But hey, you can't win 'em all, and at least I was right about one thing; the 1GHz OMAP
Then why did it score so well in this test? Taylor Wimberly, owner of AndroidAndMe.com (yes, the same website I've frequently complained of putting bad data in its articles), has performed a series of benchmarks that explains why.
His latest article, titled "What is the Fastest Android Phone?" uses a professional version of the Android benchmarking tool Quadrant that shows how it breaks down performance scores, and reveals that I/O (data read/write) performance in the Droid Shadow (aka Droid X) is why it was beating the Galaxy S (aka Samsung Captivate for the purpose of Taylor's tests) in benchmarks. As it turns out, there is a bug in the stock firmware of the Galaxy S that severely hinders I/O performance. A fix from some open source developers apparently corrects this issue, putting I/O performance back on par with the Droid X. One might argue however that until an official fix is released, the Droid X maintains its edge when comparing stock, unmodified phones. I'd be interested to know if the bug exists in all 4 US-carrier Galaxy S variants.
UPDATE: Thanks Anonymous for posting below with the following link for an I/O fix for the Captivate: http://pocketnow.com/android/two-performance-hacks-for-samsungs-captivate
Taylor's article demonstrates that for actual CPU and GPU performance, the Galaxy S / Captivate (I'm just going to call it Galaxy S henceforth) beats the Droid X in both categories. Due to the vast improvement in efficiency (via the new JIT compiler) in Android 2.2, the Nexus One (or, amazingly, even an aging HTC Magic running Cyanogen 6 Android 2.2) can outperform either of the phones by a significant margin. When the Galaxy S and Droid X get Android 2.2, they'll truly be unstoppable; at least until we see dual-core Snapdragons and Cortex A9 SoCs in showing up in late 2010 / early 2011.
The Droid X beats the Galaxy S in one field by a small margin; memory performance. Both phones have significantly better memory performance than previous phones likely due to the use of LPDDR2 memory as opposed to the much slower LPDDR used until now. The Droid X manages to beat out the Galaxy S by a fraction regardless, but I'd be interested to know if the I/O bug in the Galaxy S affects memory performance at all.
Very interesting stuff. Android And Me has redeemed itself fairly well, and Taylor seems to be making an effort to verify his data before he posts. I could nitpick about his description of the Snapdragon being a Cortex-A8 chip, but I'll let it slide this time. :-p
Posts
Hi,
ReplyDeleteDo you know if the european Galaxy S have got a gyroscope ?
It seem's some versions on US market have got it and some others doen't.
Thx.
J.C
Well, I can't tell you with any certainty, but if you look halfway down this article at the slide titled "Six Axis Motion Sensing", an international version of the Galaxy S is displayed.
ReplyDeleteSamsung performed the presentation so I'm inclined to believe it's true.
Yeah, a link might have been helpful, right??
ReplyDeletehttp://mobile.venturebeat.com/2010/07/12/mobilebeat-samsung-offers-glimpse-of-superphone-future-with-galaxy-s
Thx for your answer.
ReplyDeleteDoes 6 axis means 3 axis accelerometer + 3 axis gyroscope or could it be 6 axis accelerometer and no guroscope ?
I wonder if the demo of n.o.v.a from Gameloft ( http://www.youtube.com/watch?v=6nKSKA-gsco ) for the iPhone 4 is technicaly possible on the Galaxy S ...
J.C
My understanding is that it's a 3 axis accelerometer + 3 axis gyroscope.
ReplyDeleteCan't view that video while I'm at work, but I'll check it out when I get home tonight.
I’m new to trying to understand these and other benchmarks. Could you explain your understanding of why GLbenchmark.com has the Droid as the fastest “CPU Performance: Integer” device compared to Quadrant's results having it the slowest in “CPU Test”?
ReplyDeleteProbably due Android 2.2 running on the Droid. Take a look at this page and you can see pretty clearly that some Droids are performing 2x as well as others; I suspect those are the ones running 2.2.
ReplyDeleteGLBenchmark CPU Performance: Integer Results of Motorola Droid (Milestone Sholes)
By contrast, compare to the relative consistency in the Galaxy S results: GLBenchmark CPU Performance: Integer Results of Samsung i9000 Galaxy S
ReplyDeleteThanks, I see your point. Amazing that a "minor" OS upgrade could/would make such a big difference.
ReplyDeleteThe JIT compiler provides a truly amazing improvement in CPU performance. The original AndroidAndMe article illustrates that an HTC Magic, with a 3-year-old 528 MHz ARM11 CPU (which is about half as fast as a modern Cortex-A8 running at the same clock speed) can outperform the fastest smartphone processors on the market simply by running Android 2.2 with the JIT compiler.
ReplyDeleteReal-world CPU performance gains are shown to be between 400-500%... pretty outstanding!
Impressive !
ReplyDeletehttp://pocketnow.com/android/two-performance-hacks-for-samsungs-captivate
Thank you for that link, Anonymous... I've been waiting to see the I/O fix hit a US-carrier Galaxy S. I've updated my original post to include your link.
ReplyDeleteMuch appreciated!
I'm the anonymous of these comments :-)
ReplyDeleteWith Android 2.2 numbers will be amazing ^^
About my first question I still don't know if the Galaxy S has got a gyroscope or not :-p
I should received mine in a couple of days (french Galaxy S ^^).