EffectiveMark2 toolkit

This software is a toolkit designed especially for checking stability of overclocked components. This includes CPU, RAM, GPU and VRAM tests. All have built-in error detection.

CPU Test (CPUTest.py)

This test takes no command-line arguments. Classic CPU test launches as many processes as CPU cores are detected. It uses CPU to calculate arithmetic operations on number and compares it with precalculated results.

RAM Test (RAMTst.py)

**WARNING: This test is NOT as accurate as Memtest86, but it's good enough to test R/W operations on system

GPU Test (GPUTest.py)

Test written in OpenCL, so it should work on both AMD and NVIDIA cards. Loads GPU with arithmetic operations which uses low amounts of VRAM which is useful when debugging overclocking problems. This test accepts following arguments from command line:

• --target-device, -td: For multi-gpu setup ONLY. Select which card should be tested. This argument doesn't accept names like "NVIDIA GTX 1080 Ti It accepts IDs like "0" (means GPU-0)

VRAM test (VRAMTest.py)

This test works similar to RAM test. Writes known pattern. then loads GPU with calculations that do not use VRAM. After computations are complete it reads VRAM pattern and check if it's still the same. Following arguments are accepted:

• --target-device, -td: Tested device's ID (multi-gpu setup ONLY)
• --chunk-size, -cs: Test uses "chunks" to control and fill VRAM. Smaller number will take more time to fill but can be more precise while calculating number of chunks. E.g. Assuming 6GB VRAM and chunk size is set to 512. Quick calculations shows, that using 10 chunks will fill 5120MB VRAM. Setting this to 256 or 128 will use more chunks but can use more VRAM
• --load-loops, -ll: This parameter defines how many times GPU will be "asked" to calculate some data. If test freezes or takes too long reducing this value can help
• --free-space, -fs: How much memory should be unallocated. If you are hitting OOMError raising this value can help.
• --warmup-temp, -wt: Temperature that must be reached by GPU before continuing, allows simulation of thermal edge- and worst-case scenarios.

Render Test (torus_test.py)

Uses GPU to render rotating torus with moving light which uses a lot of calculations. this is very aggressive GPU test which allows to see rendering artifacts.

Thermal Test (GPUTempTest.py)

This test measures how fast the GPU heats up. It will stop on 85*C or other temperature is specified in args

Summary

These test should be enough if you want to quickly check for most common errors caused by overclocking or degradation of components. It's definitely NOT enough to have 100% confidence of stable system (yet). This test does NOT provide GPU temperature / power consumption monitoring. This monitoring is written in "SysView" project. Practical information: ANY failed test means system instability even occasional ones. One test is NOT enough to judge system stability, but if GPU passes around 200-300 test loops without any errors it's strong sign that everything is OK.