We love transparency, so you will find all major changes for BenchMate on this page.

Features & Improvements

  • Disabled CPU-Z update check because it simply takes too long for benching (and it is updated with BM anyways)
  • Created a manual install package for Windows 7 and 8. Check the front page for the download link.
  • Updated tools: HWiNFO 7.64, CPU-Z 2.0.8, GPU-Z 2.55


  • Reintroduce support for CPUs missing the RDTSCP instruction. That's every platform older than Intel Nehalem and AMD K8. Thanks to TheQuentincc.


  • The installer will only support installations on Windows 10+ from now on. Use the manual install package for Windows 7 and 8.

Features & Improvements

  • PYPrime 32b can now be submitted to HWBOT
  • root certificate for submission upload is now automatically installed on Windows 7 (otherwise you can start BM\res\certificates\install-hwbot-run-as-admin.bat to install the certificate manually)


  • y-cruncher workloads can now be correctly selected from CLI wrapper options
  • Sandy Bridge was not detected correctly in the result window
  • Installer on Windows 7 should correctly install the SHA2 prerequisite now

Features & Improvements

  • Bundled y-cruncher 0.7.10 because it is sometimes the faster version for older platforms


  • Fixed y-cruncher validation for new problem sizes: 5B and 25B


    • You can finally bench AMD RX 6x00 series graphics cards and hopefully RX 7x00 cards as well. The Reduction Size needs to be set to 256, because these cards have limited shared memory available.
    • GPUPI for CPU now works with Intel OpenCL 3.0 runtime. It's the latest runtime that can be downloaded without registration:
    • Added multiple tweak options:
      • OpenCL Math Optimizations: Disabled, Safe, Unsafe, Fast - if you encounter a precision error, the OpenCL runtime does not work with that level of optimization.
      • Intel OpenCL Vectorizer: Enabled (Default), Disabled - gives you control over the automatic loop vectorization for CPUs (you will be surprised that Intel's default is actually slower)
      • Intel OpenCL Vectorizer Mode - fine-grained control of the width for the vectorization, if enabled
      • Intel OpenCL Instruction Set: Auto (Default), SSE4.2, AVX, AVX2, AVX512 - Auto will set the highest possible set, be sure to select Auto or at least AVX if you have enabled the Vectorizer
      • Intel OpenCL Loop Unroll Factor - higher value might be better
    • BenchMate now hooks the underlying Intel Embree raytracing framework for better error detection. If the raytracing workload reports an error, the run will correctly be flagged as invalid. This fixes the long-time CB bug that shows very high scores when the workload was not completely rendered due to an error (mostly because of a maxmem setting).
    • Added new options to apply a selected Instruction Set Architecture (ISA) and the Frequency Level. This allows to make use of a specific instruction set like AVX, SSE 4.2 etc.
    • BenchMate's internal run measurement now starts closer to the workload for better average measurements (ie: wattage)
    • Added an option for a run confirmation dialog like in Super PI
    • Improved anti-tampering support by verifying all loaded plugin dlls
    • Added AMD Unsafe tweaks for Vermeer
  • y-cruncher
    • Currently supported version are: 0.7.10, 0.8.1, 0.8.2 (bundled)
    • Added thread count to CLI wrapper options: All Threads (BenchMate), All Threads (y-cruncher), 1 Thread. It defaults to BenchMate's thread count detection to utilize all threads. This automatically fixes a y-cruncher bug that detects 48 threads on 7900X resulting in slower runs. This might affect other AM5 CPUs as well (not the 7950X though).
    • Added new problem sizes: Pi-5b, Pi-25b
    • Improved HWBOT submission so the y-cruncher Java Submitter and BenchMate can both be used to submit scores of all problem sizes. Attention: The y-cruncher 2.5b submission had to be changed, so be sure to submit all your old 2.5b results as soon as possible using BM 10.
    • Experimental bugfix for several warnings that need user input to continue
    • Correct application of priority class setting (it was previously at Normal-1) and show it in the CLI wrapper command line (as priority:..)
  • Super PI
    • Improved security by validating the number of loops that are calculated
  • wPrime
    • Added AMD Unsafe tweaks for Vermeer

