Linpack and Qaudrant.

[carcrazygts2]

Well i decided to run Linpack and a Qaudrant after each mod I performed to my phone i created a data base. Just curious to what others are seeing. I also have had the highest qaudrant benchmark i've seen with out doing the stagefright mod. Here are my numbers from stock OTA all the way to FlyX+Tranquility.

MFLOPS/Qaudrant:
Stock OTA 2.2- 8.784 MFLOPS/ 986 quadrant
Rooted with Bloatware renamed- 9.249 MFLOPS/ 1098 qaundrant
Rooted, Deodex, No Bloat- 12.982 MFLOPS/ 1176 qaudrant
Rooted, Deodex, Tranquility- 15.571 MFLOPS / 1676 qaudrant
Rooted, Deodex, Tranquility, Build.prop mod(stagefright false) 16.836/ 1708 qaudrant
Rooted, Deodex, Tran+Flyx, Build.Prop mod(stagefright false) 17.867/ 1751 qaudrant

Why havent I seen MFLOPS in the 20's with the X? Is because of no JIT?

 

[fr3nzy]

whats stagefright false?

 

[carcrazygts2]

It has to do with the high def portion of the phone when it is set to true in build.props it will make qaudrants go well above 2000. My highest was well over 2600.

 

[fr3nzy]

ohh alrightt ahaha

 

[Dave12308]

I believe the Quadrant score jumps when you enable Stagefright because Quadrant is not compatible with Stagefright, and thus the encoding tests are skipped. It makes it appear that they have completed instantaneously, thus the jump in benchmark numbers.

 

[tobytl]

...
Why havent I seen MFLOPS in the 20's with the X? Is because of no JIT?

We do have JIT, that's why we get the jump we do. I really have no idea why people keep saying this... The reason we get lower linpack scores with JIT than others, is because of the difference between the Scorpion CPU on Snapdragon SoC's vs. the Cortex-A8 that's on the OMAP we're running. The Scorpion is very close to the Cortex-A8, due to the necessities of running the same instruction set, but one of the differences improves raw floating point operations which is what linpack specifically tests (I honestly can't remember why, I think the Scorpion just has a "wider" pipe to the FPU/VeNum/Whatever it was called). It should be noted that this doesn't equate, at all, to real world performance. Very few things outside of hardcore data analysis or media compression rely very heavily on a cpu performing floating point ops (GPU yes, CPU not so much).

What I'm shocked at is your low 2.2 OTA stock scores. Those are pretty close to my 2.1 scores from when I first got my device. When I sbf'd and went OTA, my scores before messing around with the system were pretty close to your root+deodex+tranq scores... Are those normal? (I no longer have multiple DX's to test with, just mine).

@Dave12308: Quadrant does indeed test h264 decoding with stagefright. It's just that stagefright is much, much faster, especially on initial decoder startup. Seriously, enable stagefright and load up an h264 video, startup and skipping around are much faster (until stagefright pukes, which happens occasionally and you end up with audio desync, still fast though...).

 

[carcrazygts2]

Well I have web quite happy with my phone. It is plenty fast. I'm interested when some is able to undervolt 2.2 for battery life.