1st Time Running Faster Than All Subsequent Benchmarks?

[mitchgates]

I'm seeing a strange pattern when I run VRay Benchmark 1.0.8 on a few different systems in both CPU and GPU tests. My brand new Threadripper2 2990WX (Win10) clocked the first CPU test at 26.9 seconds, but every time I've run it again it consistently hits ~27.6 seconds. The GPU test behaved similarly, reporting 44.5 seconds on the very fist run but always coming in at 45.9 every time I've run it again. I've made sure that there are no other processes running, disabled anti-virus, etc. The 2nd attempts for both tests were run RIGHT after the first set and I didn't launch any other apps or close the benchmark app between them. I tried closing and restarting the benchmark app then running both CPU and GPU tests again several times but the results are always consistent to the 2nd runs and never returning back to the original speeds.

I then did the same on my older dual Xeon Win 8.1 workstation and the same thing happened. First run CPU was 1:56, all subsequent restarts 2:01. Since this is a standalone app there's now way to "uninstall/re-install", but I'm wondering if there are any temp/user files I can delete to return the benchmark back to the original state before any tests were run. The fact it remembers and displays the previous benchmark results when you close and reopen the app leads me to believe it's caching something somewhere.

Does anyone else get this behavior?

[Muhammed_Hamed]

Couple of seconds difference is normal with Vray Benchmark.
I really think Chaos should make this benchmark last longer, specially for GPU. for me, using 4 1080 tis I get 21 seconds and using 5 1080 tis I get 20 seconds render time.. Adding more cards I only get 20 seconds render time
The benchmark itself is not really accurate with super high end hardware. Same happens with Cinebench once you hit 4500 points or so, you will start getting very different results with each run
Things like Corona Benchmark for CPU and Octane Benchmark for GPU, run longer and provide more accurate results with this kind of hardware.

[mitchgates]

Agreed on all points. What I find odd is that the very first test run on a specific system gives a better score than any subsequent runs. And several subsequent runs for me are VERY close to the same result. The only outlier is the first one.

[Muhammed_Hamed]

Really odd.. I have run Vray bench on all my machines here, but I cannot recall if this happened or not with the first run.
I would like to know why this is a thing in Vray bench
Currently I get the same score for each run on Mac OS/windows