Obligatory iPhone 4 Thread...

[Chandler55]

so with 4x the pixels and the same gpu as 3gs we're back to iphone 3g fillrate performance? why couldnt apple have made some hardware pixel doubler scalar so we can still make our apps 320x480 and have great 2d sprite rendering performance

 

[Grall]

so with 4x the pixels and the same gpu as 3gs we're back to iphone 3g fillrate performance?

This is a concern for those who care about fast 3D performance, but Apple hasn't announced any performance details for the GPU yet - do they ever actually do that though?

It's possible the GPU has seen a clock speed bump to help deal with the greater resolution screen...

 

[Chandler55]

This is a concern for those who care about fast 3D performance, but Apple hasn't announced any performance details for the GPU yet - do they ever actually do that though?

It's possible the GPU has seen a clock speed bump to help deal with the greater resolution screen...

it has the a4 chip which is the same as the one in the ipad, and I believe that has the same gpu as the 3gs. the fillrate isn't that bad, it's mostly a 2d issue where there's lots of objects to be drawn over other objects. 3d games should still be ok and get a huge boost from the A4 CPU. I was just hoping for a beefier fillrate since that is pretty much the bottleneck of most 2d games.

 

[Rys]

If you're seeing particular issues in your games on iPad or iPhone 4 and you're wondering about any potential optimisations or what have you, feel free to email me and Xmas and the other PowerVR devtech guys :smile:

There's generally always something we can do to give you a boost.

 

[Panajev2001a]

If you're seeing particular issues in your games on iPad or iPhone 4 and you're wondering about any potential optimisations or what have you, feel free to email me and Xmas and the other PowerVR devtech guys :smile:

There's generally always something we can do to give you a boost.

Uhm... if I ask "make a native Mac OS X version of all your various tools (minus DX related features)" adding the following:

1.) "please... "
2.) tons of your chips go in devices that need an SDK that runs only on OS X...
3.) sometimes the artist is the coder too... or the coder is the artist... virtualization is not a panacea...

answer ?

 

[Simon82]

Some questions to discuss...

1) It seems from the image that it's a smaller package of the Ipad Soc. Someone could say that it's the SAME soc or a simpler A4 based one?

2) Have Apple talked about the frequency or just it's based on A4? It could not be 1Ghz after all...

3) How do you compare that Soc with the S5PC110 of the Galaxy S?

 

[Ike Turner]

Some questions to discuss...



3) How do you compare that Soc with the S5PC110 of the Galaxy S?

It's exactly the same ARM core:
http://www.eetimes.com/rss/showArticle.jhtml?articleID=225402172&cid=RSSfeed_eetimes_newsRSS

 

[tangey]

Thats hardly surprising as the core was supplied by intrinsity to both companies, and the chip manufactured by Samsung.

 

[Simon82]

It's exactly the same ARM core:
http://www.eetimes.com/rss/showArticle.jhtml?articleID=225402172&cid=RSSfeed_eetimes_newsRSS

Interesting! i thoke that sane cortex architecture would not be so similar.

 

[Xmas]

so with 4x the pixels and the same gpu as 3gs we're back to iphone 3g fillrate performance? why couldnt apple have made some hardware pixel doubler scalar so we can still make our apps 320x480 and have great 2d sprite rendering performance

Who says you can't?

Uhm... if I ask "make a native Mac OS X version of all your various tools (minus DX related features)" adding the following:

We do have a Mac OS SDK, the tools use X11 though as we use a cross-platform toolkit.

 

[Panajev2001a]

Who says you can't?


We do have a Mac OS SDK, the tools use X11 though as we use a cross-platform toolkit.

I stand corrected, did browse the PowerVR website and found all tools needed... sorry... .

 

[Lazy8s]

The SGX in the S5PC110A01 should have a similar fill rate but approaching double the shader performance to the A4's, assuming frequency is fairly close (Apple might've been a little more aggressive there and could've closed the performance gap somewhat with that.)

 

[aaronspink]

Your technical description is, as far as I understand it, perfectly correct. However, I was under the impression that the technology (introduced and described here, 11th of June 2008) wasn't too interesting for higher end sensors since the larger pixel area, typically 10 times or so greater than the sensor in the iPhone 4, made the losses due to the reasons you described fairly small.
The sensor described in the Sony press release - 5MP, 1.75um squared et cetera bears a striking resemblance to the one used in the iPhone 4. Their commercial sensor products are described at Sony for the truly nerdy.

Oh, I wasn't clear. By high end I meant the 10K-1M sensor only products used as backs for cameras and telescopes. Almost all are back illuminated and many are also cryogenically cooled.

 

[Chandler55]

Who says you can't?

WHOA, did they? that actually makes sense with the ipad's 2x mode

 

[Entropy]

Oh, I wasn't clear. By high end I meant the 10K-1M sensor only products used as backs for cameras and telescopes. Almost all are back illuminated and many are also cryogenically cooled.

Ah. Telescopes. Not quite the high end I assumed. In terms of regular photography the technique is fairly new though, and has been promoted for small sensors, where it brings a limited but appreciable benefit. I wouldn't expect miracles out of the camera on any phone, but the iPhone 4 sample shots look quite good for the breed.

 

[aaronspink]

Ah. Telescopes. Not quite the high end I assumed. In terms of regular photography the technique is fairly new though, and has been promoted for small sensors, where it brings a limited but appreciable benefit. I wouldn't expect miracles out of the camera on any phone, but the iPhone 4 sample shots look quite good for the breed.

Considering they kept the same sensor area and moved to back illumination while increasing the resolution, I wouldn't be surprised if the sensor they are using is better than a lot of real consumer digital camera in many situations. Where it is going to fall down (where basically all phone camera fall down) is in the optics. Really don't understand people putting an 11MP sensor in a phone when the optics probably can't really handle more than 2-3 reasonably.

 

[flynn]

I watched the keynote yesterday and am definitely very interested in this one. I carry my iPod Touch with me all time and would be happy to have a single device for both music and phone.

As a programmer I'm pretty excited about the hardware as well. I hope the GPU can keep up with the extra pixels, but it looks good.

 

[Squilliam]

Hey can we expect an iTouch based upon this new hardware any time in the near future? How will they differ in terms of overall performance? Will we essentially have the iPhone minus phone stuff/GPS with everything else intact?

 

[Grall]

the fillrate isn't that bad, it's mostly a 2d issue where there's lots of objects to be drawn over other objects.

Why don't you draw your objects front-to-back, and let the hardware eliminate the overdraw? You should be able to stack thousands (tens of thousands) of objects at full framerate without any problems at all...

 

[Xmas]

Why don't you draw your objects front-to-back, and let the hardware eliminate the overdraw? You should be able to stack thousands (tens of thousands) of objects at full framerate without any problems at all...

2D sprites often require blending so the hardware can't eliminate overdraw (unless you make the effort to
use additional geometry for the opaque parts).