Video of Froyo on Moto Droid

[Darkseider]

I'm using Chevy's ULV 1ghz kernel, but turned it down to 550 max freq, 250 min. These are the average of 5 runs.

linpack: 4.3
short: 390
long: 1410 (any idea why this would be so high?)

Yup and the the LinPack is right in line as well. So I guess the question is how reliable is the Linpack benchmark if all it is doing is scaling linearly? With Froyo all it does is produce the same linear result as Android 2.1 x 2. So it does show that Froyo is doing something. The Incredible and Desire score a consistent 7.1 - 7.2 and a Nexus One at 1.13 Ghz scores a 7.75 so that shows .71 MFlops per 100 Mhz on a Snapdragon. From this we can infer that the OMAP 36xx series is more efficient at performing floating point calculations. It also proves that the linpack benchmark scales nearly perfectly with clockspeed based on processor.

So what does all this mean to me you and the guy next to you? Well other than comparing how well a SoC performs floating point operations compared to another SoC, nothing. There is still a bunch of other things like Integer ops, 2D/3D performance based on the GPU or lack thereof. Reads/Writes to main memory/SD Card and processor cache. So if nothing else it gives bragging rights to those of us who chose OMAP based devices. We can say we can do floating point math better than you! NYAH NYAH! but that's it.

you know a lot more than i do about all of that, but that's kinda the same conclusion i was getting to when looking at all of these numbers... plus (and fully I realize this is not the same as how froyo will be), when i've enabled JIT on my stock/non-oc'd phone, the linpack went up, but there was no noticeable difference in performance at all... the only thing that seemed to make a noticeable difference in actual phone usage (with the scores as well) was overclocking

That's because the JIT will only speed up programs that take advantage of it. The UI unfortunately does NOT so you will see no difference at all in basic function.

 

[jntdroid]

Yup and the the LinPack is right in line as well. So I guess the question is how reliable is the Linpack benchmark if all it is doing is scaling linearly? With Froyo all it does is produce the same linear result as Android 2.1 x 2. So it does show that Froyo is doing something. The Incredible and Desire score a consistent 7.1 - 7.2 and a Nexus One at 1.13 Ghz scores a 7.75 so that shows .71 MFlops per 100 Mhz on a Snapdragon. From this we can infer that the OMAP 36xx series is more efficient at performing floating point calculations. It also proves that the linpack benchmark scales nearly perfectly with clockspeed based on processor.

So what does all this mean to me you and the guy next to you? Well other than comparing how well a SoC performs floating point operations compared to another SoC, nothing. There is still a bunch of other things like Integer ops, 2D/3D performance based on the GPU or lack thereof. Reads/Writes to main memory/SD Card and processor cache. So if nothing else it gives bragging rights to those of us who chose OMAP based devices. We can say we can do floating point math better than you! NYAH NYAH! but that's it.

you know a lot more than i do about all of that, but that's kinda the same conclusion i was getting to when looking at all of these numbers... plus (and fully I realize this is not the same as how froyo will be), when i've enabled JIT on my stock/non-oc'd phone, the linpack went up, but there was no noticeable difference in performance at all... the only thing that seemed to make a noticeable difference in actual phone usage (with the scores as well) was overclocking

That's because the JIT will only speed up programs that take advantage of it. The UI unfortunately does NOT so you will see no difference at all in basic function.

and 2.2 does, i assume - so theoretically an app on froyo, if not written to take advantage of jit, could run exactly the same as 2.1?

 

[Darkseider]

you know a lot more than i do about all of that, but that's kinda the same conclusion i was getting to when looking at all of these numbers... plus (and fully I realize this is not the same as how froyo will be), when i've enabled JIT on my stock/non-oc'd phone, the linpack went up, but there was no noticeable difference in performance at all... the only thing that seemed to make a noticeable difference in actual phone usage (with the scores as well) was overclocking

That's because the JIT will only speed up programs that take advantage of it. The UI unfortunately does NOT so you will see no difference at all in basic function.

and 2.2 does, i assume - so theoretically an app on froyo, if not written to take advantage of jit, could run exactly the same as 2.1?

It may be a slight bit faster due to the overall improvement in Froyo over Eclair but not the JIT improvement that folks are talking about 4.5x - 5x, etc...