If this page doesn't answer your questions, use our forum or contact us !


What is OCCT for ?

OCCT is a stability Checking tool that includes CPU, GPU, and power supply tests. It aims at detecting hardware errors, generated by faulty hardware, badly thought overclocks, and so on.

OCCT detects an error. Should i trust it ?

Yes. If OCCT reports an error, something weird has occured. For instance, OCCT asked for 2+2 to your CPU, and it got 5 as an answer, which is obviously wrong. This indicates that something is wrong with your hardware.

OCCT does not detect an error. Does that mean my computer is stable ?

Yes and no, it only depends on how long you ran the test. OCCT only creates the optimum conditions for an error to occur. But you may have an unstable component that just didn't generate the error during the test time frame. It's all about probabilities !

To lower this, the best way is to test for a longer time. Usually, we recommend at least 1h on the CPU tests. Usually, GPU test is much faster at detecting ( 30min is great for it ).

What do you consider a good testing method for overclocking ? How do you do it yourself ?

First, this is my way of overclocking, some do it otherwise, but here's how i do it.

I use CPU:OCCT Large data set first. I read the web ( Google is always your friend ) to get a rough idea of what others are accomplishing. Then i play with the BIOS, trying to overclock from the start, in the high portion of my target. If it doesn't boot, up the voltage by a notch (if not too high of course), or go down with the frequency, until it boots Windows and goes to the desktop. That's a start !

Then i launch OCCT, and launch the abovementioned CPU:OCCT test. Most likely, if you only got your desktop to show, you're in the "unstable" zone, and the test should detect an error usually quickly. Finally, i would stabilize it by fiddling with Bios parameters (it is highly dependent on your components) until i stabilize it for at least 20 minutes. The goal being to withstand a test of 1 hour. If it passes, i do that again for 3 hours, and i then consider it stable.

Unfortunatly, i cannot give you advice on which parameters you should touch. Sometimes the memory frequency, or voltage, sometimes the CPU voltage, sometimes the PCI-E frequency being too high because linked with your overclocking... they are too numerous !

How do i translate OCCT ?

Translating OCCT has been made easier with the 4.0 release. The only tool needed is a good text editor ! I'd recommend Notepad++.

In OCCT's folder, you will find a "locale" folder. Inside, you will find several files, as on this screenshot :

The file "BaseForTranslatingOCCT.txt" is the file containing all words and sentences to be translated. Copy this file and rename it to your language, and change the extension to .lang. For instance, copy and rename it "francais.lang".

Open up the file in Notepad++ (or your favorite text editor) and you will see the following :

As you can see, this is just plain XML and is easy to understand.

In the first <NAME> tag, enter the name of the language as you want it to appear (for instance, "Français").

Next, enter in "Cultures" one "<string>" tag per culture you will be supporting. It is used by OCCT to determine if it should use this file automatically when first launched, based on the culture set in Windows. You can find a list of those tags here.

Finally, edit every "<Translation>" tag. DO NOT TOUCH THE "<key>" Tag, it is the reference string used by OCCT ! Put your translation string into the "<value>" tag.

Launch OCCT, select your language in the settings to see it in action.

Once you're happy with it, send it by mail to iench 'at' ocbase 'dot' com , i'll test it and add it to the next release.

Thanks for your help in translating OCCT !


Back to top


What is this test good for ?

This test is considered the best at detecting errors. It generates less heat than Linpack, but detects computing errors much quicker. Don't get fooled by the higher temperatures numbers of Linpack ! This one is much faster to detect errors.

What is the difference between Small, Medium and Large data set ?

It is the size of the data set OCCT will use during testing.

Small uses a very small (duh) set, which will use only a small portion of ram, and calculations will mostly take place in the CPU Cache. It will generate the most heat, but as it tends to only test the CPU, is overall a tad less efficient at detecting errors

Large uses a much larger data set, and will test thus the CPU, RAM, and/or chipset. It is thought to be the most efficient all-around mode to detect errors

Medium, as its name states, stands in between.

What is the "Logical as physical" checkbox ?

Some CPU comes with Hyperthreading technologies. Briefly, each core of your CPU will be seen twice, so that two operations can be executed at the same time on each core (with more or less success, of course).

If you check this box, OCCT will treat your 2-cores CPU as a 4 cores CPU, and thus launch 4 threads that will take care of the testing. Experience shows this is more efficient.

How long should i test my computer for it to be considered stable ?

If you didn't read the "General > OCCT does not detect an error" entry, here is a summary : having an error is a trustworthy diagnostic, not having one is always a probability. The longer the test the better. Personally, i consider a test of one hour to be the minimum. I usually use a 3-hours long test to confirm it. You'll find people using longer, they may be right, i'm not the authority on that part. It's up to everyone. I don't think that testing below one hour is enough.


Back to top


What is this test good for ?

This is the "Maximum heat" test of your dreams. It will generate a lot of heat, making your CPU hot, real hot. However, experience shows it is not really great at detecting errors that are not related to overheating - CPU:OCCT (and especially in Large data set mode) is considered the best for that.

What is the "AVX" option ? When should i tick it ?

