Snapdragon 820 Performance Preview
Testing Snapdragon 820
In usual fashion, we were invited to Qualcomm's headquarters to get a first look at Snapdragon 820’s performance. Since retail products using the 820 are not available yet (or have even been announced), we subjected Qualcomm's Mobile Development Platform (MDP) to a battery of tests.
The smartphone MDP is an oversized phone used internally by Qualcomm for testing and development. Obviously all the usual caveats apply: We’re testing pre-production hardware and software, so we could encounter some performance anomalies. Furthermore, because the MDP chassis is larger and thicker than the sleek flagship phones that will eventually house the 820, we cannot make any assessments of thermal throttling. Also, with only a limited amount of hands-on time, we’ll need to wait for shipping hardware to quantify battery life.
The data that we will be presenting then, serves as a performance preview, an early look at the 820’s potential. While the final numbers may change, our experience testing both the Snapdragon 805 and Snapdragon 810 MDPs (those were both tablets) showed good correlation with eventual shipping products.
Devices Used for Testing
Even though testing took place outside of our lab, we followed our usual testing procedures as closely as possible. We used our own benchmark files and ran each test at least twice, averaging the results, like we usually do. Because of the limited testing time, however, we did not have time to allow the MDP to cool between tests. We did not see much variation between successive runs, though, so thermal throttling did not appear to impact the results.
Apple is not actively sabotaging Android devices by offering subsidies to Android Tablet and Phone manufacturers.
Apple is not actively sabotaging Android devices by offering subsidies to Android Tablet and Phone manufacturers.
Until then, I'll survive with the 800 on my Lumia 1520.
Actually, to blackmagnum, Qualcomm is being rumored to pay companies to include their processors, not sure if they are paying to exclude the competition, but it sounds similar to what Intel did to cut-off AMD on a lot.
Found an article: http://www.zdnet.com/article/eu-to-qualcomm-youve-been-paying-to-shut-out-rival-chip-makers/
I believe he meant it in terms of performance.
Apple also doesn't allow anyone other than Apple to use their CPU... exactly UNLIKE Intel, who lets anyone use their chips in any device you can dream up.
Actual splitting the Kryo cores in two clusters whose a dumb move.
They will certainly get a performance edge in real usage where 2 cores are mostly sufficient enough (Web, apps) but in cases where you need more it will even be under the S620 so that's shooting in their own feat.
I am not impressed with Spectra or QDSP but I am impressed with GPU performance.
So in summary if they don't blow with GPU on S620 along with pricing they culd stay in the game as S620 could be a wonderful mainstream SoC all well balanced otherwise it will be a game over for them.
I won't be crying for them as I didn't for Texas Instruments or Broadcom that they pushed out (that they will finally pay for) but truth is if they go down we are all without any open stock (I don't count Exunos as one more OEM ever used them).
So true! Is the first time when i read a good comment and a very real. This is a question that many people should think about... Are we use our smartphones for playing video games or for the real communications and network connection? What do people want from a smartphone? You expecting to run AutoCAD on your smartphone, or 2 days without recharge and full network?
The core config strikes me as odd. big.LITTLE config makes more sense when the little cores are actually *little*. If the cores are going to be the same, then I think they should have just run all 4 cores at variable clock speed w/ the same ceiling (and globally limited by a power/thermal management scheme like Intel's Turbo Boost). Each core should have its own L2 cache and flush it when the core idles. Then, you could turn off that core's L2 and it burns virtually no power. The benefit would be better performance on tasks using > 2 cores. With those power savings in hand, they could afford more total L2 cache, to offset the reduction in available cache for the 1-2 core case.
But I assume Qualcomm simulated many use cases and chose this one to balance performance, battery life, and cost.
No, but Intel does have a history of pressuring companies to only use their chips instead of including AMD's
Anyway, if you don't need the performance, nobody is forcing you to buy such a high-end SoC.