Samsung Galaxy S series rumours.....

[french toast]

http://androidcommunity.com/samsung-galaxy-s-4-doubles-iphone-5s-benchmark-scores-20130318/

Pretty much the Geekbench scores I posted before...Gs4 doubles the performance of the iphone 5...also suprisingly a hefty jump over the htc one..

~likely due to obvious factors such as new software (4.2.2) higher clock speed (1.9 vs 1.7) and lpddr3. ..not to mention quality/speed of nand may also play a part.

Its likely then that perhaps samsung invests the BOM on high quality internal components rather than a fancy external unibody shell...aka htc one/asus padfone infinity/iPhone5 et al.

Edit.
Here is a look at samsungs controller...how it allows for devices with up to 6.3inch screens (note3) the partnership with EA and the 16 games optimised for galaxy s4.

Samsung is clearly (and cleverly) taking a peice of every competitors pie...in this case namely nvidia...although nothing new..this looks like a far better solution to project shield using only one expensive device (phone) over a costly 2 (phone, shield) likely with near identicle results.

By paying devs to optimise games for the gs4...they take a selling point away from nvidia on the software side also....the word gazumped springs to mind

 

[lanek]

Look like the S4 for France, will be with the S600 Snapdragon, due to compatibility with the 4G frequences ( Source PCinpact ).... I still dont know if this mean the Exynos 5 work with different LTE frequences or without 4G.

 

[Jubei]

Look like the S4 for France, will be with the S600 Snapdragon, due to compatibility with the 4G frequences ( Source PCinpact ).... I still dont know if this mean the Exynos 5 work with different LTE frequences or without 4G.

Someone on XDA found this interesting info

16Gb: SAMSUNG GT-I9505BM, GT-I9505ZKADBT, GT-I9505WF, GT-I9505ZWADBT

32Gb: GT-I950032GBW, GT-I9500WHITE32GB, GT-I950032GBS, GT-I9500BLACK32GB, GT-I9500SCHWARZ32GB


All the 16GB versions are S600 but the 32 GB version is Octa. Given the fact that Octa is not being massproduced until Q2, an interesting theory is that the 32GB version will be released later in the summer with Exynos Octa

 

[lanek]

I think we will need wait the real launch for get the final information. Untill Samsung do an official annuncement.

 

[mboeller]

it seems the S4 has a good camera: http://www.gsmarena.com/samsung_galaxy_s4-review-910p6.php a little dark maybe but sharper than the XperiaZ

 

[french toast]

Benchmarks are in.
http://www.gsmarena.com/samsung_galaxy_s4-review-910p5.php

Optimus g pro keeps up with the gs4 which is surprising. .
..still 40fps on gl benchmark is nothing to sniff at.

Edit: a virtual teardown of galaxy s4 costs.
http://www.forbes.com/sites/parmyol...4-display-drives-bill-of-materials-up-to-236/

$236 up from estimated $205 of gs3...good for samsung is that 63% of the BOM gets put back into samsung.

 

[mboeller]

Benchmarks are in.
http://www.gsmarena.com/samsung_galaxy_s4-review-910p5.php

nice...so finally a really fast Adreno 320.

 

[Wishmaster]

nice...so finally a really fast Adreno 320.

Wonder whether it's from LPDDR3 or those improved drivers we were hearing about. Probably both I'd say but if it's from those improved drivers for 4.2.2 there's a chance that other adreno 320 running devices will get slight performance boost.

 

[Jubei]

Wonder whether it's from LPDDR3 or those improved drivers we were hearing about. Probably both I'd say but if it's from those improved drivers for 4.2.2 there's a chance that other adreno 320 running devices will get slight performance boost.

http://glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Sharp+FIH-VKY

Not sure if this one has lpddr3 but it does have more recent drivers and a more recent Android build 4.2.1

 

[french toast]

http://glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Sharp+FIH-VKY

Not sure if this one has lpddr3 but it does have more recent drivers and a more recent Android build 4.2.1

Sorry I may be wrong but isnt 4.2.2 denote higher android than 4.2.1?

 

[Jubei]

Sorry I may be wrong but isnt 4.2.2 denote higher android than 4.2.1?