Features & Improvements

  • Support for upcoming Threadripper (TR5)
  • Support for upcoming Raptor Lake Refresh
  • Improved installer speed
  • Added new result dialog theme: Noctua
  • Various visual improvements for result dialogs
    • Improved text readability by using Lato as list text font
    • Added validation/warning icons for detected CPU name and bus-dependent frequencies
    • Added black icon set for ASRock AQUA and Noctua theme
    • Improved rendering of benchmark names
    • Improved rendering of Options/Tweaks by using multiple lines, if necessary
  • Added SMBIOS support to improve security with spoofing of CPUID brand strings. If data from CPUID and SMBIOS is different, both will be displayed in the result dialog to help with moderation.
  • Added CPU architecture and CPUID family, model and stepping to result dialog to help with the identification of CPUs
  • Using SMBIOS to supply full core and thread count on CPUs with disabled cores
  • Implemented new NUMA Windows API to support new multi-group node-affinity of Sapphire Rapids (Windows 11 only)
  • Improved error message for HWBOT uploads that respond with HTTP error code 403 due to the CPU or GPU missing in the HWBOT Submission API database
  • Removed F7 hotkey (flush cache) in favor of CPU-Z's hotkey to save a validation file
  • Added Intel P-Core and E-Core information to the HWBOT submissions (core count, max clocks, max temp)
  • Improved detection of unknown Windows 10 and 11 versions
  • Added Priority option to CLI Wrapper
  • Updated HWiNFO to version 7.63 (fixes Command Rate detection on AM5, improved support for RTX 40x0 GPUs, ...)
  • Updated tools: HWiNFO 7.62, CPU-Z 2.0.7, GPU-Z 2.54


  • Fixed bug with missing cpu short names of Ryzen ES CPUs in the result list
  • Fixed GPUPI frozen on startup with Korean language setting (thx safedisk)
  • Fixed detection of Windows Server 2022

Internal changes

  • Added support to detect AMD Meltdown/Spectre mitigation flags and log them to client.log (or the Debug Console in BenchMate). This will be useful in an upcoming version of BenchMate.
  • Improved detection of CPU models higher than Skylake to determine if QPC/LAPIC timer is a safe option
  • Detect Hygon Dhyana as Zen-based CPU

Features & Improvements

  • Support for y-cruncher 0.7.10 with optimized Zen 4 workload
  • New tweak for Super PI: Sloppy Wazaa - shaves off 10 seconds on 32M using a 12900K @5G. Works a bit like Copy Waza, but takes a sloppy shortcut to clear the System Cache. Works best together with RAMDisk AWE to stabilize performance after multiple runs or reboots. The tweak is applied before Super PI is started. On reboots wait a few seconds for the OS to settle in before starting Super PI from BenchMate.
  • Record new sensor information during benchmark run: Bus frequency (min/max), Memory Command Rate and RFC timing
  • Record new information for Intel CPUs: Ring frequency (min/max), Gear Mode
  • Improved CPU name detection for AMD engineering samples (AMD Ryzen instead of "AMD Eng Sample: 100-000000514-20_Y")
  • Added TSC timestamp information and skew to result dialog (needed for bus frequency reliability check)
  • Necessary bus frequency reliability check for AMD Zen to indicate that bus frequency was changed but it does not reflect in certain frequencies shown on the result window. See below for details.
  • Updated HWiNFO integration to 7.31:
    • Improved Raptor Lake support
    • Improved Zen 4 support
    • Bus frequency reliablity check for AMD Zen
    • Support for HWiNFO's freshly proposed, more reliable bus frequency read method provided by motherboard vendors
  • Switched AMD Zen measurement of Effective Clock to HWiNFO's Snapshot Polling Mode (more precise)
  • Reimplementation of a much stricter and less error-prone file integrity check of benchmark files
  • File integrity hashes are no longer shown in the result window, if they are valid (to save some vertical space and the hashes are hard to interpret anyways)
  • Added detection of Windows 10 22H2 and Windows 11 22H2 (22621 and 22622)
  • Updated tools: HWiNFO 7.31 Beta, CPU-Z 2.0.2, GPU-Z 2.49


  • Fix HPET skew bug with AMD Zen 4 CPUs (runs were flagged invalid due to an error during the timer skew check)
  • Resubmit BenchMate driver to Microsoft attestation signing to avoid possible performance degradation (KB5017241)
  • Fixed a couple of motherboards that do not communicate the number of DIMM slots through SMBIOS (ie: ROG Impact)
  • Use HWiNFO's socket count for CPU count (fixes wrong SMBIOS entries of some vendors)

