Today, we had an opportunity to play with MDP development smartphones and tablets, and we jumped on the chance to run some performance numbers. Since there’s no “design” or device “body” to review, all our attention went to “how fast” sub-components are.
It should be reminded that none of the synthetic benchmarks will provide a proper picture of the final “user experience”, but it is clear that there is a strong correlation between how powerful computers are and how good of an experience they can enable. The raw numbers below, but first here’s a quick overview of the specs.:
Inside, there are four Krait 400 CPU cores (2.3GHZ max) that work alongside an Adreno 330 GPU and and Hexagon QDSP6 DSP. Instead of using small cores (BIG.Little style), Qualcomm has chosen to use a dynamic CPU power control that lets it tune the power consumption/performance ratio in an finer fashion than core-switching would allow. This and the LPDDR3 memory are the main sub-systems that are going to be stressed in typical benchmarks.
Geekbench 2 is a math-oriented benchmark is that used to show how fast CPU cores can solve math problems. This is really a crunching-number game that does not really mimic real-world applications, but it is usually a decent indicator of how fast CPU cores are.
Interestingly, the Snapdragon 800 and NVIDIA’s Tegra 4 are running neck to neck. Although the numbers we got show a tiny advantage for the Tegra 4 reference platform, the difference is within the typical margin of error, so we’ll call this one even. It can be surprising that Qualcomm scores so well against ARM’s mighty A15 design, and this is a testament to how good Qualcomm at designing CPU cores. We would love to see which architecture consumes more power at peak performance, but this remains uncertain for now.
Antutu 3.x is a system benchmark that tries to assign an overall score to the whole machine. It does this by running a number of sub-benchmarks and computes a global score. Here, the speed of the storage and memory sub-systems do play an important role, so it is critical to test this again on actual products when they come out. In the battle of reference designs, Tegra 4 comes out slightly on top in the global score and cracks the 35000 points, which is impressive.
Next, we’re testing the graphics capabilities of the Snapdragon 800, and just like Qualcomm predicted, it offers about 50% more performance than today’s fastest smartphones (!). Many will be surprised that Snapdragon 800 scores a clear win against NVIDIA’s Tegra 4 in GLBenchmark 2.5 and I suspect that it is a slightly better triangle setup rate, but we will revisit that when we get our hands on a production unit of SHIELD. We didn’t have a Tegra 4 device on hand to test BaseMark X, but it is probable the outcome would be comparable. In terms of graphics speed, Snapdragon 800 takes the lead in those tests.
Qualcomm had the Snapdragon 800 design up-and-running for a while, but the company wanted to have all its software development tools and SDKs ready before it starts to reveal the raw performance numbers. Qualcomm points out that it’s the “whole system” that counts and it many ways, they are completely right. Still, raw synthetic numbers show us what the “speed of light” is and hopefully software developers will be able to extract all that power, at least, that’s what every one of us expect from the apps we buy.
When actual devices will come out, it looks like Snapdragon 800 will fare very well performance-wise -especially for gaming-, but the most important parts is that it comes with an integrated 4G LTE modem that has been qualified by the biggest carriers, and this is an undeniable business advantage over the competition. If you have been waiting for Q3/Q4 to upgrade your smartphone, you can smile, because this is very good…