|
|
|
|
Test Changes
- New DirectX 12 test that runs at a default resolution of 4K (3840x2160). In cases where 4K resolution isn't available the score will be scaled down and the test will run at the highest available resolution.
- Removed simple DirectX 9 test from 3D test suite. This was removed as DX9 is now fairly old and the simple test wasn't putting much load on newer video cards.
- CPU extended instructions test will perform testing using sub-tests for FMA, AVX and SSE (or only those that are supported) and take the average of the 3 (or of those that are supported) for the benchmark result. FMA and AVX are new CPU instruction sets that have become available over the last few years in Intel and AMD CPUs. They were designed to make certain mathematical operations faster. Individual sub-test duration has also be adjusted to account the new sub-tests.
- CPU Physics Test and the Advanced Physics Test have been switched to use the current 'Bullet' physics engine. In the past we were using the Tokamak engine, but Tokamak is no longer maintained. While at the moment GPU acceleration isn't stable in Bullet, we hope to enable GPU acceleration in Bullet in a future patch (limited to integrated GPUs).
- 2D Graphics, added a Fonts and Text test using Direct2D. This test will launch after the original Font and Text test complete on systems that support Direct2D (Vista and higher). Direct 2D is used by most browser and some other applications for rendering text and web pages. So the new Fonts and Text test result will now be a combination of Direct2D performance and GDI performance. GDI being the traditional method if displaying text and graphics in Windows. While Microsoft advertised Direct2D as being faster than GDI, we found the opposite in most cases. Win10 did bring some Direct2D performance improvements however.
- DirectCompute (GPU Compute) tests will now run for longer based on the test duration configuration setting. The bitonic test will now run for a length of time corresponding to the duration setting, previously it was set to 100 iterations.
- Added an OpenCL test to the GPU Compute set of tests
- GPU Compute bitonic, standard test will launch two instances of bitonic test is there are 2 identical cards in the system (not running in SLI/XFire mode)
- Advanced Disk Test
- Added sampling of individual latency results and added support for latency results plotted on the graph as a heat map.
- New Drive IOPS Test available from "Advanced->Disk->Drive Performance..." menu option.
- Advanced Memory Test
- NUMA is now supported where available. Settings are displayed when selecting results to graph. Exported results and graph will now include NUMA settings.
- Added DX11 Oculus Rift VR test, as an advanced test. This is currently a bit experimental and it has only had limited testing. The test is based on the existing DX11 test (the Lunar landscape test), but has been redesigned slightly to run on an Oculus Rift.
- Advanced 3D Test
- Changed how results were displayed and will now display frame latency times. Note: Frame latency is measured from the applications point of view, so doesn't tell the whole story, as the video card can buffer or discard frames.
- Added button for easy copy of results to clipboard
- Advanced Direct Compute (GPU Compute)
- Renamed to GPU Compute
- Added an OpenCL test
- Added option for bitonic test to run on all available cards. Previously the test would only run on a single card.
- Advanced network test, added internet and transfer protocols used to test report output. Also changed the maximum block size of 256KB and set the default to 32KB. The larger block sizes result in less overhead and higher throughput.
- Advanced Visualized Physics Test
- Now uses Bullet physics engine (see above)
- Added a new collection of preset tests, based on the capabilities of the Bullet engine. Ragdoll, compressible objects, friction, rope simulation, etc..
- Switched to using OpenGL3 (instead of DirectX) for rendering the scene.
- Changed max number of shapes to be 30000 (limited by Bullet Physics)
- Removed Overall FPS display (same as render FPS due to multithreading)
- Made some changes so GPU Compute tests are now DPI aware and should no longer automatically scale on high DPI monitors
User interface changes
- Totally new look user interface, redesigned with new charts for displaying benchmark data, your result vs the world (using all submitted baselines) and your result vs same hardware model (CPU/Video/RAM/Disk).
- Added User editable string file allowing for customisation and translation of strings used in PerformanceTest (pt_localization.txt). Nearly all menus and headings can be customised/translated by editing the text file. This will allow future localization of the software into French, Chinese, etc..
- Added option to enable/disable heat map plotting in Customization dialog (in the advanced tests)
- Added real-time temperature collection and display to main 3D window display
- In system information window physical disks are now displayed in disk number order (as opposed to the SMBIOS order)
- Added baseline drag and drop loading to main baseline information window, baselines can be dragged form windows explorer into PerformanceTest for loading.
- Added right-click menu to benchmark graphs for exporting results to clipboard/file
- Removed "Show numeric results column" from Preferences
- Temperatures are now updated after corresponding test suite has completed (eg. CPU temperatures updated after running CPU test suite, etc)
- When RAM manufacturer/part number from the SPD data is not available, use SMBIOS details instead
- Added mouse over tooltips for advanced memory graphs
- Added DPI scaling for 2D font test. Previously the test window would be tiny on high DPI displays.