Internal changes

  • The client application is now compiled with the latest Visual Studio 2022 making use of C++20. The driver still needs 2019 to be able to compile a 32 bit version.
  • Fine-tuned compiler optimizations for the client and improved performance of several classes and functions

Important information regarding AMD Zen

Starting with HWiNFO 7.30 and BenchMate 10.12 all generations of AMD Zen CPUs may not show correct CPU/bus frequencies as well as Effective Clocks. Due to changes in HWiNFO dealing with unreliable bus frequencies during high load scenarios, the bus frequency is now only read once during startup of BenchMate/HWiNFO. Any changes of the bus clock after that point in time will not be reflected, hence showing unchanged values in the result dialog. There is a long-term plan to fix this at least for Zen 4. For now BenchMate detects these cases, marks the affected frequencies as unreliable and issues a warning.

Reworked Super PI benchmark integration and added new RAMDisk mode called "RAMDisk AWE". You'll also find a couple of bugfixes and smaller improvements like latest HWiNFO integration for upcoming CPU platforms.


  • Due to changes in the Super PI integration, you can no longer use the Task Manager to change CPU thread affinity! There is a new menu entry that helps you to select the affinity at any time.
  • The new RAMDisk AWE needs the SeLockMemoryPrivilege applied via group policy editor in Windows. See this FAQ entry for more details.
  • Updating an existing BenchMate installation sadly takes very long due to the huge number of files that need to be validated. If you want to speed up the process, you can uninstall the current version and install the new one. Your configuration and result files will be left in tact.

Features & Improvements

  • Super PI: The integration now uses a thread for the calculation to separate all UI events from the calculation. The window is snappy now and interacting with it should only have little influence on the score.
  • Super PI: Added a new menu entry to set thread affinity. Due to the newly introduced calculation thread, you can no longer use the Task Manager to change affinity! That would have a bad impact on the whole bench, because you would force UI and calculation to use the same core. Use the menu instead.
  • Super PI: Added a new RAMDisk mode called "RAMDisk AWE" and it is fast! It uses a technique called AWE (Address Windowing Extensions) to map the intermediate buffer files directly into memory. The score is also more stable now, producing similar numbers even after hours of benching.
  • Super PI: The previous default RAMDisk mode is now called "RAMDisk PM" (PM stands for Process Memory).
  • Super PI: The previous RAMDisk LPM mode is now called "RAMPDisk LP" (LP stands for Large Pages). It is technically the same as RAMDisk PM but uses Large Pages locked into memory.
  • Super PI: Renamed "LPM" tweak to "Standby Process Memory" so it is not confused with RAMDisk LP. This tweak essentially tells the OS to reserve up to 500 MB for Super PI, if needed.
  • Improved error messages for missing SeLockMemoryPrivilege OS setting for RAMDisk LP and RAMDisk AWE
  • Improved CLI Wrapper benchmark result stability
  • Added new option to select HWBOT submission server. Current options are: (Production), (Testing)
  • HWiNFO integration updated to latest beta of 7.27.7 to support the upcoming platforms as good as possible at this time. A big thank you to Martin Malik over at HWiNFO for his hard work!
  • Tool updates: HWiNFO 7.27.4840 Beta


  • Fixed CLI Wrapper thread affinity
  • Fixed Tweaks/options submenu not being shown after a fresh installation
  • Fixed layout of scores without categories for custom result dialog themes
  • Fixed CPU-Z icon going missing on result dialog button
  • Fixed GPUPI device name with duplicate vendor name (NVIDIA NVIDIA GeForce ...). This is actually a GPUPI/CUDA bug.
  • Fixed font size of result dialog's save button for huge DPI settings

With this release we will start a new versioning scheme. We simply dropped the 0 going from 0.10.9 to 10.10. This solves issues with the installer not being able to upgrade minor version in-place. It's also best to install the new release into a fresh new folder. Upcoming releases should be easily upgradable via the installer.

