Prime95 is a great stress testing tool for many reasons; it can be used by end-users for normal stress testing, it can be used by lab technicians for advanced, high-level stress testing, users can choose how aggressively they want to stress test their hardware by choosing the number of tests that are run, and testing can be stopped at any time.
Typically, a stress test yields two types of results — either quantitative or qualitative. The test allows a network admin to collect quantitative results, for example, the frequency of errors or number of crashes. Furmark is a stability and stress testing tool designed especially for graphics cards and runs a very intensive “Fur” rendering algorithm which is very good at pushing the GPU to its absolute limits. The freeware stress test tool HeavyLoad was developed to bring your PC to its limits. It can be installed on any edition of Windows and client/server versions (32-bit and 64-bit) currently supported by Microsoft. HeavyLoad puts your workstation or server PC under a heavy load. Stress testing is a good way to check the reliability and stability of your system after buying or building a new PC, upgrading your graphics card, or overclocking your GPU. It can help you identify faulty hardware or a need for better cooling.
-->System designers should run stress tests and long-duration tests on their Modern Standby systems to help identify and resolve potential reliability issues. Modern Standby enables the system to keep running, even when it is in a low-power, screen-off state. This state is different from the traditional ACPI Sleep (S3) and Hibernate (S4) states, in which much of the system hardware and software is stopped and then stays inactive until it is later restarted on resume.
Modern Standby enables the system to stay up and running for a much longer total time and can therefore expose hardware and software reliability issues that would not be discovered on a system that supports only S3 and S4.
Entry and exit
Every Modern Standby system should be validated to enter and exit Modern Standby for at least 1,000 cycles without failure. Entry to and exit from Modern Standby is the user's primary interaction with low-power operation on the system and should be extremely reliable.
Successfully entering and exiting Modern Standby validates a number of hardware, firmware, and device-driver components, which include:
- The platform hardware that manages power-button operation, including the power-management IC (PMIC).
- The display-panel management and initialization hardware.
- The Wi-Fi and networking device firmware and driver.
- The graphics device driver.
Stress-testing of Modern Standby entry and exit can be automated using the PwrTest tool. PwrTest should be installed on the target system as part of the Windows Driver Kit (WDK), which includes additional software for automating the system power button on Modern Standby systems.
| Test scenario | Expected result | Diagnostic notes |
|---|---|---|
The system can enter and exit Modern Standby reliably for at least 1,000 cycles. | Use the PwrTest tool and the /cs command-line option to automatically cycle the system through Modern Standby for 1,000 cycles. The expected result is that the system completes all 1,000 cycles. | We recommend incrementally increasing the stress test to 1,000 cycles. First, test for 100 cycles. If an error is found, connect the system to a kernel debugger and to the SoC hardware debugger, and repeat the 100-cycle test to capture and determine the root cause of the issue. After the 100-cycle test successfully completes, extend the cycle count to 500 cycles and then to 1,000 cycles. |
SoC low-power state transitions
The firmware and drivers that are responsible for managing SoC transitions between the idle and active power states must be highly reliable to withstand the stresses of operating for long periods in Modern Standby. SoC low-power state transitions should be stressed through long-duration Modern Standby testing. This testing helps ensure that the system remains reliably operational during long Modern Standby durations, such as over the weekend. This test should be performed while connected to AC power.
| Measurement scenario | Expected result | Power notes |
|---|---|---|
The system can stay in Modern Standby for 100 consecutive hours and is functional on exit. The system maintains Wi-Fi connectivity during the 100 hours and Wi-Fi connectivity is functional on exit. | Put the system in Modern Standby and wake it with the power button after 100 hours. The expected result is that the system powers on instantly and the Wi-Fi connection is operational without additional configuration or selection of a Wi-Fi network. | We recommend incrementally increasing the long-duration test to 100 hours. First, test for 24 hours. If an error is found, connect the system to a kernel debugger and to the SoC hardware debugger, and repeat the 24-hour test to capture and determine the root cause of the issue. After the 24-hour test successfully completes, extend the duration to 100 hours. |
Windows HLK Modern Standby stress test
The Windows Hardware Lab Kit (HLK) includes a Modern Standby stress test named Connected Standby Stress with Driver Verifier's Concurrency Stress that exercises automatic modern standby transitions at the same time that device drivers are exercised for device operation. The test is designed to verify that the device and its driver(s) continue to function as the system transitions to and from the Modern Standby power state.
This test is a critical part of validating that the system continues to operate as expected after it exits Modern Standby. This test is included as part of the Windows HLK and is required for system certification.
Test operation
The test uses the Windows Device Testing Framework (WDTF) SimpleIO interfaces to exercise devices that are enumerated on the system. These devices include sensors, cameras, audio, graphics, Wi-Fi, storage, and Bluetooth devices. The test places the system in Modern Standby for one minute, and then transitions the system out of Modern Standby and exercises the devices for 30 seconds. This cycle repeats 150 times.
