System Performance

The P8 Lite is the first device with Qualcomm's Snapdragon 615 that we've reviewed. The Snapdragon 615 implementation in the P8 lite has two clusters with 4 Cortex A53 cores. With this bin the first cluster has a peak frequency of 1.5GHz, while the second has a peak frequency of 1.11GHz. This is slightly slower than the 1.7GHz implementation in the HTC One M8s, but it still represents a 25% improvement in clock speeds on the big cluster over Snapdragon 410. In theory since there's support for HMP the OS could schedule threads to use all 8 cores at once, but in essentially every user workload it will act as a quad core Cortex A53 CPU.

Kraken 1.1 (Chrome/Safari/IE)

Google Octane v2  (Chrome/Safari/IE)

WebXPRT 2013 (Chrome/Safari/IE)

WebXPRT 2015 (Chrome/Safari/IE)

The P8 Lite shows good improvement over devices that use Qualcomm's Snapdragon 400 and 410 SoCs. In both versions of WebXPRT and in Mozilla Kraken we see improvements of about 20%, with Google Octane improving by over 30%. This is in line with what one would expect from the 25% clock speed improvement on the big cluster when compared to Snapdragon 410.

PCMark - Web Browsing

PCMark - Video Playback

PCMark - Writing

PCMark - Photo Editing

PCMark - Work Performance Overall

PCMark is a benchmark that focuses more on real-world scenarios where race to sleep speed is paramount, but the user workload also increases with faster CPUs. Software can also be as important to the test as hardware, and we've seen gains on devices running Android Lollipop over those running KitKat. In it we see that the Huawei P8 Lite doesn't do very well overall. The main cause of this is its score in the writing sub test. This is not a new situation for Huawei, as all of their KitKat devices score somewhere around 500 points in that test due to problems with Dalvik. If that test were to be removed, the P8 Lite would perform slightly better overall than the Moto E. Ultimately the result in PCMark is just an example of how performance is hardly software agnostic, and demonstrates why devices should not be launching with KitKat this far into 2015.

Basemark OS II 2.0 - System

Basemark OS II 2.0 - Memory

Basemark OS II 2.0 - Graphics

Basemark OS II 2.0 - Web

Basemark OS II 2.0 - Overall

In BaseMark OS II we again see the P8 Lite holding a lead over Snapdragon 400 and 410 devices, with the largest improvement being in the graphics sub test.

Overall the P8 Lite shows significant improvements over the performance of lower end Snapdragon 400 and 410 based devices. CPU improvements are right in line with the 25% greater clock speed than Snapdragon 410 on the big cluster of Cortex A53 cores, and BaseMark OS II shows us that there are substantial GPU improvements which will be detailed on the next page.

Introduction, Design, and Software System Performance Cont'd: GPU and NAND
POST A COMMENT

45 Comments

View All Comments

  • Buk Lau - Wednesday, July 29, 2015 - link

    It is capable of QC 2.0, they just don't ship you with a charger and you just have to get it yourself from somewhere Reply
  • zebrax2 - Monday, July 27, 2015 - link

    Error in the specifications table showing A7 instead of A53. Reply
  • jjj - Monday, July 27, 2015 - link

    Disturbing that for a SoC known for it's overheating capabilities (yes this one too), you fully ignore the issue and don't look for throttling.

    "Snapdragon 615 is really the best you're going to get at this price point"
    That is clearly a false statement and not sure how can you make it. SD615 is likely the worst A53 based solution in this price range clock for clock and not just in CPU. If you factor in that others are going higher clocks (without overheating) in the same price band and even higher clocks for slightly costlier and more capable SoCs , your statement becomes even more absurd.
    But that's when you factor in the SoC price not the device price. If you go by device price, the OnePlus One is 250$ and if you factor in China, a bunch of faster devices are at or very close that price.
    Reply
  • T1beriu - Monday, July 27, 2015 - link

    Source? Reply
  • jjj - Monday, July 27, 2015 - link

    For what? SD615 can't do 1.7GHz and pretty much any device with those clocks is know to have problems. At 1.5GHz it throttles little if at all but that can depend on the device and requires investigation. As for how it does against competing 8xA53 SoCs from Hisilicon (hey those are actually worse), Marvell, Mediatek and Samsung it would take some time(that i don't have) to find a bunch of reasonably good links since too many can't be bothered to look beyond Qualcomm, just like AT loves to do. Reply
  • djw39 - Monday, July 27, 2015 - link

    How can you say the Snapdragon 615 is "likely the worst A53 based solution in this price range" then? Reply
  • LiverpoolFC5903 - Tuesday, July 28, 2015 - link

    The MTK 6752 is significantly faster than the SD 615 and runs cooler too, despite all the ores being clocked at a relatively high 1.7ghz

    The 615 flatters to deceive and is hamstrung by memory bandwith issues and thermals. If they had provided a dual channel 32 bit RAM like the older Snapdragon 600 and paired it with the Adreno 320, it would have been a killer mid range chip. Alas,this is par for the course these days with Qualcom chipsets.

    One more observation, the Zenfone 2 is absolutely suberb in terms of performance/price/ Nothing else in the midrange comes close to it in terms of both CPU and GPU performance. The 2.3 GHZ and 4 GB model at 300 USD performs as good as any Snapdragon 801 device and trades blows with the 805.
    Reply
  • AndrewJacksonZA - Tuesday, July 28, 2015 - link

    Totally off topic, so apologies in advance, but...

    "T1beriu" as your nick? Is it meant to be in reference to the Tiber River? Tiberius Caesar? Tiberium? Brigadier General Mark Jamison Sheppard must know! :-)
    Reply
  • twizzlebizzle22 - Monday, July 27, 2015 - link

    What's the benefit of using big little with the same core architecture between the clusters.

    Can't one cluster scale back under lighter workloads?
    Reply
  • cmikeh2 - Monday, July 27, 2015 - link

    You can fab the transistors for the two different clusters differently. For the high performance cluster, you would use a higher leakage, faster transistor (and power gate when not in use) and for the low power cluster use slower, but lower leakage transistors. Reply

Log in

Don't have an account? Sign up now