Nokia N93 - first Nokia with HW accelerator

I am glad to let you know that N93 sales started yesterday. Just submitted some benchmark results with final sales SW: http://jbenchmark.com/result.jsp?benchmark=hd

Just a reminder, check out Khronos booth at SIGGRAPH06 to meet me and N93, the happy couple :)

Edit: for those who are wondering the poor quality score in bilinear texturing test, jbenchmark uses tiny (4x4, if I remember correctly) texture for this test. Our HW scales all textures to 8x8 minimum to optimize memory accesses, and the the pixel accurate test fails. Free bilinear with 8x8 or larger textures works just fine in N93.
 
Last edited by a moderator:
Simon F said:
So are you using compressed textures?
Well, in theory, we could compress the textures M3G passes on to OpenGL ES, but we do not. It's just how our HW handles tiny textures internally.

Anyways, this is not a big issue. Real world use cases that a) use so small (1x1, 2x2 or 4x4) textures b) with bilinear filtering and c) expects accurate results are quite rare. I can name just one :)
 
jkemp said:
Well, in theory, we could compress the textures M3G passes on to OpenGL ES, but we do not. It's just how our HW handles tiny textures internally.
On your hardware, I know of certain of arrangements to do with compressed textures but I wasn't aware of similar for non-compressed.

It's a pity, with those benchmarks, that java is there to handicap the HW. :???:
 
Last edited by a moderator:
Simon F said:
It's a pity, with those benchmarks, that java is there to handicap the HW. :???:
Well in that sense you are correct that with native code the performance difference between any of the competition is even greater than observed +2x seen here.. Too bad there are no heavy enough native benchmark content to test against with, even 3dmarkmobile is too light weight..
 
The Java platform option should mainly be useful for compatibility with mobile game development of an older origin.
 
Is Futuremark's 'kung-fu robot atop spire' demo any more complex than their 3DMarkMobile06 game-type demos for potentially serving as a better OpenGL ES native benchmark?
 
Lazy8s said:
Is Futuremark's 'kung-fu robot atop spire' demo any more complex than their 3DMarkMobile06 game-type demos for potentially serving as a better OpenGL ES native benchmark?
Yes, that content is considerably heavier. However, it was done by FM for Nokia (premiered behind closed doors with development macro @ E3Expo 2005), and Nokia owns the content. Thus, it cannot be used by FM for benchmarking as such.

Which reminds me.. that demo was left to background for a while since we were busy to get N93 drivers in shape. We will put some polishing effort into it during flight to Boston.
 
Call me crazy, but i think that by now i've played more on my mobile than i ever did with my Gameboy, and needless to say i spent years playing with that thing when i was younger.
Now on mobiles it's just so easy, you're on the tube and u pop it out. They might be silly games, but Sudoku on my N80 is just perfect for a mobile game.

There is so much potential, because i wouldn't want yet another "thing" to carry in my pockets just to play games. Heck, even my MP3 player now isn't getting much use, cause i just don't have enough pocket space available.
 
At what speed is the N93's 3D processor clocked -- about 55-MHz? As Nokia mentioned that N-Gage smartphones might end up being more powerful than the development kit specification, are/will the actual phones clock the 3D processor higher -- perhaps to 110-MHz?

Since N-Gage smartphones will vary in specs like processor and memory, how will the games be ensured to run well on all of the different configurations?
 
Since N-Gage smartphones will vary in specs like processor and memory, how will the games be ensured to run well on all of the different configurations?
I should imagine it will be no different to the situation with PC games where there are many different system configurations.
 
i always play tetris or some casino game on my phone . it kills the time on the train!
 
How far will Nokia go to address the performance discrepencies of different system configurations with a hardware design approach, such as clocking an MBX Lite in an OMAP2430 twice as high as an MBX in an OMAP2420, versus a software solution like simply having developers scale the level of graphics detail in their games for different phones?
 
How far will Nokia go to address the performance discrepencies of different system configurations with a hardware design approach, such as clocking an MBX Lite in an OMAP2430 twice as high as an MBX in an OMAP2420, versus a software solution like simply having developers scale the level of graphics detail in their games for different phones?

