JBenchmark Pro launched

What are the major differences in workload between the HD and Pro tests which might be the cause of the switch in rankings of the GoForce 4800-powered W900i and the MBX-Lite/PNX4008-powered trio of Sony Ericsson phones?
 
I now see the breakdown of the different tests in Pro. The PNX4008 Sony Ericssons are the only devices to pass both of the high quality tests for 3D graphics.

Why don't some of the 3D hardware accelerated phones really outpace phones like the Nokia 6280 in the low and medium quality 3D graphics benches?
 
Why don't some of the 3D hardware accelerated phones really outpace phones like the Nokia 6280 in the low and medium quality 3D graphics benches?

Whereas in pure native benchmark, performance difference comes pretty much from formula:

gl_driver+gpu_hw+memorybandwidth

in java M3G there are much more variables in the formula:

jvm+cpu_perf+memorybandwith+m3g_engine+ gl_driver+ gpu_hw + how_efficiently_all_of_these_have_been_glued

In devices without HWA, the formula is somewhat more simplified and thus easier to get close to theoretic maximum.

On top of that, Nokia's M3G implementation is pretty much industry benchmark - ask that from any serious M3G developer. I guess it pays to be active in standardization :)

Edit: Great benchmark suite by the way, kudos to the development team!
 
Last edited by a moderator:
N95 results have started showing up. It's taking the top spot, as claimed. Can't tell yet how the trilinear tested.

An application for uncovering which generation of ARM core is being used in a phone and also the clockspeed at which it's running has been launched, called JBenchmark ACE (ARM CPU Estimator).

http://jbenchmark.com/index.jsp

Quite a large variety of clockspeeds among OMAP2420 phones from Japan and the US is indicated by the tests.
 
Back
Top