During test execution, Driver Verifier is enabled to help identify driver bugs and memory leaks.
The test helps identify the following system or device driver problems:
- A system crash or hang during device operation after a Modern Standby session.
- The inability for the system to enter the low-power state (deepest runtime idle platform state, or DRIPS) after device activity.
- Driver issues that are identified by Driver Verifier, including system corruption, driver failures, and memory leaks.
- Driver issues after resume from Modern Standby, including unresponsiveness, crashes, or problem codes.
Resolving test failures
The test exercises multiple devices, which can result in different types of test failures. Identifying the type of test failure is the first step to finding the root cause of system or driver issues.
The test typically fails in one of the following three failure modes:
- The test fails and the failure is recorded in the Windows HLK logs, which contain data about the detected failure.
- The test fails, but the system does not report to the Windows HLK server as a result of the failure; however, the system is responsive and works with local interaction.
- The test does not complete and the system under test crashes or hangs (frozen at a black screen).
Debugging test failures that are recorded in the Windows HLK logs
There are two common failure types when test failures are recorded in the Windows HLK logs:
- The system failed to enter the low-power state (DRIPS) during the test.
- The test detected that it could no longer communicate with a driver, and a time-out occurred.
You can use the SleepStudy report, which is included as part of the test logs, to identify which components are responsible for preventing the system from entering the low-power state (DRIPS). There are several common causes:
- Test setup and configuration problems, including using a wired Ethernet adapter that does not support NDIS 6.3 and Modern Standby functionality.
- DHCP server problems on the wired LAN network.
- A device and/or driver that does not correctly idle to its own low-power mode during Modern Standby.
The test logs might also include a failure message that indicates which devices did not respond to I/O requests in a timely manner. This condition is considered a test failure because it can prevent the user or an app from being functional when the system resumes from Modern Standby.
The test logs indicate the last devices to perform I/O operations—these devices are the source of the test failure. The test log output in the following example shows that the ACPIXXXX2&DAFA3FF&1 device timed out.
Message | 7/16/2013 12:50:24.333 AM | WDTF_SIMPLEIO_STRESS_PROC : - WaitAsyncCompletion(Some Location Sensor Device ACPIXXXX2&DAFA3FF&1) |
Message | 7/16/2013 12:59:50.333 AM | WDTF_SIMPLEIO_STRESS_PROC : - WaitAsyncCompletion(Some Other Device XXX_XXXUART_XXX3&2F829BAD&0&F00D) |
A common cause of failures is poor GPS reception, which causes the GPS device to take extremely long amounts of time to reply to I/O requests. For more information about running this test on systems with GPS devices, see Notes for systems that are equipped with GPS.
Debugging test failures without logs (and a responsive system)
If the system under test is still running with no signs that the test is still running, the most likely cause is that the system has encountered a fatal error or restarted. To debug these issues, check the system directory for any dump files, and disable any hardware watchdog that might reset the system.
Debugging test failures when the system is unresponsive (black screen)
If the system is frozen on a black screen, a kernel debugger must be must connected to the system to diagnose the problem.
If the kernel debugger is already connected and the system is not responding to the kernel debugger, a hardware debugger is required to identify the reason that the system locks up. You can consult with the core silicon/SoC provider for additional assistance with debugging.
Additional HLK Documentation
Notes for systems equipped with GPS
If the system-under-test has a GPS device or location sensor device, the following Windows settings must be enabled before running the test:
- Control PanelHardware and SoundLocation SettingsTurn on the Windows Location platform
- PC SettingsPrivacyLocation: Let Windows and apps use my location
You can use the Sensor Diagnostic Tool in the Windows Driver Kit (WDK) to confirm the reception of the GPS signal at the test site. For more information, see Testing sensor functionality with the Sensor Diagnostic Tool.
If you’re having problems with your computer and it isn’t behaving correctly in certain situations, then it could be a cause for concern. One of these situations could be when you are performing resource intensive tasks and the system becomes unstable or crashes while being put under stress. This could be a problem such as the CPU overheating and shutting itself down, or even the graphics card running into similar trouble. Or maybe the power supply simply cannot cope anymore with all hardware components running at full intensity…
Whether you’re someone who wants to test their PC components because they suspect a possible fault, or have bought / built your own PC and want to make sure it runs stably and reliably at maximum load, you need a program to be able to put the system under the desired levels of stress for periods of time to monitor the situation. Here are 10 tools (9 of them free) to do just that and put your system under huge amounts of stress to check for faults or problems.
