     |
| |
Samsung drew criticism for inflating the benchmark scores of Exynos devices earlier in the year, but the company appears undaunted; it's reportedly boosting test numbers for other hardware as well. Ars Technica has discovered that the Snapdragon 800-based Galaxy Note 3 (and possibly the new Note 10.1) includes code that runs all CPU cores at full speed during certain benchmarks. The tweak gives the smartphone a minimum 20 percent higher score in any affected app, or enough to claim an artificially large advantage over an LG G2 using a similar chip. There may also be a graphics boost, Ars says. We've asked Samsung for its take on the findings. Whether or not the company responds, we don't envy its position -- it's hard to form an alliance devoted to accurate mobile benchmarking when you're accused of doctoring results.
Filed under: Cellphones, Mobile, Samsung
Source: Ars Technica
It's called MobileBench: an industry consortium planning to offer "more effective" performance assessments on mobile devices -- most likely centered on, but not limited to, Android. Unsurprisingly after recent developments, Samsung joins as a founding member, alongside Broadcom, Huawei, Oppo, and Spreadtrum. While that's who's in, who isn't? Well, both NVIDIA (responsible for the Tegra series of mobile chips) and the increasingly ubiquitous Qualcomm, which makes the Snapdragon mobile processor range. Between them, they power the likes of Microsoft's Surface series, Amazon's new Kindle Fire range, not to mention numerous flagship devices from LG, Samsung, Sony and Motorola.
The group gathered for the first time yesterday in Shenzhen, China and outlined how it aims to offer more useful tools for mobile platform designers and "more reliable indices" for assessing user experience. MobileBench plans to establish impartial guidelines and a more sophisticated evaluation methodology for both its first benchmark tool, MobileBench and MobileBench-UX, for testing system-level applications. The benchmarking tool will assess hardware performance, including high-level processes like video and image viewing, camera use and other real-life use cases, with one of the primary aims being result consistency and less deviation between repeated tests. Another app is planned for consumer use in the future, likely similar to the benchmarking apps Engadget uses in its reviews. The bigger question is how much the consortium can achieve without wider adoption inside the industry -- it's apparently "actively seeking" more members.
Source: MobileBench consortium (PDF)
Overclocking may yield impressive benchmark results, but it turns out scores from Windows 8 PCs may not be reliable. The management at overclocking community HWBOT has discovered that tests provide inaccurate stats when then CPU base clock frequency is fiddled with from within the OS. Hardware-based real-time clocks (RTCs) help keep accurate track of time, but the operating system's timekeeping somehow slows down or ramps up when processing speeds are tweaked. When underclocked by six percent, the outfit's Haswell-infused system lagged 18 seconds behind actual time, fooling the benchmark into a higher score since it seemingly finished in a shorter period of time. Conversely, a boost to CPU speeds results in a lower mark as the internal timepiece ticks away faster than usual. However, modifying processor speeds at boot time avoids these issues.
As a result of the revelation, HWBOT is no longer accepting benchmarks from computers running the eighth iteration of Ballmer and Co.'s software, and will invalidate those already in its database. "Simply no benchmark - not even 3DMark - is unaffected by Microsoft's RTC design decisions," the outlet adds. The timing issues are said to stem from Windows 8's support of disparate hardware setups, including embedded and budget PCs that don't have a fixed RTC. If you'd like to see the inconsistencies for yourself, head past the break for video proof.
Filed under: Software, Microsoft
Via: ExtremeTech
Source: HWBOT
While benchmark leaks can be wobbly info-stones to tread upon, the arrival of some purported Kindle Fire HD scores could be the first steps toward a timely refresh. The results in question refer to a "Amazon KFAPWA" device with a reported Adreno 330 GPU, which by implication would likely mean a Snapdragon 800 SoC. The field that really leaps out, however, is that 2,560 x 1,600 resolution. If true, it could mean the 8.9-inch version is getting a bump from the previous iteration's 1,920 x 1,200 display. This certainly chimes with earlier reports, and, if nothing else, would certainly make that new browser UI all the more pretty to look at.
Via: Android Community
Source: GFXBench