It turns out that a new version of linpack was released by Intel, supporting AVX instructions, their latest instruction set like SSE and alike. It runs on any CPU, and takes advantage od AVX-oriented instruction when thy're available. However, for some reason, this version is less efficient than the older one on non-AVX capable CPUs.

So, to make things short, tick only this box if your CPU supports AVX instructions. Leave it unchecked otherwise.


Back to top


What is this test good for ?

This is both, at the time i write this, the best heat-generation test, and also one of the few with error checking. It's not me that says so, it's AMD themselves :)

Why is the test tab greyed out ?

Your DirectX version isn't up to date, or a required file in DirectX is missing. This happens with DirectX9, and with DirectX10 : for some reason, a core file of DirectX10 was removed from Windows 7 SP1 install, while it was present in the 7 SP0 and Vista install.

To get it fixed, you need to update your directX version to the latest one, by using this small utility from Microsoft. Once updated, restart OCCT, and the tests will be available.

Some call it a power virus... and some stock cards cannot withstand it...

Well, believe it or not, it is a pretty simple 3d application. Nothing fancy, i do not real-time overclock your card, i just use standard DirectX functions, and a few shaders. Really, it is surprisingly simple. But it turns out that in 3d, the most efficient is not always the most complicated or visually great !

The test is composed of a 3d model (the donut), one shader for the fur generation (inspired of shaders used in 3d movies), a few squares and textures for the background, and a few textures for error checking. That's it ! So, if it is simple, why all the talk around it ?

That's because Graphic cards makers design their cards with one thing in mind : games. And games do not use 100% of your GPU. Games will use advanced graphic effects, which in turns implies not effective instructions, stalls for calculation and branches, and games also rely heavily on CPUs for calculating the current game state, give them the geometry, data, and so on. If i had to give an estimate, i'd say that your GPU is only working at around 50% of the time when playing a game. Really. The other half is spent at waiting for the CPU to do stuff.

Mind you, not all cards are like that, and things are getting much, much better !

So when designing a card, the manufacturers get a margin around that, and make it so it will be sufficient for 70 or 80%, sometimes 100% GPU usage. So, what happens when you encounter a 3d app, that is not a game, and does not wait for the CPU ? You get it : the card overheats, when it does not sometimes shutdown due to the lack of current it requires for operating.

Am i a power virus ? If so, Prime95, CPU:OCCT, CPU:Linpack and others are too. It is just more blatant as for GPUs, manufacturers pushed the limit too low, and it started to get noticed.

Personally, i think that a card should be able to handle 100% GPU Usage. When i buy a GPU, i see all those beautiful numbers, GFlops and others, numbers you will never be able to reach due to faulty design for cost killing.

In the end, GPU:3D got limited (the card cuts its frequency when it is detected), but i will continue to get past that limitation as possible, because i do think it is right !

The donut doesn't move when i enable error checking mode !

That is intended. To do error checking, i compare each frame against the previous one. If the donut was moving, there would be no way of detecting an error ! If you want it to move again, untick the Error Check Box.

GPU:3D detects errors when moving the window accross two screens...

This is a driver issue, as when you do so, a few frames will be partially rendered by the GPU, and this results in OCCT counting it as errors. To avoid that, do not move GPU:3D window while it is running, or, even better, run it in a fullscreen mode.

Do you support Crossfire/SLI ?

Yes. Just make sure you use Fullscreen mode ! SLI and crossfire are disabled in windowed mode by the drivers.

Which resolution should i use ?

The best is always the native one of your monitor.


Back to top

Power Supply

What is this test good for ?

This test is designed with one thing in mind : trying to get the most out of your Power Supply. And for that, a combination of a GPU:3D on the GPU and Linpack on the CPU is the best.

What about making HD spin up, making the DVD drive busy ?

Compared to a CPU and a GPU power requirements, they are negligible. I do not think this is really required...


Back to top


Is this version usable in my company ?

Yes. While the free version is only free for personal use, this one is allowed in any context.

Which additional features does it comes with ?

Just check the OCCT Pro software description.

I'm a single individual doing overclocking. Do i need it ?

Most likely you don't. OCCT Pro comes with features aimed at professionnal, and that does not include additional monitoring capabilities or additional stability tests. The free version has everything that's required for personal. I made this version cheap for end-users, but the idea is sort of not to sell it to people not in a company that probably won't need it.

How does the custom test work ?

This test is based on batchs. You'll have to define 3 batch file : one that will start your test, one that will close it, and one that'll check if it works. Just make sure your return codes are as follows : 0 is the OK status, any other number is an error, and that for every batch file.

You can find more detailed information and an example here

You mentioned a "Monitoring Only" test ?

In the "Custom Test" tab, just check the "Monitoring Only" checkbox. OCCT won't launch anything as a test, and thus behave as if it was monitoring only. The idea here is to produce graphs without launching anything. You can make it manual by selecting 'Infinite" test (and thus start/stop it manually), or program it using the usual "Custom" test.

How can i customize the graph output folder ?

In OCCT's folder, you'll find a text file named "GraphPath.txt". Edt it and put your own path in it. Environment variables are supported, and a new one has been defined : %OCCT% which points to OCCT's starting folder.


Back to top