CPU Performance

The Galaxy S 5 marks the second Snapdragon 801 based device we've reviewed at AnandTech, the first being HTC's M8. I've gone through the Snapdragon 801 in depth already, but we're basically dealing with a reasonable upgrade to Snapdragon 800 on an improved 28nm HPm process. The bulk of the improvements impact GPU and ISP performance, but the SoC is just better overall. GS5 owners are lucky as all versions of the device that use Qualcomm silicon feature the MSM8974AC v3 SKU, which includes four 2.5GHz Krait 400 cores and a 578MHz Adreno 330 GPU.

Snapdragon 800/801 Breakdown
  SoC Version Model Max CPU Frequency Max GPU Frequency ISP eMMC DSDA Memory IF
MSM8974VV v2 S800 2.2GHz 450MHz 320MHz 4.5 N 800MHz
MSM8974AA v2 S800 2.3GHz 450MHz 320MHz 4.5 N 800MHz
MSM8974AB v2 S800 2.3GHz 550MHz 320MHz 4.5 N 933MHz
MSM8974AA v3 S801 2.3GHz 450MHz 320MHz 5.0 Y 800MHz
MSM8974AB v3 S801 2.3GHz 578MHz 465MHz 5.0 Y 933MHz
MSM8974AC v3 S801 2.5GHz 578MHz 465MHz 5.0 Y 933MHz

Although Samsung was the first major OEM to be caught cheating in Android benchmarks, it appears to have completely abandoned the practice with the Galaxy S 5's shipping software. Not only was I unable to find any evidence of the old cheats, I couldn't find any evidence of HTC's new subtle cheating either. The Galaxy S 5 appears to be clean as far as I can tell. Kudos to Samsung on doing the right thing, and I hope all other OEMs take this as a sign to stop the silliness.

For our performance tests I turned to our usual suite of browser and native applications. If there's one obvious takeaway from our CPU tests it's that despite having faster silicon than HTC's M8, the GS5 isn't always faster. I believe this has more to do with thermals than anything else. HTC's metal chassis is able to do a better job of dissipating heat than the GS5's plastic chassis. I don't believe there's a substantial impact on user experience, but it's interesting to note how choice in materials can have a performance impact like this.

SunSpider 1.0.2 Benchmark  (Chrome/Safari/IE)

Kraken 1.1 (Chrome/Safari/IE)

Google Octane v2  (Chrome/Safari/IE)

WebXPRT (Chrome/Safari/IE)

BaseMark OS II - Overall

BaseMark OS II - System

BaseMark OS II - Graphics

BaseMark OS II - Web

GPU Performance

GPU performance remains where we see the biggest benefit from Snapdragon 801 vs. 800, and since the GPU gains are almost entirely due to frequency scaling it's not too surprising that the M8 pulls ahead of the GS5 here in most cases.

There aren't any surprises here. The Adreno 330 in the Galaxy S 5 is more than capable of driving the device's 1080p display both in current and near term future 3D games.

3DMark 1.2 Unlimited - Overall

3DMark 1.2 Unlimited - Graphics

3DMark 1.2 Unlimited - Physics

BaseMark X 1.1

BaseMark X 1.1 - Overall (Medium)

BaseMark X 1.1 - Overall (High Quality)

BaseMark X 1.1 - Dunes (Medium, Offscreen)

BaseMark X 1.1 - Hangar (Medium, Offscreen)

BaseMark X 1.1 - Dunes (Medium, Onscreen)

BaseMark X 1.1 - Hangar (Medium, Onscreen)

BaseMark X 1.1 - Dunes (High Quality, Offscreen)

BaseMark X 1.1 - Hangar (High Quality, Offscreen)

BaseMark X 1.1 - Dunes (High Quality, Onscreen)

BaseMark X 1.1 - Hangar (High Quality, Onscreen)

GFXBench 3.0

GFXBench 3.0 Manhattan (Onscreen)

GFXBench 3.0 Manhattan (Offscreen)

GFXBench 3.0 T-Rex HD (Onscreen)

GFXBench 3.0 T-Rex HD (Offscreen)

GFXBench 3.0 ALU Test (Onscreen)

GFXBench 3.0 ALU Test (Offscreen)

GFXBench 3.0 Alpha Blending Test (Onscreen)

GFXBench 3.0 Alpha Blending Test (Offscreen)

GFXBench 3.0 Driver Overhead Test (Offscreen)

GFXBench 3.0 Driver Overhead Test (Onscreen)

GFXBench 3.0 Fill Rate Test (Offscreen)

GFXBench 3.0 Fill Rate Test (Onscreen)

GFXBench 3.0 Quality/Accuracy Test (High Precision)

GFXBench 3.0 Quality/Accuracy Test (Medium Precision)

NAND Performance

The GS5 ships with 16GB or 32GB of NAND internally on an integrated eMMC device. Expansion is supported through a microSD card slot behind the removable back cover. Although the Snapdragon 801 inside supports eMMC 5.0, that alone doesn't guarantee a substantial increase in NAND performance. Keep in mind that most OEMs find multiple sources for their internal eMMC/NAND solutions, so what I'm testing here may only be representative of a portion of all GS5 devices.

Samsung sampled a 16GB GS5 review device. I put it through our usual random/sequential IO tests on a 100MB span of LBAs.

Random read performance is disappointing, it falls behind all modern devices we've tested. Random write performance is middle-of-the-road at best. It's unclear to me if this is a cost optimization or a lack of concern for NAND performance, but either way I'd rather see these metrics improve rather than regress.

 

Internal NAND - Random Read

Internal NAND - Random Write

Sequential read/write performance both improve handsomely compared to the Galaxy S 4. I can see why Samsung would want to optimize for these two cases as they are quite common in regular usage, but random read/write performance can also significantly impact user experience.

Internal NAND - Sequential Read

Internal NAND - Sequential Write

Display WiFi, Cellular & Download Booster
Comments Locked

296 Comments

View All Comments

  • tipoo - Tuesday, April 8, 2014 - link

    They certainly aren't measly cores. Extremely wide for mobile.
  • tipoo - Tuesday, April 8, 2014 - link

    I believe Nvidia will be following this concept next, two big cores rather than four low power small ones.
  • Streamlined - Wednesday, April 9, 2014 - link

    I was shocked that last years iphone still seemed to outperform Samsungs latest. Holy crap!
  • twotwotwo - Tuesday, April 8, 2014 - link

    On the camera page, no mentions of the G2, which, to me, seems to produce better-looking output in the low-light shots than the rest (though, for all I know, that might be mostly a difference in postprocessing)? What's up--was that added after the text was written, or am I an outlier for thinking the G2's shots look nice?
  • coburn_c - Tuesday, April 8, 2014 - link

    That OIS really makes the difference in low light. The stock camera app actually does stupid things in low light, XDA camera apps make it even better.
  • peterfares - Tuesday, April 8, 2014 - link

    Still PenTile. Yuck.
  • pppp6071 - Tuesday, April 8, 2014 - link

    I guess the review deliberately avoided PenTile RGBG issue. Its just cheating in my view. LG G2 and Nexus 5 and HTC One M7,8 uses RGB pixels.
  • tipoo - Tuesday, April 8, 2014 - link

    Who cares anymore when in every measurable aspect it's a great display?
  • melgross - Tuesday, April 8, 2014 - link

    Not great by any means, but better than before.
  • kaelynthedove78 - Tuesday, April 8, 2014 - link

    For starters, it means that the resolution is not actually 1920x1080, but 1493x840.
    1,25 megapixels instead of 2, so in reality it's closer to a 720p RGB display than a 1080p one.

Log in

Don't have an account? Sign up now