Separate Windows installs are used for each chipset-i.e. The newest Windows version at time of writing was used, and we will update as necessary for major changes. NVIDIA driver version 445.75 was used for all of today’s tests on the EVGA RTX 2080 Ti XC Ultra.Īll tests are run with the Windows High Performance power plan selected or the AMD Ryzen High Performance power plan. Starter CPUs include, but aren’t limited to: Again, rather than asking “WhY iSn’T MY CpU oN HeRe?”, remember that the goal today is an outlay of the methodology. We’ve found them, at this point, but we’re also going to publicly present a few of those shortcomings (and our solutions to fix them). Our selection of CPUs here was to cast a wide net to try and locate shortcomings in the testing methodology. In other words, we know not every CPU is on the chart yet, but this isn’t a review. We’ve been hard at work running our inventory of CPUs through this new test suite and recording the results, and that work is ongoing. HWiNFO64 (frequency, thermal validation).CPU-Z Validation report (internal use only).Chromium Windows build (using clang-cl).Here’s the suite (graphics settings noted in each section): GAMES: We try to limit this, but time often gets the better of us, and it’s good to have reference points to ensure ongoing accuracy. a 2600K, but didn’t have a fresh retest, we could reference data from GTA V, Tomb Raider, Civilization, and ACO to form an understanding without fully retesting. For example, if we were testing a 10700K and wanted to reference its performance vs. This is to keep producing data that we can still compare to old data, which is useful for rapid analysis of parts that may not have been re-tested in the current year. Gaming gets a total overhaul, too, with a big suite of new games added.Īdditionally, we’ve updated several existing game and production benchmarks from last year’s suite, with a few left unchanged. New code compile benchmarking (with more usefulness) has been added for 2020, alongside the addition of Handbrake H.264 to H.265 transcoding (ranked by time), updated Adobe Premiere video rendering and Adobe Photoshop benchmarks, updated file compression and decompression benchmarks, and more. We ditched that test (and consulted Wendell of Level1 Techs on it in this video), and we’re just now replacing it. The test featured GCC, Cygwin, some other environments, and ended up being a top-to-bottom sort by cache. Last time we updated our testing methodology, we added a code compile benchmark that was short-lived. We’ve done some significant streamlining behind the scenes that make these tests easier to run and the results easier and more accurate to process, but on the public side, we’ve completely overhauled the software suite we’re using. It’s time again for our CPU testing methodology to be updated, alongside the test bench.
0 Comments
Leave a Reply. |