Pandora Not Powerful Enough?

[Exophase]

OMAP3550 doesn't exist. Maybe you're referring to 3630 or 3640.

When OP first chose to use a TI chip the 3530 wasn't even officially available. OMAP3430 (which as far as I'm aware is identical) was announced, but you couldn't get any yet. Naturally something better than 3530 wouldn't have been available. I don't even know if it's possible to order anything better.

 

[Awakening]

I wonder if the Pandora can outsell the GP2X (which had sold over 60'000 units 31 Aug 2008). The GP2X was primary sold as an English learning tool for Koreans so it might be hard to beat that number.

 

[Monk]

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:

It might be a nice goal, but it's unlikely. We seem on the whole to avoid the middle of the road. That way, at least we avoid getting run over?



I'll get me coat...

 

[Dead1nside]

OMAP3550 doesn't exist. Maybe you're referring to 3630 or 3640.

When OP first chose to use a TI chip the 3530 wasn't even officially available. OMAP3430 (which as far as I'm aware is identical) was announced, but you couldn't get any yet. Naturally something better than 3530 wouldn't have been available. I don't even know if it's possible to order anything better.

Thank you. I did mean 3530 versus 3430.

 

[Phawx]

I thought 3430 didn't have the DSP?

Also, did anyone see the Cortex A9 clocked at 500Mhz keeping up with an N270 Atom clocked at 1.6Ghz? It's on Engadget.

http://www.engadget.com/2010/01/05/arm-demos-the-cortex-a9s-web-browsing-skills-on-video/

 

[Exophase]

Pretty sure OMAP3430 has the DSP, as part of the IVA2+. 3430 is only intended for high volume wireless devices, ones like Pandora don't qualify.

I would hardly say that the Cortex A9 was keeping up with the Atom, it looked like it was way behind on some sites to me. But I think web browsing is an awful CPU benchmark and it's kind of embarrassing that ARM keeps resorting to this. I can't listen to the audio right now so I don't know the exact variables. But no matter what, web browsing will always be limited by a lot of factors aside from just CPU, even if it's offline.

Really there's no way a 500MHz Cortex-A9 can be competitive with a 1.6GHz Atom at most things where CPU performance really makes a difference. I'm sure that the real point is just to show that it's competitive just with web browsing, but I have a good feeling that people will get from this that the Cortex-A9 performs as well in general.

Maybe next they'll show that the Cortex-A9 is as fast as the Atom at disk defragging.

 

[mali]

ARM had some web browsing comparisons for A8 as well. They timed page rendering, Atom and A8 were roughly equivalent clock for clock. I agree though, web browsing is a common usage scenario, but not a real benchmark.

 

[streak]

Pandora have enough horsepower to cut your lawn, clean your room and make a dinner for two with candles. If you want you could also start some prawn movies and "she" will not be protesting at all. Damn she's that good..

 

[Exophase]

ARM had some web browsing comparisons for A8 as well. They timed page rendering, Atom and A8 were roughly equivalent clock for clock. I agree though, web browsing is a common usage scenario, but not a real benchmark.

I actually expect that clock for clock A8 will outperform Atom, so long as it's not involving scalar floating point code. On paper the pipelines look fairly similar, with A8 getting a bit of a boost from having lower address generation stall and branch misprediction penalty and Atom getting bit of a boost from having lower load-use penalty. But in practice I think that having far fewer registers, two address arithmetic, and inability to do much without every ALU flag getting stepped on makes it a lot harder to schedule x86 code for in-order very effectively. x86 benefits heavily from out-of-order execution and register renaming which Atom just doesn't have.

Cortex-A9 should beat Atom per clock even more so, but not by a margin of over 3x, that's for sure.

 

[Phawx]

@Exophase

I'm just excited for the A9 to be showing up. The demo only showed off a dual core A9 running at 25% of its expected frequency. So seeing it, in my words "keep up", with the n270 is nice for me. Especially considering that the A9 is only supposed to take .25 watts per processor (at 2GHz?) compared to the atom at 2 watts or even the N450 with built in memory controller and IGP at 6 watts. I'm just really excited at the next generation of mobile stuff coming.

Hopefully Pandora 2 can rock (quasi)Quad-core A9 which they can scale up to now. Also, since the little DSP conversation a C67x chip would be nice, but I don't see it on any of TI's OMAP series.

I also thought over the Pandora 2 some more, and the screen actually couldn't get any bigger. It can get wider, but I don't know if that would really be *better*. So I'm all for Pandora 2 just getting a PCB enhancement. Maybe use the breakout cable (if possible) for HDMI (if OMAP4xxx has a secondary Digital output nm, I see the sheet and it does support HDMI out.)

 

[krosfyah]

I also think that now OMAP3 isn't amazing anymore, while I thought it was 2 years ago.

This is partly why I ended up cancelling my pre-order. I might be proven wrong but I just think that in 2010 the Pandora isn't nearly as attractive as it was in 08/09. I want to keep my options open for now in case something amazing is released later this year. Of course, if that doesn't happen then I may well end up with a batch 2 or 3 unit but for now the delays were enough to put some doubts in my mind.

 

[streak]

I also think that now OMAP3 isn't amazing anymore, while I thought it was 2 years ago.

Its not, but it making no difference. Only thing that matters is software. If it would be good, then im happy.
Pandora cannot run x360/ps3 emu, and should it bother you? PSP and DS are more underpowered, but people dont care as long as developers releases new shit every week on it.

ant btw. Pandora is very well prepared for HQ 3D games that could be compared to the best ones from iphone.

