[SOLVED] AIDA64 Engineer system stability test

[sbtang]

Hi everyone, I have some questions regarding the AIDA64 Engineer system stability test.

Here are the specs for my Windows 10 PC, which was custom-built for me by a local IT company here in Melbourne, Australia:

• AMD Ryzen 9 3900X 3.8 GHz Processor (12 Core / 24 Thread)
• Gigabyte X570 Aorus Elite WiFi Motherboard
• Motherboard has WiFi and Bluetooth Built-In
• 64 GB DDR4 3200 Team T-Force Vulcan RAM
• 1 TB PCIe Generation 4 Gigabyte NVMe Solid State Drive
• Gigabyte Nvidia GeForce GTX-1050Ti 4GB Graphics Card
• Gigabyte Case with 750-Watt Power Supply

Michael, the hardware guy at my local IT company who built my PC, suggested that run a “System stability test” in AIDA64 Engineer. “You should select ‘Stress GPU(s)’ in the top left,” he wrote in an email, “and then press start in the bottom left. That Window will allow you monitor system temperatures as well as CPU usage and CPU throttling (if there is any).”

In a follow-up email, Michael explained what the AIDA64 Engineer system stability test does: “The application synthetically simulates a load that stresses the CPU, motherboard, RAM, GPU and drive(s). These are all the components that can be tested by any software package, and are the key points of failure.”

I totally forgot about running the stress test for months. I got around to running it yesterday. Michael has since left my local IT company. I rang the IT company yesterday and spoke to Ryan G. He informed me that he’s not familiar with AIDA64.

I was unsure whether to tick the box for “stress local disks” before running the system stability test. After re-reading Michael’s email which said that the AIDA64 system stability test “synthetically simulates a load that stresses the CPU, motherboard, RAM, GPU and drive(s)”, I decided to tick the box for “stress local disks”. I then ran the system stability test for three hours and 32 minutes with the following boxes ticked:

• Stress CPU
• Stress FPU
• Stress cache
• Stress system memory
• Stress local disks
• Stress GPU(s)

Please see below for the results:

I stopped the system stability test after three hours and 32 minutes because I discovered an AIDA64 message board thread which said: “It’s not recommended to use the write tests on SSD drives, due to them wearing out SSD flash memory cells.”
I then ran the system stability test for one hour, 11 minutes and 57 seconds with the following boxes ticked:

• Stress CPU
• Stress FPU
• Stress cache
• Stress system memory
• Stress GPU(s)

Please see below for the results:

Here are my questions:

1. Are the results of my first system stability test good?
2. Are the results of my second system stability test good?
3. How much did I wear out the flash memory cells in my 1 TB PCIe Generation 4 Gigabyte NVMe SSD when I ran the system stability test for three hours and 32 minutes with the “Stress local disks” box ticked?
4. Going forward, which boxes should I tick when I run a system stability test using AIDA64?
5. Going forward, how long should I run the AIDA64 system stability test for on my PC?

How do I read the AIDA64 system stability test results? What’s a good result?

 

[MeanMachine41]

Aida64 is an excellent stress tester and it's free. I use it with all my builds.

1 and 2: Both tests indicate high CPU temps at close to 90C so your Cooling system is barely coping. you should aim for no more than 80C under load. List your cooling system.?

3. Impossible to tell the wear on your SSD however i doubt it would be much.
According to Aida64 forum admin the System Stability Test stresses all local disks simultaneously, but only as long as they contain a partition table and they have files on them. This is because that test is a file system test, so it relies on the existence of files on the tested drive(s).
They say It's not recommended to use the write tests on SSD drives, due to them wearing out SSD flash memory cells. You can however use the new Linear Write + Verify test (of AIDA64 Disk Benchmark) once on each new drive to see if it has proper data integrity. That will only consume 1 write cycle for each cell, so it's not excessive in terms of wearing the memory cells out.

4. The initial stress test should be run with check boxes for CPU, FPU and Cache and other tests run separately. Initially test for 10mins and stop the test if temps exceed 80C. Once your satisfied with overhaul stability, run the stress test for 1hr.

Use AIDA64 in conjunction with HWinfo64 and your looking for temperature reading on the die when under load and your Rail voltages to check for out of tolerance (5% +/-) spikes.

I don't use the RAM test but for testing DIMMs I use Memtest86 from a USB.

 

[sbtang]

Hi, thanks very much for your helpful message. I just want to clarify how I should use AIDA64 going forward.

My understanding is that you're saying that I should:

1. Run the AIDA64 stress test with only the check boxes for CPU, FPU and Cache ticked for 10 minutes.
2. See if the CPU temperature exceeds 80 degrees Celsius on that first test.
3. If the CPU temperature does not exceed 80 degrees Celsius on the first test, then I should run the same test again for one hour.
4. See if the CPU temperature exceeds 80 degrees Celsius on the second test.

Is my understanding set out above correct?

What did you mean by "other tests run separately"?

Should I use the "Stress GPU(s)" test? If so, how? What should I look for in the results?

Should I use the "Stress system memory" test? If so, how? What should I look for in the results?

 

[MeanMachine41]

Yes the setup as listed is correct.
Testing the CPU, FPU and Cache is primarily testing CPU thermals.
I don't run initial testing on all tests simultaneously but if I did it would be one test at a time. There are more stressful GPU tests than what AIDA64 offers and I use benchmarks such as Heaven, Cinebench.

Aida64 is but one stress tester in your toolbox.

AIDA64 with Just Memory. Is a good choice for evaluating the stability of a memory overclock on its own, or as a supplemental test combined with other workloads.
The Memory Latency benchmark measures the typical delay when the CPU reads data from system memory. Memory latency time means the penalty measured from the issuing of the read command until the data arrives to the integer registers of the CPU.

I consider Realbench the ultimate real world tester and if you pass that test then your system is 100% stable.