just make crap up.

I'm not going to pick through all your arguments because I really don't have the time. You've made some good points and some really bad and irrelevant ones, but you're clearly just as set in stone and immovable on a lot of points as I am, but I will admit when I am wrong. I don't think you're capable of that, though. If you are I never witnessed it....

I always admit when I'm wrong.

Just unlike you I make more of an effort to not talk about things I'm clueless about.

Unless you're willing to actually show how I've been wrong or what I've said that was such a bad or irrelevant point, keep your mouth shut.

so yeah, think whatever you want about what you "think" I meant by any argument or if you think I'm the rebirth of some troll or whatever gets you off, man. I'll be busy with other things.

Let's get something straight, I actually don't think Epicenter is a troll at all. He's just an annoying person who keeps arguing about things he has no business arguing about.

 

[StealthBagel]

I've got a little more time to burn so let me clear a couple things up (in a civil tone, don't get the impression I'm in some kind of a rage here...... this is definitely a misconception).

I know PC-Engine has no YM2612, it has an embedded proprietary audio generator in the same IC as the CPU.... DaveC originally was commenting on another emulator, I think called Hu6280 or similar and how he was disappointed the original author left the forum and didn't perfect the audio. I didn't say the PC version had better audio and that being on a handheld made the emulation inaccurate. That is sometimes the case, but in this one it isn't. Emulation for a lot of systems is just underdeveloped, like PC-Engine, Lynx, Jaguar, Virtual Boy, Neo-Geo Pocket and some others. I wasn't trying to make an irrational blanket statement, you ran two of my points together to say something I never intended.

Yeah, I haven't used most Gp2x emus in a long time.... I got bored of the machine fast due to how it was built so I don't keep up with all its software. My remarks were suggestions for better software and hardware in the future based on my past observations, not criticisms of gp2x software in the state it might exist today. It's no crime to make potentially-helpful suggestions in hopes the pandora will combine better hardware with better software to take advantage of it; I think that's pretty common sentiment.

 

[atomicthumbs]

hay guys can we put the pandora clock singal in a wire so we can ues it liek a radio

 

[icurafu]

Are you high?

 

[Squidge]

Could you post the GP2X ones too please?

Sure, I'll post them tonight.

So Pandora's cpu is running at 500MHZ and it's getting 138FPS in a NeoGeo game. This is with sound disabled ?

Sound isn't output, but it's certainly mixed.

 

[Clad]

Hey, I didn't work on a PSX emulator, I coded a GBA emulator.

Don't worry, that still qualifies you for godhood.

Anyway, it's no the first time you argue with an user that way, and I don't quite understand you.

Usually, and this time is no exception, you make your point pretty early in the debate, and everyone reading it would agree that you are right. In this particular example, when you say "No, the Amiga is a 68000@7mhz and GP2X dosbox only emulate <186 CPUs at full speed", well, the argument is kind of over, no matter whether you did or did not code a GBA emulator, your point is valid enough by itself.

What I don't understand is why, once it's clear for everyone that you are right, you keep involving yourself in what then become a quite sterile conversation. It's only trolling for the sake of trolling, and won't make anyone learn anything new, which is kind of the point of a forum (sharing knowledge).

I don't think anyone here would challenge your technical knowledge, as long as you say something is wrong and explain why with sensitive arguments, we'll believe you. Please stop acting like you need to prove something : you don't.

 

[fusion_power]

Here's my original release of fba2x ported to the Pandora:
=======================================