 

[ldesnogu]

I would hardly say that the Cortex A9 was keeping up with the Atom, it looked like it was way behind on some sites to me. But I think web browsing is an awful CPU benchmark and it's kind of embarrassing that ARM keeps resorting to this. I can't listen to the audio right now so I don't know the exact variables. But no matter what, web browsing will always be limited by a lot of factors aside from just CPU, even if it's offline.

Really there's no way a 500MHz Cortex-A9 can be competitive with a 1.6GHz Atom at most things where CPU performance really makes a difference. I'm sure that the real point is just to show that it's competitive just with web browsing, but I have a good feeling that people will get from this that the Cortex-A9 performs as well in general.

Maybe next they'll show that the Cortex-A9 is as fast as the Atom at disk defragging.

A9 is much faster than Atom at the same frequency and A9 will surely go well above 1GHz. It can also have 2 to 4 cores (and no, hyperthreading can't compete). You'll have to take my word as I can't provide figures

BTW I know of no 500 MHz A9 that will go into production (the one shown is a test chip); the lowest frequency I have seen advertised is the 720 MHz from lower-ends OMAP4.

Atom really is a piece of shit. Try to get access to one and run some benchmarks.

Also if ARM keeps on comparing ARM vs Atom on benchmarking stuff it's because that's what most people use. You know that most people don't care about emulation, don't you?
And you'd be surprised how in fact browsers are very dependent on CPU performance, especially in the presence of Flash and Javascript, but not only. Run oprofile on your PC and look at the results.

 

[ldesnogu]

I want to keep my options open for now in case something amazing is released later this year. Of course, if that doesn't happen then I may well end up with a batch 2 or 3 unit but for now the delays were enough to put some doubts in my mind.

I would be surprised if something similar to Pandora while being as polished as it seems to be and as powerful will be out this year.

EDIT: sorry, I keep on forgetting about multiquote

 

[MDave]

I also think that now OMAP3 isn't amazing anymore, while I thought it was 2 years ago.

Its not, but it making no difference. Only thing that matters is software. If it would be good, then im happy.
Pandora cannot run x360/ps3 emu, and should it bother you? PSP and DS are more underpowered, but people dont care as long as developers releases new shit every week on it.

I guess you don't know who Laurent is, but he definitely knows his stuff, and not a kid that wants a x360/ps3 emu. He's more then a developer from the hints dropped by certain individuals

ant btw. Pandora is very well prepared for HQ 3D games that could be compared to the best ones from iphone.

Maybe, but I don't think so. Especially more so in June/July when they inevitably release the 4th gen iPhone, perhaps using Cortex A9. Any bets if the Pandora will be released by then?

 

[laurencevde]

They used a dev-board with a chip that didn't have proper graphics acceleration. No wonder rendering's a bit sluggish...

What they're saying is that that chip is (almost) just as good for web-browsing as an atom, so if that's all you need(which is the case for most people most of the time), pure power doesn't really matter anymore, and you can base your buying decision on other factors, such as price, power usage, weight etc.

Said differently: suppose(completely fictional numbers) you've got a task (web browsing) that requires X horsepower. You've got two possible chips for that task. One gives you X horsepower, costs $10, and uses 0,1W. The other gives you 4X horsepower, costs $80 and eats up 5W(and that'll require extra cooling and a bigger battery). Which one do you choose?

And that's just for a 500Mhz A9. The real chips will be sold at (far?) higher clockspeeds. But I guess they'll rarely clock beyond just a few 100Mhz while browsing... That's another area where the OMAPs (or just about any ARM) wallop Atom: power management.

 

[Monk]

Maybe, but I don't think so. Especially more so in June/July when they inevitably release the 4th gen iPhone, perhaps using Cortex A9. Any bets if the Pandora will be released by then?

I feel pretty confident about July?

 

[streak]

I guess you don't know who Laurent is, but he definitely knows his stuff, and not a kid that wants a x360/ps3 emu. He's more then a developer from the hints dropped by certain individuals

Excellent, but i still dont see his point. I didnt say he really wants a x360 on Panda ,

Maybe, but I don't think so.

Of course everything in dev's hands. I dont expect tons of quality stuff on pandora even in peak of panda's time.

 

[streak]

Any bets if the Pandora will be released by then?

I feel pretty confident about July?

My cancelation order will be send after 15 of march, then..

 

[Exophase]

A9 is much faster than Atom at the same frequency and A9 will surely go well above 1GHz. It can also have 2 to 4 cores (and no, hyperthreading can't compete). You'll have to take my word as I can't provide figures

BTW I know of no 500 MHz A9 that will go into production (the one shown is a test chip); the lowest frequency I have seen advertised is the 720 MHz from lower-ends OMAP4.

Atom really is a piece of shit. Try to get access to one and run some benchmarks.

I have an Atom, and obviously I know everything you just said (see the post I made after the one you quoted), no need to be defensive of Cortex-A9. Nonetheless, this has nothing to do with what I said, which only pertains to comparing a 500MHz Cortex-A9 with a 1.6GHz Atom. Whether or not 500Mhz A9s will ship doesn't matter, nor does the actual speed they'll ship at, nor how they compare clock per clock.

Also if ARM keeps on comparing ARM vs Atom on benchmarking stuff it's because that's what most people use. You know that most people don't care about emulation, don't you?
And you'd be surprised how in fact browsers are very dependent on CPU performance, especially in the presence of Flash and Javascript, but not only. Run oprofile on your PC and look at the results.

I think most people do MORE than just web browsing, ARM just keeps comparing it because they know they aren't PURELY CPU driven. This is especially the case for comparisons like this one that were completely visual; when you load Google, are you going to tell the difference between one CPU that took 20ms to render it vs another that took 50ms? I doubt it. Is it useful to show? Probably. But people are still going to get the wrong idea. Which is what I said in my post.