Die Shrunk Omap3: Lower Power, Higher Speeds, Better Graphics

[joa]

Until we hear from Craigix I'd say it's way too late. I think they're being a bit optimistic with the increase in performance. Aren't they always?

My worry is that this appears in future pandora's and emulators are optimised for the faster chips. thoughts?

Simple. Lock out the increased speed, and then you gain an even bigger increase in battery life. Sony did something similar with the PSP, releasing higher spec PSP's but were still locked down to old specs for games.


That would make a lot of sense.

 

[benjiro]

There's no need to react the way you did.

I disagree. Even if you were joking, this is thread number 9267352 of this nature so it's not funny anymore, just annoying.


To be fair, this is only the second thread with this topic. If you want to complain, do so with all the 15, or 20 keyboard topics, and whatever else topic's.

Because off the smaller 45nm process, they are able to increase the default cpu speed to 720Mhz, and the 75% better graphic performance seems to mean that the GPU speed has increased from 110Mhz to +- 190Mhz. Whats unclear, if the chip delivers a 25% decrease in power combined with a increase in cpu/gpu speed ( PR releases can be a dangerous thing ).

But, if its just Engineering samples, yes, then we have no us for the moment. On the other hand, if its already going into production ( what was unclear ), then the OP's question is valid. Just like mine was, but some people seem to jump on it like rabbit dogs, looking for a kill... :angry:

This is almost the exact same question / problem like with the 128/256MB version. At first it was not possible for the 256MB version, then it became possible and reality. So, ofcourse people have the right to ask. Asking never hurt anybody. And Craigex said it are ES ( Engineer samples ), so no luck seeing them any time soon ( unless it takes a year to finish the Pandora ).

 

[WizardStan]

Simple. Lock out the increased speed, and then you gain an even bigger increase in battery life. Sony did something similar with the PSP, releasing higher spec PSP's but were still locked down to old specs for games.
That would make a lot of sense.
It works for Sony because Sony is the defacto authority in what the PSP can and cannot do. If Sony says "thou shalt not write anything that exceeds 220 mhz (or whatever it actually is)" then the game producers need to stick with it, even if the CPU is actually capable of much more.
The Pandora is fully open by design and necessity. If a TI releases an OMAP3 almost identical to what's going in the Pandora except it boasts an extra 200 mhz, you will divide the community: someone will write a something that takes advantage of that extra 200 mhz (or extra memory, or extra whatever that may go into it) and the result will be "those with Pandoras bought after X can run this, those bought before cannot." At which point you may as well just upgrade everything and release a Pandora 2.

 

[Loonie]

I can out-shorten that:

yes.

1

 

[Monk]

Imagine we would wait for the first IBM PC, delayed due to problems with the 256kb RAM bank. Intel would have announced the 8088 but our design sticks with 8086. If we waited for 8088, the 80186 would be around and so on.

It's amazing to me now that the first IBM PC (not the Jnr) came with just 16KB of RAM, with room on the motherboard for a total of 64KB. Astonishing.

 

[Tor]

Soon we'll get used to getting 16GB RAM on our computers, a million times more memory than the first IBM PC. And my first comp had only 2KB.. it's not even that long ago. Oh well.

 

[lulzfish]

And as they say, software has expanded to fill the new hardware.
They're saying Windows 7 won't be as demanding as Vista, but that's not saying much. I'll probably be staying with Ubuntu or Debian for my PCs for the foreseeable future. Since the whole DRM thing [Which basically appeared in Vista and will be getting worse in W7], I don't trust Microsoft with my computer anymore.

 

[fischju2000]

Soon we'll get used to getting 16GB RAM on our computers, a million times more memory than the first IBM PC. And my first comp had only 2KB.. it's not even that long ago. Oh well.

(You can already get 4gb sticks and mobos with 6 RAM slots)


It took months to get the PCB right and work out manufacturing defects, switching at this point would be disastrous for the project..

 

[Tinnus]

My point for the future: of course if at some point the 36xx equivalent becomes cheaper than the 35xx one in the Pandora (say if it's discontinued or any crazy reason), then it would just be plain stupid to use the old version.

edit: Of course provided they're pin-compatible or the difference in price compensates a board redesign.

 

[Eternalodyssey]

I am very interested in seeing this used in an HTPC. I have been contemplating purchasing a Beagleboard and seeing just how far I can implement it as an extremely low power HTPC. Heck its fanless and the only noise I would hear is the laptop DVD drive spinning when in use. I wonder how long it might be before we see systems on a chip like this completely integrated into home television sets for more than simple weather updates and email capabilities. That would be sweet if I could just fire up a PiP feature to get more information on something that was broadcast on the television. Its the future I tell ya!