Features & Improvements

  • Improved rendering of result dialog with resize support and much better theming abilities
  • Added ASRock AQUA result dialog theme ❤
  • Added option dialog under Tools menu to change the theme and result dialog size (both at Auto by default, so BenchMate selects it for you)
  • Improved installer looks, handling and install size and performance
  • Automatic installation of the digital signing certificate for old OS versions
  • Trimmed CINEBENCH R11.5 to its absolute minimum which reduces the number of files from ~6660 to ~1460 (removed OpenGL test and alternate languages)
  • Updated HWiNFO integration to latest beta with early support for Intel Raptor Lake and AMD Raphael
  • Integrated and bundling for y-cruncher 0.7.9 (support for 0.7.8 was removed in favor of the new, faster version)
  • Improved CLI wrapper's benchmark output handling by showing inline progress. The output is now exactly the same as running the benchmark from the console.
  • Added timer support for mobile Alder Lake SKUs
  • Added preliminary timer support for Intel Raptor Lake
  • Improved OS detection for Windows 11
  • New, faster and unified hashing algorithm to check benchmark file integrity. Sadly all hashes differ from results of previous BenchMate versions, but now we don't need a CRC32 fallback for legacy CPUs and it's much easier to maintain.
  • Added more file integrity checks to multiple benchmarks correctly highlighted in yellow or the result dialog theme's primary color
  • Added support for the upcoming PYPrime 2.3 (or 3.0, whatever it's going to be called)
  • Prepared HWBOT submission for uploading results into the real categories
  • Updated tools: HWiNFO 2.27.4820 Beta, CPU-Z 2.01, GPU-Z 2.47


  • Fixed GPUPI, pifast and multiple other results not being shown in result list after restarting BenchMate
  • Fixed bug report tool - feel free to submit detailed bugs again
  • Fixed mouse clicks on buttons of the result dialog and CLI wrapper, that only activated the window, but did invoke the button action at the same time
  • Fixed support for Windows 7
  • Fixed installer error "Source file not found"

Internal changes

  • The client configuration and app configuration are each saved into their own json file now (client.json and apps.json). This was necessary for an upcoming feature to download benchmarks on-the-fly. Old config.json files are not compatible with the new structure.
  • Refactored CLI wrapper to use a single executable and .app files to launch CLI benchmarks like pifast, y-cruncher
  • Moved to vcpkg and dedicated manifests for project dependencies
  • Removed old dependencies used for never fully working DX9 support of legacy 3DMarks
  • Fixed CPUID TSC frequency detection for rare CPUs (Goldmont)

Official Alder Lake support, new theming for result dialog, ASUS theme and updated default theme

  • Official Intel Alder Lake support
  • Support for HWBOT submissions into their new P categories
  • Result dialog theming
  • If you are using an ASUS ROG motherboard, you are in for a treat. Our designer made a special ASUS skin that will automatically be applied to the result dialog. More to come!
  • The launcher for command line benchmarks is now a single common executable for all workloads (pifast, 7-Zip, y-cruncher, PYPrime). This is an import step for the new download feature that's coming soon.
  • New config option for y-cruncher to select a specific workload executable
  • Updated HWiNFO to 4.14a
  • Updated CPU-Z to 1.98
  • Updated GPU-Z to 2.43

Big thanks to ASUS and G.Skill for supporting us with the necessary hardware. ❤

New installer, official Rocket Lake and Windows 11 support, several important bugfixes

BenchMate is now operated by Lentils e.U. instead of is now a nonprofit association and therefore can't have an extended validation certificate, which is needed by the BenchMate kernel driver. This is purely an organizational change without any practical impact.

