Discussion in 'iPhone 5' started by ITS ME DAVID, Sep 16, 2012.
Found this at macrumors here http://forums.macrumors.com/showthread.php?t=1443297
Mmmm, two-cores then?
Yup same as 4s just over clocked it seems but the 4s had a score or 699 I think white the 5 is over 1600.
Wow. That is over 2x the integer performance, almost 3x the FP, and over 3x the memory performance. That is seriously smoking. That's more than a frequency bump, that's a BIG gain in performance per mhz. Bigger leap than from the 3G to 3GS, bigger than the 3GS to the iPhone 4, AND bigger than the iPhone 4 to iPhone 4S
For the specs geeks out there, a GeekBench score of 1600 means that in terms of artificial benchmarks, the iPhone 5 now beats many of the PowerPC G5 based Macs (including the baseline system which is a score of 1000), all of the G4s, and the first 1.5ghz Core Solo Mac Mini. http://browser.primatelabs.com/mac-benchmarks
What OS are those PPC G5's running? I'm thinking that's I big part of the poor bench results.
All the ones in that chart are running Mac OS X, I don't think that has much of an effect on benchmark scores. Benchmarks typically get all the horsepower they ask for unless the user is running some high priority process
For the sake of posterity, I can, however, find several results that the iPhone 5 beats on non-OS X machines, by typing platform:"Windows" or platform:"Linux" into the Geekbench search box and sifting around.
AMD A6 1.4ghz quad core - http://browser.primatelabs.com/geekbench2/478296
Atom D525 - http://browser.primatelabs.com/geekbench2/482690
Granted, I'm basically cherry picking benchmark scores that make the iPhone 5 look good against low-end x86 chips. But still, that the Apple A6 is encroaching on modern low-end x86 territory to begin with is really amazing to me. I still find it hard to believe you can squeeze so much performance out of such a small dual core chip without sucking up the same amount of power as x86. It boggles my mind.
Those Apple engineers are at it again. The iP5 scored higher than quad core phones and tablets that have higher clock speeds and double the RAM! Not to mention with a smaller footprint in size and power consumption. How did they do that?! It was assumed they used the ARM Cortex A15 but low and behold, it's an Apple designed chip. Simply amazing! I wonder why the US version of the Samsung Galaxy S III is not on the list?
Edit: I forgot the GS3 has the similar S4 (dual core) as the HTC One S so it would have scored at approx. 1258. I'm still in awe of what Apple achieved.
Wait, 9MB are unaccounted for... there's supposed to be 1024MB... What's up with that??
Is there any way to find out where the benchmark app was run?
It seems to be a random odd discrepancy. From the iPhone test results I linked, it's missing 12mb from the iPhone 3G, 3mb from the 3GS, 9mb from the 4, and 8mb in the 4S). So, my best guess is that it's simply getting the report of total memory by tallying system memory usage info and running into some rounding errors, or getting inaccurate data from something.
My first thought was that it's reserved for the frame buffer for the GPU (since 9mb happens to be roughly how much 1136x640x32bpp would take up if triple buffered), but since it's not a constant difference in each of the results of the other iPhones, I can't be sure.
Interested to find out how it will do against the galaxy note 2 quad core CPU.
The Note 2 has the same quad core Exynos 4412 as the international S3, just clocked slightly higher (1.6ghz vs. 1.4ghz), but it's a Cortex-A9, so likely, it will continue to beat the A6 in integer and floating point due to sheer number of cores, but still be behind in memory interface.
Anand is on fire, already analyzing the GeekBench results: http://www.anandtech.com/show/6298/analyzing-iphone5-geekbench-results
Nice read ^
Best of three runs and to be fair I am running a custom kernel but it is my daily so it's not sacrificing stability or battery for performance. Actually it's specifically for boosting battery. Also while I did shunt running apps such as Facebook and Twitter, I didn't uninstall anything so likely have room for minor improvement.
Anyway, I guess 1600 is good against the carrier build but nothing on an optimized setup.
Lastly, this is the international quad version. No clue what the dual core S3 can post.
Sent from my GT-I9300 using Tapatalk 2
I'm kinda surprised there's enough overhead in Android that you can gain almost 800 points by changing the kernel.
From my history of devices there seems to be a foundation approach to commits to the kernel for general product releases. Thereafter the root community tests & applies known and as discovered upgrades that are vendor, SOC and device specific. One obvious example would be when the general Linux kernel is updated between Google's platform updates. Another would be porting of performance & stability code developed in the Linux PC space. Another would be exotics such as call recording which is a kernel based hack.
Historically I've found that by halfway into the life of a phone, kernel and ROM modifications can more than double the performance and battery experience.
Sent from my GT-I9300 using Tapatalk 2
Separate names with a comma.