/ # ./fba2x ./donpachi.zip
Attempt to initialise 'DonPachi (ver. 1.01 1995/05/11, U.S.A)'
1 attempt to open './donpachi.zip'
seems ok, moving to check the other stuff
2 prgu.u29 ... 512Kb program (OK)
1 atdp.u44 ... 2048Kb graphics (OK)
1 atdp.u45 ... 2048Kb graphics (OK)
1 atdp.u54 ... 1024Kb graphics (OK)
1 atdp.u57 ... 1024Kb graphics (OK)
1 text.u58 ... 256Kb graphics (OK)
atdp.u32 ... 1024Kb sound (OK)
2 atdp.u33 ... 2048Kb sound (OK)
BurnDrvInit exiting with return value 0
Lets go!
239 630 225 231 535 377 199 206 199 200 = 304 fps avg
188 227 188 198 182 178 182 188 200 179 = 191 fps avg
176 184 176 178 199 224 234 435 455 191 = 245 fps avg
207 196 197 187 237 207 187 173 169 187 = 194 fps avg
176 172 175 162 183 208 245 233 236 229 = 201 fps avg
575 310 202 201 197 196 197 208 202 203 = 249 fps avg
165 192 201 191 193 185 221 237 420 483 = 248 fps avg
190 207 197 196 187 216 199 204 185 170 = 195 fps avg
177 198 199 180 174 184 178 167 192 210 = 185 fps avg
247 283 605 195 209 197 196 195 212 223 = 256 fps avg
191 183 206 245 274 623 202 210 196 196 = 252 fps avg
195 203 199 207 209 219 239 394 510 188 = 256 fps avg
209 196 197 188 215 199 205 184 170 177 = 194 fps avg

Notes: It's still not optimised yet for the omap, only for gp2x. fps counts taken each second. After ten seconds, average fps is outputted for the last 10 seconds.

Opinions?

MHz-FPS comparisons between Pandora/GP2X?
DonPachi is a good example, because I can play this on GP2X and it runs Full Speed at 200MHz (always between 50-60 FPS) Could you clock your Pandora to 200MHz and see, what happens with the Game?

And.. I hate to burst your bubble, but you're not going to see anyone using hardware acceleration to emulate SNES or GBA's video subsystems while retaining any kind of accuracy.

Hm, thats new for me to. So the Devs of Emulators can't use Hardware Features of the Pandora to support faster emulation? Is this only for some Emulators or an overall fact?

 

[Orkie]

Granted, GBA games often spend a lot of time in Thumb mode which is simple, but games can and DO spend a lot of cycles in ARM. Go play Asterix 3D and tell me that your Neo Geo can do that.

You couldn't do that even if hypothetically the CPU were powerful enough, since it can't access VRAM, there are no bitmapped modes and I *think* it can only read graphics from ROM, not RAM (so no rendering things with the CPU) .

 

[craigix]

Opinions?


MHz-FPS comparisons between Pandora/GP2X?

And.. I hate to burst your bubble, but you're not going to see anyone using hardware acceleration to emulate SNES or GBA's video subsystems while retaining any kind of accuracy.

Hm, thats new for me to. So the Devs of Emulators can't use Hardware Features of the Pandora to support faster emulation? Is this only for some Emulators or an overall fact?


You can speed emulators up a little with hardware blitting, blitting to the screen/buffer but you can't speed them up like Epicentre used to claim, ie the 3D/2D hardware in the pandora is nothing like the custom hardware in the SNES or GBA and would be no use at all.

To see how much hardware blitting speeds up emulators see the software sdl and hardware sdl version of the neogeo emu, I think the hardware one is around 15fps faster, so it's a decent gain but nothing like the gain you would get if you could do all the graphical effects directly on the graphics chip.

Where the Pandora hardware should become interesting is in emulating 3D systems.

 

[HIVexploder]

well I guess it does't matter too much either way because even if the devs just want to use software graphics
for everything with hardware blitting it should still be pretty damn fast if the numbers that squidge posted are
similar to the real performance when we get units. but was that in-game or just sitting at the title screen or something? the difference isn't always so huge but there is a difference. since games sometimes slow down so sporadically when there is a lot more goin gon onscreen.

 

[atomicthumbs]

Are you high?

no, just silly (+ a certified flamewar defusal technician)

 

[pepone]

To see how much hardware blitting speeds up emulators see the software sdl and hardware sdl version of the neogeo emu, I think the hardware one is around 15fps faster, so it's a decent gain but nothing like the gain you would get if you could do all the graphical effects directly on the graphics chip.

IIRC, you're right. gngeo gain 10/15fps with the hardware sdl version.
On the other hand, the neogeo video system is actually quite simple, and using the hardware capability to do the sprite rendering might be possible (I think the neogeo emulator for psp do it for example).
But seeing donpachi running so well on pandora, putting a lot of work in that kind of rendering engine may not be useful.