  • Official support for Rocket Lake
  • Official support for Windows 11
  • New installer with a classic look using basic MSI functionality. It's also smaller in size and should be much more stable than the last one. There is only a single offline installer for now, which is the basic use case for most overclockers in our opinion. A smaller installer and download of benchmarks and tools inside the client is next on our list.
  • Added the latest HWiNFO SDK to fix the Rocket Lake 4C/6C memory frequency bug
  • Added OS detection for the latest Windows 10 and Windows Server version
  • Added detection for upcoming CPU and platform releases
  • Memory frequency and timings are no longer mandatory to produce a valid result file
  • Improved score security by adding more result validation
  • Improved executable obfuscation
  • Improved error messages for result file creation errors. It will exactly show what went wrong like for example which mandatory data is missing that prevents a valid result file.
  • Fixed a bug where some benchmarks showed "HWiNFO error (reboot?)" when mandatory sensors where missing
  • Fixed a bug where a skewed timer could result in a negative score/time
  • Fixed a bug where some benchmarks were not visible in the result list after restarting the client (thanks to L0wpull from HWBOT)
  • Fixed DIMM slot detection of EVGA Z590 DARK
  • Fixed Rocket Lake timer detection warning in GPUPI. Important: GPUPI will show a 20 MHz HPET instead, which is actually using QPC with an appropriate factor for timestamps internally. It's a workaround to avoid the unsafe timer warnings and does not mean that you really have HPET enabled!
  • Fixed HPET bug on legacy platforms (X58 and below) where time could not be measured for a longer period than up to 4 minutes (thanks to L0wpull from HWBOT)
  • Fixed a bug that closed the launcher before displaying an error message
  • Updated all tools to their latest versions: HWiNFO 7.07, CPU-Z 1.96.1, GPU-Z 2.41.0
  • Compiled with the latest Visual Studio version with all libraries at their latest versions
  • New Extended Validation Certificate because the old one ran out and was validated for
  • Switched to DigitalOcean Spaces for faster file downloads and CDN support using our own subdomain (

Big thanks to ASRock, Nick Shih and Splave. Without your help we would not have been able to implement and test the Rocket Lake support. ❤

Maintenance update to fix false detection of Antivirus tools, HWiNFO update to support latest hardware

  • Wrongly detected malware warning fixed for multiple AV engines by improving the code and compilation to circumvent detection
  • Smaller improvements to security and performance for all executables
  • Updated HWiNFO SDK to the latest beta version (7.03)
  • Updated HWiNFO tool to the latest beta version
  • Updated all libraries/dependencies to their latest versions
  • Updated Qt installer framework to 4.1
  • Reported the executable of BenchMate to Microsoft so they do something about their false positives

Integrates PYPrime and HWiNFO, improves support for AMD Ryzen CPUs, adds new sensor values to a trimmed down result dialog and fixes several bugs.

  • Added PYPrime, a single-threaded prime number benchmark written in python by Monabuntur
  • Added HWiNFO 6.42 as additional monitoring tool. This version matches the integrated HWiNFO version for sensor measurement and can be run next to BenchMate.
  • Improved HWiNFO sensor measurement performance by avoiding redundant CPU scans
  • Effective Clock will no longer be shown if the CPU's TSC timer is skewed. It's unreliable due to the skew and will show wrong values.
  • AMD Zen support: CPU temperature, wattage support, FCLK and UCLK
  • Improved readability of raw CPUID brand string in result dialog by replacing (r), (tm) and (c) with the appropriate unicode character
  • Added measurement of average CPU load during the run. It's a running average of all collected sensor measurements, so don't expect a precise value here. But it can help especially in continous multi-threaded workloads to see if the work was done.
  • Renamed Mainboard to the correct term: Motherboard
  • Merged CPU, memory and motherboard information in the result dialog to reduce its size
  • Updated CPU-Z to version 1.95
  • Versioning of benchmarks now relies completely on the internal version detection inside the benchmark process. If the version can't be detected, the result will be invalid.
  • Improved workflow for CLI wrapper's "RUN" button by avoiding that you need to press it twice due to the application not having the current focus
  • Launcher: Added a maximum operation timeout to the update check (5 seconds)
  • Improved security for guarded benchmarks
  • Improved security and performance of Super PI integration. For those that had a problem with capturing Super PI results, this is now fixed as well.
  • Allowing the presence of an enabled Hyper-V hypervisor, but will show invalid results when used inside a virtual machine.
  • Fixed a bug with wrong detection of Engineering Sample flag on AMD Ryzen CPUs
  • Fixed DIMM slot count for an upcoming GIGABYTE motherboard
  • Fixed a bug on Ryzen where the HPET timer was disabled after returning from standby
  • Fixed a rare bug where previous run errors were overwritten
  • Fixed a bug where the launchpad is not fully visible/correctly updated when the client is initializing itself
  • Fixed a bug with GPUPI 3.3 results not being visible in the result list when BenchMate is restarted
  • Fixed a bug for y-cruncher on Skylake-X not showing any output
  • Fixed a rare bug in kernel driver where a specific error message could result in a system hang
  • Fixed a rare bug in kernel driver where a process error of a child workload was not correctly inherited by the parent
  • Fixed a bug with CLI wrapper showing "Client was not ready" although no run was started yet
  • Fixed a bug with skewed running load average if multiple GPUs are in the system
  • Fixed a bug where CINEBENCH installed as Windows app is detected by BenchMate and added to the launchpad. Windows apps run in a sandbox and therefore can not be guarded by BenchMate. Use the bundled versions of CINEBENCH that come with the BenchMate installer.
  • Multiple bugfixes for the Bug Report tool

