Decreasing performance from Corona Benchmark to Corona in 3Ds max

[Polaris]

Trying to get the maximum from my processor, after some simple overclocking, I noticed a disappointing negative difference in CPU performance between Corona Benchmark and Corona in 3Ds max.

Seems like Corona in 3ds max renders on lower frequency (about -0,3 GHz) but with extra CPU core voltage (over the safe threshold of 1,4 V) in comparison to Corona Benchmark rendering. Both IR and straight render, no difference between different scenes or if running Max as user/administrator.

Screenshots in attachment :
- Corona Benchmark : 5 GHz – 1,376 V – 70 C after 1 minute rendering
- Corona in 3Ds max : 4,7 GHz – 1,440 V – 75 C after 40 seconds rendering (lower performance, higher voltage and temperature, less stable)

Any ideas how to overcome these illogical performance limits ?..

[sprayer]

And what is not logical?
How you comparing speed of different coronas versions and different scenes?
Temperature is fine, you set high voltage and it gives you higher temperature, how you set multiplayer at cpu it's not counts

[Polaris]

Sprayer, thank you for reply. You’re right, it’s not obvious to compare the rendering speed on different scenes.
But what counts here is the reached CPU frequency during rendering. Being overclocked to reach the target 5GHz with adaptive voltage, is produces fine results in bench but lower performance in 3Ds max Corona.

you set high voltage and it gives you higher temperature

As you can see on the screenshots, the OC settings are exactly the same between Benchmark and 3ds max version.

Update : done some tests in CineBench, the result is the same as Corona Benchmark : stable 5GHz with adaptive voltage under 1,38 V.

As a result, I have my CPU performance decreased in Corona 3Ds max version compared
to Corona Benchmark and other software.

[Nejc Kilar]

I could be wrong but didn't we figure out that newer Corona versions use AVX instructions while the Benchmark doesn't?

[maru]

I'll try to get some answers, but generally I wouldn't worry about the differences between benchmark and regular rendering.
We will also have a new benchmark at some point in the future, which should be more "relevant" to regular rendering.

[maru]

Some time has passed, and I'm back.

The possible explanations are:
-Difference between different versions of Corona used for the benchmark and inside 3ds Max
-There may be some differences between the Benchmark and 3ds Max plugin - for example AVX
-Performance measurements may not be accurate (e.g. we can see the total CPU usage, but we can't see how much Corona/Benchmark uses)
-There could be some other app running in the background

We will also update the benchmark at some point in the future, but it is not high priority for us.