OCCT does use AVX

A lot of you may know this already but I discovered this morning that OCCT does engage the CPU's AVX instruction set. I'm not talking about the Linpack tab with AVX option checked, I'm talking about the main OCCT tab.

I emailed OCCT with this question a couple of weeks ago but haven't heard back from them. So I was experimenting with the AVX Ratio Offset in my bios for the first time and had HWMonitor open at the same time. I could see that although the CPU core ratio was set to 48x when I ran the OCCT CPU stress test it dropped back to 46x, just like it does when I run Prime95. What I don't know is if OCCT is using the AVX 1 or the AVX 2 instruction set. I tried to register for their forum so as to be able to ask some questions like this but it has been "down for maintenance" for a long time and is probably innactive.

On the other hand, the XTU stress test does not seem to engage the AVX instruction set as it does not cause the CPU ratio to drop. I mention that because I have read where people said that XTU uses the same stress test method that Prime95 does. Well, if that's true it must be a pre AVX Prime95 version that is being referenced.

A three hour OCCT stress test has become my go to OC stability test. It requires less voltage to pass than the newer Prime95 versions (and generates lower temps) but gives adequate assurance that I am stable in the range of applications I use. I cannot say that for XTU which even if I can pass it when run overnight still leaves me with random blue screens and lockups.

AVX2/FMA3 needs more voltage then AVX1, quick way to check is try to encode something x265 in Handbrake or try Prime95 latest version...

So I would guess then that OCCT uses the AVX1 instruction set and not AVX2.

What is the relationship between AVX2 and FMA3 as you seem to present them together in contrast to AVX1.

Honestly no idea apart from needing the same voltage to be stable. As far as I know it's not really used, but everytime I read something about AVX2 its always presented as AVX2/FMA3 ?

FMA3 is a specific set of instructions which were introduced same time as AVX2 on Intel, which has other stuff in it. On Intel at least, if you have AVX2, you have FMA3, but they're not the same thing. I guess it is a bad habit of many, including myself, to use them interchangeably. FMA3 is the magic that makes Intel much faster at Prime95 like applications than anything AMD.

Thanks for the clarification.

What are AVX instruction sets? Do many programs use these? If not, why would I use a torture test that does?
In the EUFI bios for my motherboard, there is a setting in AI Tweak that will adjust the core ratio when it detects an AVX workload. Apparently, an AVX workload can seriously heat up your CPU. So I set this offset to 1, as I am trying to get a stable OC at 4300 and 1.3v adaptive. Now when I torture test on p95 it runs at 4200 and 1.29v. Seems logical, detecting AVX.
And if I test on cinebench, it runs at 4300mhz and 1.3v. Awesome. Is this cheating the OC test. I mean, can I still say I am OCing at 4300 and have stability when I am using this setting to dance around AVX in P95?
Also, does having a 6 core chip make it more sensitive to AVX heat than a 4 core?

AVX can provide a significant performance boost for software that uses it. It can run hotter, but it is doing more work. And that is where arguments start... do you need to be stable with particular software if you never use certain CPU features? Even if you don't use any now, who knows what you might use in the future. Personally I'd like to know my system will be stable no matter what software I ever run on it, so I'd make sure everything as far as possible is stable.

6 cores will simply produce 50% more heat that an equivalent clock/voltage/architecture quad core. Just make sure your cooling is up for it.

There's 2 schools of thought here, one that says "your PC must be stable for everything just in case" and the other states "your PC only needs to be stable enough for your daily use". Either has it's valid points and to be fair it will always be entirely up to you.

stable for everything just in case : Stress test for every possible scenario including AVX/AVX2/FMA3, higher voltage and temps = lower overclock. Prime95 - 2h bare minimum, 24h to guarantee 100% stability

stable enough for your daily use : Stress test only for whatever programs you use, like gaming or web browsing, lower voltage and temps = higher overclock - Aida64/RealBench/OCCT - 2h bare minimum, 6h+ recommended

Thank you for the insight. I'm very new to overclocking and I feel like I am just scratching the surface with this. Using AVX core ratio offset in my bios I can get a higher default overclock and still be stable because it will back off automatically when an AVX instruction set appears. That about right?