Major release to ease the pain while we are working on BenchMate 1.0 and the validation platform.

  • Big improvement for compatibility! New minimum requirement for CPUs is now SSE2, which was introduced with the Intel Pentium 4/AMD Athlon64.
  • Introducing y-cruncher, CINEBENCH R15 EXTREME and CINEBENCH R23.
  • Online and Offline Installer, so you can customize what benchmarks, tools and prerequisites are installed. Updates can also be applied using the installer. The launcher will notify you of new updates.
  • New hotkey for result saving: Just press F6 anywhere after a benchmark has finished and the result dialog will pop up.
  • All major versions can be installed and run right next to each other. For example BenchMate 0.9.3 runs without hiccups next to 0.10.6.
  • Latest HWiNFO SDK, improved startup time, less impact on the score by sensor readings, works next to HWiNFO now
  • New hardware supported: AMD Vermeer, Intel Comet Lake and Rocket Lake, NVIDIA Ampere, AMD Navi.
  • New logo, a launcher with a splashscreen and improved visuals for the client.
  • New result dialog to give your results a more professional look.
  • Added more details to the result dialog:
    • Raw CPUID brand string
    • Max Effective Clock
    • Full transparency on used timers
    • Motherboard DIMM count
    • Amount of memory available for the OS (the MAXMEM setting)
    • Windows OS build - needed for score judging on HWBOT, but will later be only visible internally for moderators
    • HWiNFO DLL and driver version
    • Precise and secure benchmark version
  • Improved integrations for Super PI, GPUPI and CINEBENCH in terms of performance, workflow and security.
  • Major changes to wPrime:
    • Rewrite of the workload priority tweak (was called Child Process Priority or CPP before), it should be faster now
    • New tweak to automatically apply the workload process affinities. The is exactly what Process Lasso does and helps with the synchronicity of the workload processes (the calculation should be more stable)
    • Shows every result on the GUI, not only the best. This was always a big security concern for me and I finally decided to remove that "feature" from wPrime.
    • Improved performance for high core count and NUMA CPUs. Try the HCC tweak with various thread options to let the benchmark breathe.
    • MSVMVB6.DLL version is now shown in the wPrime GUI as well as the result dialog. Read my FB post on why this is important.
    • Code cleanup and minor performance improvements
  • Many improvements for the CLI wrapper that is used for all command line benchmarks (y-cruncher, 7-Zip and pifast). It is now more stable, more secure, dark-themed and you will see the command line that is executed. For y-cruncher I added the ability to select important command line parameters. These settings are saved in a configuration file, so you don't have to select them on each startup.
  • Rewrite of the security background services so they "just work".
  • Improved handling of CPUs with multiple NUMA nodes and high core counts (> 64 Threads) like the Threadripper 3990X.
  • Lots of fixes for Windows 7 including automatic installation of all prerequisites like the mandatory SHA2 patch (be sure to select the component in the installer!)
  • Better error message boxes and more as well as more detailed error messages.
  • Support for GPUPI 2.3.4 so you can finally break these unbeatable old records. You have to add it manually though - just open it while BenchMate is running and it will be added instantly to the launchpad. Also be sure to set the "Timer Emulation" option to "Driver" if GPUPI cries for a valid timer (right click on the entry in the launchpad => Options/Tweaks)
  • Improved security by enhanced detection of the benchmark and its version from within the running benchmark process. For example you will now find the precise Super PI version in the result dialog.
  • Memory Violation error for CINEBENCH is now fixed, so you can use old graphics cards now.
  • Launch errors for benchmarks are now correctly shown. For example if OpenCL or the VC Redist is not available, the appropriate error message will be displayed.