How is that hardly fair? It's both phones on 2.1, we're testing hardware performance and software updates don't mean anything. Just because 2.2 can get you better benchmarks doesn't mean when 2.2 for the Epic comes out those numbers won't be squashed so software doesn't mean anything.
That's not even wrong.
Optimizations and maker-added kernel changes are always at play.
With 2.2 compared to 2.1 performance, an Epic might get 1.5x, 1.2x, 2x, or 3x better.
Things don't scale linearly no matter how much you want them to.
Do I expect the Epic, having been better than an Evo w/ both running 2.1, to be better than a stock Evo when both are running 2.2?
Well, yeah - kinda.
But the point of testing is simple and in my business, we have a saying - one test is worth a thousand expert opinions.
The test you posted is irrelevant now - Epics are to ship with 2.2 and the Evo is officially at 2.2.
Testing like this by definition includes all hardware and the execution profile defined by software.
To say that software doesn't mean anything is to not understand how software works.
This is my opinion based on professional experience. Cray used to test their compiler upgrades, in part, using my digital signal processing routines, designed specifically for that purpose.
And you're wrong the 65nm Snapdragon sucks in comparsion to the 45nm Hummingbird proven time and again, look it up.
Please carefully consider these remarks of mine -
http://androidforums.com/support-troubleshooting-evo-4g/150025-how-hummingbird-so-much-faster-than-snapdragon.html#post1377369
You'll note I claim that the reviewers are all attributing a GPU metric in mega-triangles per second that is _9 times_ greater than Samsung's own Hummingbird datasheet.
Further - the link there shows a pro output of one benchmark that interesting breaks out the OpenGL (graphics) performance on the Snapdragon as much higher than thought. So - the Snapdragon GPU might not be the dog of the world.
That was with a custom kernel.
What does this tell us?
It tells us what our common sense tells us:
The operating system (Android) is one size fits all - and making it shine is up to the maker's getting their customizations to that right.
To date, no maker has produced a phone that independent devs haven't been able to significantly improve.
Note in that same referenced thread that one Vibrant (Galaxy-class) owner is getting a stellar Quadrant performance of over 1800 using a single root trick with the stock kernel.
Neither the Snapdragon, Hummingbird, nor OMAP are dogs.
Their true differences - despite what tech wannabes with websites proclaim based on stock performance and benchmarks (that many of them openly proclaim to not know what the benchmark means other than bigger is better) - have yet to be learned.
The only thing I'm terribly certain of is that the Epic, out of the box, won't have the 30 fps cap that some people care about - and some don't.
Out of the box - better fps - worse GPS.
As I already cannot stand the sight of the Epic, I'll be replacing its launcher at minimum, and so I won't be stock and I'll have to learn if that helps, hinders, or is net-no-change to its performance.
(Assuming they fix the GPS thing and we go thru with the purchase.)
We don't carry benchmarks - we carry phones.
They run apps. When I start seeing performance assessments on the apps on the various phones, maybe then, for me personally, I'll care.
I simply do not understand this fascination with CPU architectures - especially when so little about them is truly revealed due to industrial secrecy.
The only time I've ever cared about CPU architecture was when I writing or modifying kernels.
These three - Snapdragon, Hummingbird, OMAP - are now creating what I consider superphones. They'll continue to incrementally leapfrog each other in performance as their models progress - due to user demands from the _apps_ and the new devices/uses.
I just don't see how you can go wrong buying a phone with any of them on the inside.
