12 replies to this topic

[Stevos]

Does anyone know why the Galaxy S doesn't get anywhere near the performance boost that the Nexus gets for 2.2?

The Galaxy 2.2 should wipe the floor with the nexus 2.2 benchmark results considering how they perform relatively with 2.1

I assume it can't be related to the lagfix/filesystem issues as they would hamper 2.1 and 2.2 roms on the galaxy similarly.

...
(based on quadrant results)

Edited by Stevos, 02 December 2010 - 04:04 PM.

[john306]

Stevos, on Dec 2 2010, 16:02, said:

Does anyone know why the Galaxy S doesn't get anywhere near the performance boost that the Nexus gets for 2.2?

The Galaxy 2.2 should wipe the floor with the nexus 2.2 benchmark results considering how they perform relatively with 2.1

I assume it can't be related to the lagfix/filesystem issues as they would hamper 2.1 and 2.2 roms on the galaxy similarly.

...
(based on quadrant results)

My Quadrant results on my SGS upgrade which came with 2.2 FROYO was about 900, once I applied the lag fix it went to about 1900, the Nexus One 2.2 score is about the 1300 mark.
I don't know how accurate these readings are or how they relate to real life usage but mine definately feels much

[Stevos]

What I mean is that the froyo benchmark improvements for the nexus one seem to be based on JIT causing the processor scores to be much higher, which doesn't seem to happen with the Galaxy S.

Stock quadrant scores for the nexus one on 2.1 are lower than the Galaxy S on 2.1, whereas stock scores on 2.2, the nexus destroys the galaxy by an enormous margin.

2.1 stock ROMs (quadrant score):
Nexus one: 500ish
Galaxy S: 800ish

2.2 stock ROMs (quadrant score):
Nexus one: 1300-1400ish
Galaxy S:  900ish

So the 2.2 JIT benefits and optimisations only give the Galaxy S a ~10% improvement, whereas they give the Nexus one a ~270% improvement. Pretty much purely on the CPU section of the test.

The lag fix problems are common to the 2.1 and 2.2 Galaxy ROMs so shouldn't be skewing the results.

This thread (and the benchmark charts in it) addresses some of the questions: http://androidforums...nt-lag-fix.html

The lag fix is boosting a different part of the overall score, and NOT the CPU part which is improved with froyo on the nexus one.


It seems that Froyo does almost nothing for the Galaxy S CPU performance, whereas it massively boosts the Nexus. Something seems very wrong here. Maybe the JIT implementation is buggy or something?

Edited by Stevos, 03 December 2010 - 05:05 PM.

[Stevos]

Oh well, assuming it's due to Froyo being heavily optimised for Snapdragons then we may see some good boosts in Gingerbread builds if the Nexus S plays out with a Hummingbird variant.

Shrug.

[Genesys]

Snapdragon processors have the NEON unit which boosts the performance using the JIT. Galaxy S doesn't have that, so the JIT improvements are not the same, but less more.

Edited by Genesys, 05 December 2010 - 03:43 PM.

[Stevos]

Thanks for this info, it looks like there's more info googlable about the NEON issue.

[Stevos]

The upshot seems to be that whilst the Hummingbird is generally a stronger processor than the Snapdragon, it isn't compatible with the optimisations that were built into Froyo/JIT.

Seems like an odd feature to leave off (arguably stupid? considering the apparent performance benefits), but it basically appears to mean that it won't be as fast as it should be, and it's unlikely that the performance will be improved much if we get gingerbread.

I guess Samsung will put out the next version of the chip with Advanced SIMD/NEON support and just chalk it up to obsolescence...

[Stevos]

Although this Hummingbird press release says that it does support NEON

http://www.samsung.c...do?news_id=1030

The plot thickens. Maybe it's an inferior implementation, or maybe the problem is something else.

[Stevos]

And here, we have indications that Froyo CPU/JIT performance can be sorted on the SGS

http://www.androidpo...d-in-beta-form/

Check out the graphs vs Nexus One 2.2.  This is very encouraging.

Edited by Stevos, 06 December 2010 - 05:35 PM.

[dagrim1]

Stevos, on Dec 6 2010, 18:33, said:

And here, we have indications that Froyo CPU/JIT performance can be sorted on the SGS

http://www.androidpo...d-in-beta-form/

Check out the graphs vs Nexus One 2.2.  This is very encouraging.

From what I read the boosted CPU score was because of one of the CPU tests not completing properly (it terminates immediately).

[Stevos]

Ah, okay well that would skew the results somewhat... Interesting that it skews them in a manner that almost exactly replicates the Froyo cpu score improvements though.

In other positive news, the announced Nexus S hardware looks to be a Galaxy S, but with a few minor changes (NFC hardware, slightly different storage memory - 16 GB iNAND with no SD, missing FM radio, maybe a few other bits)

That means the Galaxy S is almost certainly going to get Gingerbread fairly quickly - possibly from the community, and should be very well positioned for future updates (The Nexus S will get Honeycomb for sure, and I'd guess probably at least one update after that, so we could be very well set indeed.)

It also means that Google are confident in the basic hardware platform and performance, so I expect we may see some good boosts when we get Gingerbread. They wouldn't release a new flagship phone and OS without doing some serious work on chip optimisation, and presumably making sure it performs better than the old Nexus in most (all?) categories.

I wonder what the effect of that different storage is, and whether it's a simple way to address the IO / RFS / lagfix issues that the Galaxy S has had.

Edited by Stevos, 07 December 2010 - 05:57 PM.

[lgkahn]

Stevos, on Dec 7 2010, 12:48, said:

Ah, okay well that would skew the results somewhat... Interesting that it skews them in a manner that almost exactly replicates the Froyo cpu score improvements though.

In other positive news, the announced Nexus S hardware looks to be a Galaxy S, but with a few minor changes (NFC hardware, slightly different storage memory - 16 GB iNAND with no SD, missing FM radio, maybe a few other bits)

That means the Galaxy S is almost certainly going to get Gingerbread fairly quickly - possibly from the community, and should be very well positioned for future updates (The Nexus S will get Honeycomb for sure, and I'd guess probably at least one update after that, so we could be very well set indeed.)

It also means that Google are confident in the basic hardware platform and performance, so I expect we may see some good boosts when we get Gingerbread. They wouldn't release a new flagship phone and OS without doing some serious work on chip optimisation, and presumably making sure it performs better than the old Nexus in most (all?) categories.

I wonder what the effect of that different storage is, and whether it's a simple way to address the IO / RFS / lagfix issues that the Galaxy S has had.

i'm running 2.2.1 jpy now on my galaxy s and qudrant score is bassicaly the same as the nexus 1 with 2.2 about 1275 average.. this is with no lag fix applied.

[Eicodi]

Hi, I bought a Galaxy S today. and i have android 2.1. how can i update to froyo?
Sorry my english.

Thanks