Overclock.net banner

1 - 1 of 1 Posts

3 Posts
Discussion Starter #1
It's Probably Fine - Stability Test [v1.13]


First I'd like to start by saying that I have a background in computing, overclocking and scripting but this is my first public release of any of the tools I have created or used over the years. If anyone has any issues with what I've created please get in contact. I will be more that happy to engage to resolve any issues that may occur, I understand that I'm the outsider right now (though I have been lurking in the shadows for quite some time) and I could be stepping on some toes unintentionally.

This specific tool is something that came to me after seeing how the "Custom x264 with Loop Functionality and Other Improvements v2.06" works [you guy's have to come up with a catchier name]. I love the method used by that tool but figured I could improve it. I want to make it very clear that, not only was the tool the inspiration for this project but I have immense respect for Darkwizzie in particular and the information he puts together for the overclocking threads. I also mustn't forget Angelotti and JackCY who also contributed to that tool. Although my script started based largely upon their code, I think they would agree that I have heavily modified the code from what they originally created and I hope they might think that I've created something better at the end of it. A quick shout out to stackoverflow too; without which I don't think I would have the first clue what I was doing.

What is this?

Right, well. Essentially what I've tried to create here is a "one size fits all" Stability Test.

There is so much focus on creating or using Benchmark's in the overclocking community that people often forget that user's actually need something to test the stability of their overclock for 24/7 use, rather than just a benchmark showing them a score. So let's get it out on the table from the start: This Test Will Not Score You. You are unlikely to see whether your PC is running faster using this tool. It is purely aimed at "proving" stability, as much as technically possible with the tools available.

What's different?

So essentially I've combined lots of tools (freely available) into one giant stability test.

It will run through a whole variety of the tools one by one in a loop much like the x264 test (which is also included for good measure).
I have also included temperature monitoring software and a whole variety of other bits and bobs that will provide you info on what your system is doing at any given point in time.
The test is automated and it will even fit itself to a 1080p screen automatically so almost everything is visible.
I've also adjusted all of the settings for all of the tests so that they're optimised for most standard configurations.
Most importantly, there is "downtime" and "load variation" within the tests. Often stability tests throw 100% load at systems constantly assuming that's the best way to trigger faults but most users actually run between the 20-75% range most of the time. So certain tests within this tool will use a varied load in the hope of creating a more realistic environment, synthetically.

How long does it take?

The test has a built in 20 Hour Timer. You should see what happens at the end...

You should be leaving your system overnight, ideally. I would say that over 10 rounds should be a pretty healthy indicator of stability or when the timer reaches 00:00:00.

1 Round takes approx 1 Hour on a Stock 6700K (or did last I checked, which was a few versions ago I must admit). Slower systems take longer, up to 2-3 hours per round has been seen. If it's taking longer than 3 hours for 1 round; either something is wrong with the tests (check the logs) or your system is super slow, please upgrade [or you have 128GB of RAM, I mean seriously is that even necessary?].

Why use this?

I like to think there are many reasons. My main reason is that I don't think that any single tool currently available provides a good stress for a system. Very few of them provide a realistic workload for the computer. My opinion changed when I saw the x264 test and it was ground breaking for me [long term Prime95 believer until Haswell/Skylake heat issues became truly apparent]. However, I have a fair bit of evidence to show that this tool alone isn't very good at proving stability; regrettably I can't share this with you because it's protected data but I can share my experience and my solution. I'm also not saying that my tool is perfect. I have already seen systems pass a few hours of this test and when a real application is loaded the system BSOD with an overclocking stability code (124) but it's far fewer than any other tool I have used before [though it's still early days].

Things Worth Noting

I'm pretty confident that the software is in a usable state. It's been tested on over 5,000 systems since it's first release about half of which were overclocked in one form or another. However there are some thing to be aware of.

First - "cmdow.exe" shows up as a "false positive" in certain Anti-Virus applications:
I'll leave it up to you to decide. In my script the tool is used to move the Windows around on screen. It's not essential to the functionality of the tool but it's a nice feature. For some reason the anti-virus applications deem this sort of behaviour to be "risky" because it could be used to hide malicious applications. Just delete the file cmdow.exe from the Tests directory if you're not sure and it won't run.

Second - Matinee was showing some issues with the AMD 380 GPU's. The tool will repeatedly crash during testing. It's essentially only there to keep your system from going to sleep. You can use something like Unigine Valley in Windowed mode instead if it's crashing for you, but I can't redistribute this for licensing reasons. I believe this to be fixed in the latest version though but I don't have a 380 GPU to hand to confirm.

Third - if there is anything within my package that I shouldn't be redistributing please let me know and I'll work out an alternative. I'm pretty confident I've read all the terms and conditions but this has been a long time coming so I could be wrong or missed something crucial. I'm certainly no expert in that field.

Fourth - on that note of redistribution. Please do not redistribute this tool without my express permission. I will happily discuss any propositions made privately and if anyone wants to contribute to the project please let me know and I'm sure we can work something out.

Fifth - it should be obvious, but don't do anything else on the system during testing. It's designed to use all of your RAM at points and if you run lots of software along side it it won't have enough space available. It will either grind to an almost halt whilst thrashing your virtual memory [Linpack] or it will give an error about not enough memory [y-cruncher]. On the note of RAM it will only work with certain "standard" values; 4GB; 8GB; 16GB, etc up to 128GB. If your RAM doesn't match ones of those values it will skip certain tests. I haven't got a solution to this just yet.

Sixth - Throttling - yes your system may start throttling during the Linpack stage of the test. This is nothing to be concerned about, though it shouldn't throttle the whole way through or too low during, the Linpack stage is absolutely thrashing the CPU & RAM at the same time. It's designed to be relatively short term test though you'll notice that the more RAM you have the longer it takes I've limited this as much as possible and only at 32GB or higher RAM sizes does the test get exceptionally long. Throttling also shouldn't occur on any of the other tests, if it does so it indicates an issue with the cooling.

Where do I get it?

Well, if you've made it this far well done

Links are below:

7-Zip: https://drive.google.com/file/d/0B_vQYAge3O9gWGIwbU9DbVIzYXM/view?usp=sharing
Zip: https://drive.google.com/file/d/0B_vQYAge3O9gMEluaGVRc053Zjg/view?usp=sharing

Extract the files and run Launcher.bat - it will automatically ask for Admin rights or you can right click Run as Admin (up to you).

Most importantly! Please enjoy using this.
Please contact me or reply to this topic if you have any questions or queries or feedback at all.
1 - 1 of 1 Posts