 

[Exophase]

It works for Sony because Sony is the defacto authority in what the PSP can and cannot do. If Sony says "thou shalt not write anything that exceeds 220 mhz (or whatever it actually is)" then the game producers need to stick with it, even if the CPU is actually capable of much more.
The Pandora is fully open by design and necessity. If a TI releases an OMAP3 almost identical to what's going in the Pandora except it boasts an extra 200 mhz, you will divide the community: someone will write a something that takes advantage of that extra 200 mhz (or extra memory, or extra whatever that may go into it) and the result will be "those with Pandoras bought after X can run this, those bought before cannot." At which point you may as well just upgrade everything and release a Pandora 2.

Look at GP2X - some people can overclock it over 300MHz, while others are stuck at 240MHz. No one writes software that needs 300MHz, but a lot of it can benefit from it. Really it's hard to make something require a certain clock speed - at worst, it'll have slowndown if it's not clocked that high. I doubt anyone would set a minimum requirement in the higher MHz range if they can avoid it, but if the app is going to be too slow anyway then the extra MHz sure doesn't hurt.

 

[WizardStan]

Look at GP2X - some people can overclock it over 300MHz, while others are stuck at 240MHz. No one writes software that needs 300MHz, but a lot of it can benefit from it. Really it's hard to make something require a certain clock speed - at worst, it'll have slowndown if it's not clocked that high. I doubt anyone would set a minimum requirement in the higher MHz range if they can avoid it, but if the app is going to be too slow anyway then the extra MHz sure doesn't hurt.

Right, but that's overclocking: the official rating is still only whatever it is; anyone who gets something beyond that is totally bonus.
If the OMAP in the Pandora is changed in such a way that the new boards are officially superior to the old ones, that's when you risk contention. I'm not saying that that's the way it will happen, just that it's more likely than if the only difference is who is lucky enough to be able to overclock their system more.

 

[Phawx]

I think you should treat the Pandora like a PC. Upgrades happen. In the end its going to be like a computer. Well at least I don't want to treat the Pandora like a console. I want to treat it like a handheld PC.

 

[cb88]

people are worried that they are gonna buy a device and 3 months down the road a new one is gonna be faster for the same price...

WHAT ELSE IS NEW!!!

so just get over it if they can make a pandora for the same price with a faster processor then it is just natural progression and won't decrease the number of titles available for the original pandora but might actually increase them since if the new HW keeps with the times it will remain more relevant in the market bringing more developers and since the hardware is mostly the same with only a chip size upgrade you have retained backward compatibility which is what is really important

you can't force people to make good games for the pandora but you can make an attractive platform that people will want to dev for and if a game is too slow to run on the Pandora but runs on later versions what have you lost? nothing because the game wouldn't run anyway

 

[MiniSinisterMinister]

If you have a fixed handheld, there will be more of an incentive for homebrew developers to code optimised emulators for it and spend more time developing for the system in general. Look at the effort the guy behind the N64 emu for the PSP is still putting into his emulator and the attention this earns him. If the PSP was an "evolving" handheld, people would just shrug their shoulders and think that a simple port of an existing N64 emu will run fast enough sooner or later anyway. I think this would kill the enthusiasm for programming for such a handheld and also that of the audience. And what is more, this only gives you the illusion of faster progress: you would have to buy another unit every 2 to 3 years only to run software that wastes 50% or so of the hardware resources.

 

[lulzfish]

It also seems like it would be a lot cheaper for the dev team to produce one design for as long as it sells, so they can make money on the volume without needing to redesign and re-prototype everything.

Also, that's really the point of consoles, to have a not-cutting-edge hardware / OS platform that's highly standardized.

 

[Zeno Arrow]

It also seems like it would be a lot cheaper for the dev team to produce one design for as long as it sells, so they can make money on the volume without needing to redesign and re-prototype everything.

The point is that use of this new set of OMAP3 chips is unlikely to require a redesign of the Pandora PCB. If the OpenPandora team can offer improved battery life on later versions without increasing the consumer purchasing price why shouldn't they go for it?

The only reason I can think of is splitting the community but splitting the community would only happen if software was written with an overclocked processor in mind, which is not going to be the case for the most part.

Think about it like this, how much 'designed for F200 only' software is there compared to software designed for all GP2X models?

