Post your Android benchmark scores

[Cuchulain]

Download Linpak and Quadrant from the Market, both free, run them both and post your phone model, custom rom (if applicable), overclock (if any) and scores:

HTC Desire, deFrost 0.6 (Froyo), no OC.

Linpack: 39.945 MFlops
Quadrant: 1387

HTC Desire, deFrost 0.6 (Froyo), 1267mhz

Linpack - 43.656 Mflops
Quadrant - 1531

 

[Cuchulain]

I think it's important to stress that you are using JIT - which despite having no massive improvement in real world, does massively inflate the linpack result.

The revised JIT (from my limited experience) improves performance in ALL applications written in Java, and from what I gather aren't all applications for android written in Java? So as for "no massive improvement in real world", how do you justify that?

 

[Cuchulain]

The linpack score will go up ~300% but you are not going to see that kind of improvement with real apps. The OS is not written in Java and considering linpack is a benchmark of the JVM I'm just saying it's important to note that the JIT is causing the high number - and it's 'unfair' to compare it's results to other devices that are not running JIT.

Sorry but I'm not sure how you can say that you won't see that kind of improvement with "real apps", the platform java is running on is largely irrelevant.

If the JIT increases the performance of the DVM, the DVM is required to interpret the code from all apps at runtime so a faster DVM = a faster app.

 

[Cuchulain]

Fine, JIT helps a lot.

BUT what is the point in this thread? If it's to compare benchmarks between different hardware setups then i stick to my first point - it is worth nothing that JIT is responsible for the high linpack score NOT the hardware.

It's not meant to be a competition, Android covers such a huge hardware/software base it would be foolish but, if someone came on and posted great results from a custom 2.1 rom (for example), other people then have the opportunity to try it for themselves, perhaps people's results are not what they should be and they want to find out why.

 

[Cuchulain]

It gives people something new to play with.

The benchmarks are pretty meaningless because there is no baseline benchmark for each phone to start off with. You will probably not even manage to reproduce the same benchmark with two identical phones unless you can get them into the same state. So as with all benchmarking of this kind all you are doing is playing a numbers game be it with a computer or a mobile phone.

You can't even reproduce the same benchmark score on subsequent runs I think you'd need to run linpack a bunch of times and take the average.

 

[Cuchulain]

Benchmarks are always only a guideline, GeX is right, your benchmark score maybe 5x faster but your phone performance through all applications is not going to increase by that amount, you have to remember that at their heart benchmarks are very simple programs with little real world application.

Can you give me some examples as to how your phone/app performance has improved by 500%? What differences have you noticed? Do your games run 5x faster? Does your app drawer scroll 5x smoother?

Benchmarks are the only way of actually measuring performance, this is why I posted two to run, not just linpack (conveniently missed out by the neighsayers but nevermind).

Take a look at the hacked Froyo rom run of Quadrant compared to a stock Nexus one, it's nearly 3 times faster.

 

[Cuchulain]

Update:

Swapped back to a 2.1 rom, Bravox. Flashed the Jit enabler to it as well, but even with the JIT enabled on 2.1 it's still not a patch on 2.2

No O/C:

Linpack: 10.762
Quadrant: 695

Still substantially faster than every other handset with a stock rom on 2.1 but miles behind Froyo.