- Replaced 'Upload baseline' button with checkbox that is checked by default. Checkbox is enabled only if full baseline is available and baseline has not already been uploaded.
- Replaced some icons with higher quality ones
- When running a single test a reduced size status window is now displayed
- Removed Vsync options in main preferences, will now always default to off. Script command will remain and allow it to be changed when running a script.
- Added German, French, Chinese (simplified and traditional) translation
Misc
- Scripting, added REPORTSUMMARYHTML script keyword to output report in HTML format when using the LOOPS command
- Scripting, added optional parameter SINGLELINE to commands EXPORTCSV, EXPORTTEXT for outputting the computer's system information and results to one line
- Added shortened operations descriptions to CSH/HTML/TXT exports (eg M/ops, Mb/s, Fps, etc)
- Added error message when the number of selected baselines cannot fit on one page when printing
- Added error message when saving results to image fails due to the image file height being too large (eg too many selected baselines)
- Added a warning message when submitting a baseline if different video card names were detected as being tested. This is to alert the user and try to prevent situations in systems with low power/high power video cards using the low power card in some tests
- Support for new CPUs and better identification of AMD graphics card (some AMD card have never identified themselves correctly and only refer to themselves as 'Series 200' or the like).
- Added export function to export results and graphs to HTML file from drive performance
- Changed an error message when then DirectX12 test could not be launched
- Added a delay before launching 3D tests when running a script and using their single test command to prevent monitor resolution switching occurring too quickly and causing errors
- Scripts will now exit and no longer continue to run if an error occurs during the SETDISK command
- Changed the behaviour of the Dx9 and DX10 when run from a script, previously they would run for much longer than the duration used for SETDURATION
- Added network drives to Advanced Disk -> Drive Performance disk selection
- Changed scripted CSV output to always print test names and results, if no result recorded then output N/A, to allow for easier combination of results from multiple machines
- Added option to 'Clear chart cache' in Preferences window. All .xml files under ProgramData\Passmark\PerformanceTest9\ChartData\ are deleted.
- Added a check for Nvidia G-Sync, if detected (and if result seems to be capped at a framerate between 54 and 60) will block basline submission until G-Sync is turned off and 3D tests re-run
- Changed how script files are processed to better handle the loop command
Bug Fixes
- Fixed some places where files were saved to the local user directories instead of the executable directory when running with command line parameter /p (eg not saving to USB drive as expected)
- Fixed a bug in the memory read cached test where the buffer size used was in steps from 16KB to 148KB instead of 16KB to 8MB. Test results in this test will now be slightly lower than they were as a result.
- Fixed baseline search in baseline manager where searches of turbo speed 5001MHz+ were not working correctly
- Fixed a bug in the Advanced Direct Compute bitonic test where the displayed results were being calculated incorrectly
- Added some checks to try and avoid a situation where a graphics driver crash could cause an abnormally high Direct2D result
- Fixed a crash when running Direct Compute test
- Now closing DirectIO.sys library when PT exists to stop the .sys file being in use until automatically closed. This should avoid an error some people saw on installation that the DirectIO.sys was locked and couldn't be overwritten.
- Fixed incorrect values for average speed + data transferred when exporting results from the advanced network test
- Fixed crash in advanced network test if test was started without server being available.
- Fixed a bug where the ADT_SETEXPORT graph settings were not being read correctly and the graph type was not changing
- Fixed a bug where DX11 test would often pick a low refresh rate (eg 24hz)
- Fixed a bug where the "raw disk read" options for the advanced CD/DVD test was unavailable in windows versions newer than XP
- Fixed some CSV output so "," characters in CPU names no longer break the formatting
Test Score Scaling
- Added 6.5% scale down of CPU Mark so PT9 results are closer to PT8 results, to compensate for the increased scores in the Extended Instructions and Physics test changes.
- So the PT8 CPUMark should be broadly comparable to the PT9 CPUMark. Same for the overall PassMark rating. Still, when possible, for improved accuracy, compare PT9 results to PT9 results and PT8 results to PT8 results.
- DX12 test score is penalised if it cannot be run at the desired 4K (3840x2160) resolution. Running at lower resolution increases the frame rate. So the scaling compensates for this.
- The weighting of the 3D tests was changed due to the removal of the DX9 simple test and the addition of the DX12 test. The DX9 (complex) test is now weighted lower than the other tests (approx. 25% less). Systems that cannot run the DX12 test will received a 20% penalty to the overall 3D mark score.
|
|
|
|
|