 

[cb88]

@MiniSinisterMinister the PSP is a changing platform some of them have twice as much ram and so forth it is not a fixed platform

also from what i understand the n64 emu isn't even possible on the lower end version but don't quote me on that

also this is not windows ... according to kernel developers over the past few years only a slight decrease in proformace in Linux has occured at the kernel level the rest is up to the distrobutions so if you choose to run bloatbuntu well you are democratically shooting yourself in the foot

you can't stop people from writing crappy code get over it... thankfully the pandora seems to have attracted a good number of developers that care to take pains with thier code

Oh btw i have an 11 year old mobo here with dual PIIs that can still hang in there and be usable with Linux on it so don't tell me that just because computers are getting faster the open source arena is getting slower as a whole although I do agree that some areas are

 

[benjiro]

@MiniSinisterMinister the PSP is a changing platform some of them have twice as much ram and so forth it is not a fixed platform

also from what i understand the n64 emu isn't even possible on the lower end version but don't quote me on that

True, the PSP is not a fixed platform. But, none off the games for the PSP are designed with 64MB memory in mind. Just some games run better, then others ( less < 30fps = frame drops ) because off the 32MB memory.

We can not exactly compare the PSP as a marvel experience in the upgrades. They lower the manufacturing process. Good. But then they also decrease the battery size, in a non correct correlation with the gain in power saving. Resulting in a drop in playtime. But thats a discussion for a other topic

From that posts/discussion about the n64, lets say that reality set in now ( compared to a few months ago ), and while the n64 is possible, do not expect to see it running on the cpu @ 300mhz ( what may be possible with a lot of other emulators ).

Personally, given the change, IF there is a 45nm OMAP3 3630/or3640 ( 1Ghz, yea baby! ), version down the line, i'll gladly buy it.

Why:
- Power Saving:
If the claims are correct that it has a 25% lower power drain, with a higher clocked gpu, that means the power gain is even more. But, lets assume its just 25%. Thats maybe a hour more battery juice ( note: took in effect that the omap is not the only power using component ).

- Speed:
Default speed increase of 120Mhz, on the 3630, and 400Mhz on the 3640... Not bad at all.
But a 75% speed increase on the GPU! No offense, but thats not a small mater anymore.

From the 10 MPoly/sec to 17.5 MPoly/sec... Thats a major upgrade on the GPU. That open a even large door for any commercial game developer. Correct me if i'm wrong?

- Overclocking:
The fact that the OMAP3640, is 1Ghz, means that there is a high chance that the OMAP 3630 can do 1Ghz, or higher OC'ed.

- Possible other tweaks?
Now, i'm going to jump on a limb here, and assume that we see the same thing like with normal CPU upgrades. With new revisions, the design gets tweaked, you see this the most clearly with changes in the manufacturing process. Like going from a 65nm to 45nm Intel. Clock both on the same bus, cpu etc speed, and there are a few % "free" speed gain. Who's to say the same thing will not happen here.

- Future proof?
How long is TI going to support the 35xx serie? Lets be honest, the 65nm version is going to cost them a lot more money to manufacture. Anybody who has been in the PC business knows that any CPU manufacture will gladly drop the older manufacture process as soon as possible.

There is no "new" manufacturing process thats going to be ready for mass adaptation and production the first year and half ( 32nm ), and thats Intel, who in general are a year in advance off most other manufactures. So, the lifespan for the 45nm is at minimum 2.5 to 3 years. Can we say the same for the 65nm process? Intel -> 45nm, Amd -> 45nm, Memory producers -> 45nm...

- Other changes?
This is pure speculation, but there has been no mention off the memory size. The question is, are they also going to die shrink the memory? In other words, they can place more memory in the same package size then before. Pure speculation, but something that i expect we will see.


The problem is, the games that "may" use the advanced CPU/GPU speed, that normally are not possible ( or to limited ) with the current. Going back to the PSP, we all know that the PSP was locked at 200mhz, but Sony did unlock the 300Mhz range ( homebrew makers have been using that a along, but thats a other mater ). In other words, even Sony's PSP has not been so "fixed". Hell, did the X360 also not get several upgrades?

So, given the choice in the future, i will gladly buy a second Pandora 1.5 with a 3630 or3640 in it.

 

[t3hmadhatt3r]

Would it be nice?For sure...Would it be inconvenient for the developers and for us?Probly...It would take 4-6 months longer and would most likely cost more.In the future it would be great to use the OMAP4 in a newer pandora.

QUOTE

Power just slight of a Xbox-360 in the palm of your hand


Yes!It would be nice but, right now its just not going to happen.The pandora should and will finish the first one and get their profits before starting development for the pandora 2.0 .It is something to look forward to and I'm am excited to see how the pandora will progress.I think it's an interesting post...Cheers!