Will be handled :D
 
How far will Nokia go to address the performance discrepencies of different system configurations with a hardware design approach, such as clocking an MBX Lite in an OMAP2430 twice as high as an MBX in an OMAP2420, versus a software solution like simply having developers scale the level of graphics detail in their games for different phones?

To answer a bit more than Patric (without actually saying anything:), we simply cannot achieve equal level of performance through-out product portfolio. It wouldn't make any sense either. However, we will do what ever it takes to minimize the fragmentation, mainly from feature side. It is then up to application developer to either detect the performance level and adjust content accordingly or provide manual adjustments to features (like in PC world).

While we do have a transition period to handle (performance difference of sw rasterizer vs. hw acceleration), in long run, once hw acceleration becomes commodity, this won't be such a big deal - generally more powerful devices tend to have higher screen resolutions and thus higher performance requirement for the same content. If your favourite game runs 15fps vs. new model runs it 30fps... you do the same as on pc and upgrade your gear, if you really want the higher performance.
 
While we do have a transition period to handle (performance difference of sw rasterizer vs. hw acceleration), in long run, once hw acceleration becomes commodity, this won't be such a big deal - generally more powerful devices tend to have higher screen resolutions and thus higher performance requirement for the same content. If your favourite game runs 15fps vs. new model runs it 30fps... you do the same as on pc and upgrade your gear, if you really want the higher performance.[/QUOTE]

How long a transition period are we talking about. Since I am hoping that by 2007 we shall see existing OMAP 2420 devices supplemented by OMAP 2430 S60 platform and OMAPV2230 and maybe 2008 some OMAP3430 devices...oh yes then theres SGX with OPEN GL ES2.X
 
Last edited by a moderator:
Seems like Nokia knows how to commit to this platform for the long-term.

OpenGL ES should provide a good evolutionary path for the technology. The evolution should be very dynamic, too, since new phone models are released almost seasonally and smartphone users replace their hardware within every couple of years or sooner.

How big of a selling point will gaming need to become to get Nokia to adopt new processor technology at near the speed of the Asian markets? Seeing OpenGL ES 2 phones before the third quarter of 2008 in western markets would be good. Also, for the current generation 3D, one of the more gaming-oriented phone models should go all-out and use a ~165-Mhz-clocked graphics core.
 
Seems like Nokia knows how to commit to this platform for the long-term.

OpenGL ES should provide a good evolutionary path for the technology. The evolution should be very dynamic, too, since new phone models are released almost seasonally and smartphone users replace their hardware within every couple of years or sooner.

How big of a selling point will gaming need to become to get Nokia to adopt new processor technology at near the speed of the Asian markets? Seeing OpenGL ES 2 phones before the third quarter of 2008 in western markets would be good. Also, for the current generation 3D, one of the more gaming-oriented phone models should go all-out and use a ~165-Mhz-clocked graphics core.

Nokia has been one of the most active companies in OpenGL ES 2.0 specification, and will certainly be one of the first to productize it (if not, consider me sacked:). Perhaps the most appealing feature of ES2.0 is the fact that industry gets rid of fairly large amount of vendor specific extensions that causes fragmentation to market and gives grey hair to the developer.

As said earlier, we try to do everything to minimize the feature differences between the devices. Let's take N93 for an example. We decided not to expose any of the hardware's vendor specific extensions, including vertex shaders. This was a painful decision but removed need of quite substancial double-codepath from applications. However, we did leave vendor specific texture compression available, since omitting that would have caused significant performance hit. In this case though, it is quite stright-forward from developer to include both compressed and non-compressed textures. This is one reason we drove OES_compressed_ETC1_RGB8_texture to ES2.0 spec. Every Nokia ES2.0 device will feature this compression format, guaranteed. And it seems to be no problem for every gfx core vendor out there to support it.

To answer to your question, my feeing is that gaming alone will never justify the adoptation speed of Asian market. There is no point of being the first in the market, but to be the first profitable in the market. However, gaming is not the only use case for graphics. Our S60 UI already is totally based on SVG, and OpenVG acceleration is becoming more and more important. In addition, OpenGL ES will be featured in number of non-gaming applications in future. And all of these combined justify the existance of graphics hw in a convergence product like our handhelds are.
 
Back
Top