07-13-2015 03:20 PM - last edited on 03-06-2024 02:31 AM by ROGBot
07-13-2015 05:19 PM
07-13-2015 06:22 PM
Korth wrote:
Not a lot of detail in your questions. What are you benchmarking? With what hardware and what software?
Benchmarks don't always accurately represent useful "real world" performance, especially when they attempt to combine multiple metrics into some sort of overall performance score. At the worst end you have things like Microsoft's Windows Experience Index, perhaps useful for comparing devices side-by-side on a store display shelf but otherwise only tenuously connected to actual performance.
Better data can be obtained by benching specific components (processor performance, memory bandwidth, drive R/W speeds, etc) but it's only useful if you understand what you're measuring and why it's meaningful. 3D games tend to push hardware extremes - performance leans most heavily on the GPU card(s) but also depends on processor, memory, drives, and other hardware performances - different 3D/game engines will output different fps scores, although most tend to go a little light on stressing physics/particles/complexity because they want to advertise smooth playability more than crazy eye-candy.
Any benchmark can produce minor variances on repeated runs. Complex code and hardware can behave a little differently each time it's run, the benchmark tool itself may use some arbitrary level of precision, astronomically large numbers of samples can be measured and condensed into final results so it's almost more of a statistical thing than a hard speedometer.
Stress tests attempt to push your hardware (or specific hardware components) to 100% and beyond. To simulate absolute worst-case heaviest loading, to strain your performance and power and cooling systems to their max, to confirm no weak failing parts are present. Many stress tests also serve as benchmarks. But again, these typically don't accurately represent useful "real world" performance.
07-13-2015 06:54 PM
07-13-2015 11:25 PM
07-14-2015 08:23 AM