Yes. But my point was that there might be a performance boost with already at 4.2.1

HTC One is running 4.1.2 and hitting 34 fps while the Sharp device has 4.2.1 and hitting 38 fps

 

[Nebuchadnezzar]

Let's be clear here: there are no performance differences between Android versions. And especially not between minor versions. It all boils down do vendor driver versions and compiler optimizations.

 

[Jubei]

Let's be clear here: there are no performance differences between Android versions. And especially not between minor versions. It all boils down do vendor driver versions and compiler optimizations.

Absolutely true. Not my intention of confusing the matter. Altough from what i can tell the Sharp device also has more recent drivers compared to the One

But based on my layman understanding, i believe Qualcomm releases "blobs" for the latest Android version, but its up to the OEM to implement it. Hope you can correct me if im wrong

 

[Nebuchadnezzar]

But based on my layman understanding, i believe Qualcomm releases "blobs" for the latest Android version, but its up to the OEM to implement it. Hope you can correct me if im wrong

Yes correct, however there's no correlation between Android and the drivers unless there's some kind of big jump like ICS -> JB. So it's pretty futile to argue about which device has what Android version, you're better off just to compare the graphics driver release string.

The latest Qualcomm devices come also with a new advanced CPU governor from Qualcomm that might come into play, reports say that's it's pretty smooth.

 

[Nebuchadnezzar]

But even if that's the case that doesn't mean it's going to be sucking 9+W while you're running a game, first and foremost because Android games aren't going to run four Cortex-A15s at 1.9GHz. They wouldn't know what to do with that.

Heavy games at least have one "heavy" thread so I see no problem with them maybe loading one core, and since they're all in the same frequency plane, there goes your power efficiency in that use-case.

 

[Jubei]

So it's pretty futile to argue about which device has what Android version, you're better off just to compare the graphics driver release string.

Thx for the correction. Do you have any info on what kind of speed Samsung is using on their LPDDR3 RAM?

I know Octa supports up to 12 GB per sec with what i assume is dual-channel 800 MHz. But Qualcomm boasts about that bandwidth for the SD800 but only mentions LPDDR3 support for SD600 wich makes me think its going to be lower speeds

 

[Exophase]

Heavy games at least have one "heavy" thread so I see no problem with them maybe loading one core, and since they're all in the same frequency plane, there goes your power efficiency in that use-case.

It doesn't matter what frequency plane the cores are on if the cores are off. If mobile games don't know what to do with > 2 cores then those last two cores will be idle most of the time and can be power gated for a majority of that time. The fact that Cortex-A15 uses a shared inclusive L2 means that turning the cores on and off should be relatively low latency since they just have to save internal state and push L1 to L2. Those other cores aren't going to need high frequency on and off switching. They'll probably tend to go to sleep and wake up once per frame, so can handle latency even as high as a few ms.

And I don't agree at all that games having a "heavy" thread means they're going to run 100% CPU time at any clock speed you set it at. CPU time requirements for games is mostly fixed and only scales somewhat with frame rate, which is also going to quickly hit fixed at 60Hz or so (and a lot of games will cap it at much lower). Consider that most games will want to run on iPads, so compare with that use-case: the single-threaded CPU capability is much lower (1.4GHz Swift vs 1.9GHz Cortex-A15), and the GPU power is similar. So what good is it going to do for the game to peg a 1.9GHz A15?

According to a Heise news about Project Shield, the console needs between 4 and 8 watt running a game. So not so far from the 9W.

The relevant quote is this here:
So 38Wh/5h = 7,6W for the complete system

Link: http://www.heise.de/newsticker/meld...lkonsole-mit-Tegra-4-und-Android-1778002.html

Yes, 7.6W worst case for the complete system.. that isn't close to 9W just for the SoC. Where french toast really got that 9W number was from Linley and it was just for the CPUs. I'm saying that number doesn't apply even if it's correct.

 

[wco81]

Is there actually a Samsung Shield product?

 

[Ailuros]

(deleted.)

 

[Shifty Geezer]

nVidia Shield discussion moved to the existing Shield thread.