1. HeavyLoadHeavyLoad is a utility that aims to stress the main component areas of a PC, namely processor, memory, hard drive and graphics. It can also run these tests individually or altogether which is obviously the time maximum stress will be placed of the system hardware and also the power supply. The Disk space and Free memory tests are not really there to put undue stress on those components, but rather continually writes a large file to the drive and allocates / deallocates memory to the system simulating heavy load when all major components are utilized.
There is an option though to add more stress to the hard drive by using Jam software’s other popular included tool Treesize Free to simulate more heavy disc access. HeavyLoad is available as portable and installer versions and is a very useful overall system stability tester to keep in the USB toolkit. Works on Windows XP to Windows 8 32-bit and 64-bit.
Download HeavyLoad
2. FurMark
Furmark is a stability and stress testing tool designed especially for graphics cards and runs a very intensive “Fur” rendering algorithm which is very good at pushing the GPU to its absolute limits. There are a few settings that can be changed such as resolution, full screen mode and anti aliasing, and a few presets are available such as running in the HD resolutions of 720 / 1080, or running a burn-in test for 15 minutes. The benchmark’s default run time and an alarm for the maximum allowable temperature for the graphics card is found via the Settings window.
Benchmark scores can be compared or viewed online. FurMark is compatible with Windows XP and above.
Download FurMark
3. StressMyPC
This is a simple, tiny and portable utility of around 20KB that can run a stability test on your single, multi core or multi threaded processor. In addition it can also perform a couple of other tests such as a simple GPU graphics test and also one for the hard drive. The “Paint-Stress” GPU test is enabled by default and the “HD-test” and a more aggressive CPU test which will push your processor towards 100% (the standard test used about 60% of a dual core CPU during testing) can be enabled by the buttons at the top of the window. StressMyPC works on all versions of Windows 2000 and above, including 64-bit.
Download StressMyPC
4. System Stability Tester
System Stability Tester works by simply using the well known and famous method of telling the computer to calculate the value of Pi up to 128 million digits. This will completely consume your processor for as long as the test runs and can also be used as a basic benchmarking tool to see how long it takes your CPU to calculate the specified number of digits. The range can be between 128 thousand up to 128 million and can be run continuously up to 50 times using up to 32 threads. There are 2 methods to choose from, Borwein and Gauss-Legendre which is also used by the classic SuperPi tool.
Works on Windows XP and above, portable and installer versions are available.
Download System Stability Tester
5. IntelBurnTest
Despite the name, IntelBurnTest actually works fine for testing on AMD processors as well, and is called as such because it makes use of the Intel Linpack libraries which Intel themselves use to stress test CPU’s. Usage is easy and all you have to do is set the number of times to run the test, the number of threads to use and the test stress level. This can be Standard, High, Very High or Maximum and if your available RAM is less than what the test requires, choose Custom and set the amount of memory to use accordingly.
The program is portable and runs on Windows XP and above.
Download IntelBurnTest
12Next › View All You might also like:
11 Tools to Find out Information About Your Computer’s Hardware5 Free Tools to Check and Test USB Flash Drives6 Free Programs to Check Your Video Card Memory For Errors8 Free Tools to Test Read and Write Speed of USB Flash Drives2 Ways to Test and Determine If Your BitTorrent Speed is Limited or ThrottledA nice collection of stress tests, but I am currently searching for DOS-based software. Will keep looking, but perhaps you could add something in this vein the next time you update the article?
ReplyBurninTest software is good for testing, but unfortunately it’s not free
Replyany commandline tools, i need them in my automation
Reply3d mark is good for higher end graphics cards
ReplyAida64 is also good program. It is almost head-to-head with BurnIn Test made by PassMark
ReplyI generally run Orthos and MemTest. For the hard disk, when I get a new one, I copy tons of files big and small to it and then try using these files. Last drive showed problems immediately after this.
I also always do a Nero Recode and x264 encodes since these are heavy CPU programs.
ReplyUseful tool and have noted contributors very helpful warnings.
Free Cpu Test Windows 10
ReplyLooks like a decent benchmarking software. Thanks Ray!
ReplyMy suggestion regarding this type of program:
Run it well BEFORE your warranty expires. Make sure you didn’t get a system with faulty components. Don’t experiment on an older system (with gigs of files & programs at risk). If something is “marginal”, these programs WILL find them – often destructively!
Second this – it is actually possible for stress testing software to fry your PC. E.g. normal use of a video card will not come close to the temperature and power consumption of running a 100% stress test like FurMark, and this could cause an otherwise working component to fail. It does mean the component is running out of spec, so it could even be recommended to run this just before your warranty expires, but not after.
ReplyAn example of how testing software can be destructive if it does not change or write anything new or over existing files, code or registry? If a tool writes its own fix without showing you changes before they happen then don’t use it.
ReplyThanks again for a useful tool Raymond
ReplyThank you Raymond.
Nice tool.
ReplyA very good and large hardware store here uses Everest for stability testing
ReplyThanks for the info Raymond!