Maxwell Render | Official public forum

Hello,
i've got a request for a feature, if it's technically possible: a hardware independent benchmark.
this one could be a good indicator for how far a scene is optimized...
eg. if you are playing around with displacement, which makes the benchmark go down very fast, but until now it's always only comparable, if rendered on the same machine. So this benchmark would also be a good measurement for the complexity of the szene.

cheers,
kami

Well... what you ask is the opposite of the common benchmark idea...

Wikipedia wrote: Benchmarking is usually associated with assessing performance characteristics of computer hardware

So, instead of running a standard scene on a different hardware you propose a number that define the complexity of different scenes runing on a standard hardare?

yes, more or less...
but it's not the opposite of a benchmark like you assume. It's intended as a benchmark for the scene, not for the hardware.

Wikipedia wrote:Benchmarking in simple is the process where you compare your process with that of a better process and try to improve the standard of the process you follow to improve quality of the system,product,services etc.

when you upload a material to the mxm database, you can add a lot of unnecessary information about SL, your machine, benchmark etc. My point is: the benchmark number of mxcl is useless, since you can't compare it with other computers. if you know the benchmark and the hardware on which it was rendered, you could estimate the benchmark on your machine, but it would be a fairly rough assumption.
so what would be the use of this new benchmark? since the introduction of displacement, a lot of new materials are beeing created with the use of displacement. their rendering performance is strongly dependent of the detail, reflexions, etc... so not every material renders more at more or less the same speed. there are huge differences. but there is no value to compare them to each other!
how would this benchmark number help? since my hardware is rather weak, I have to use materials which don't take ages to clear up.

other purposes? eg. one user uploads an image into the wip-section complaining about the render time ("20h and still a lot of noise"). Some other user says: "after 20h this image should be clean". The reason for this image beeing so noisy after 20h could be either a very old hardware on which it was rendered or the inefficient scene setup. With a "scene benchmark" you could better evaluate the problems...

hope you can understand what i mean, and what would be the use of it.
sorry for my weak english. i'm trying my best
greets,
kami

Since we're getting scientific here, you can probably find a linear relationship between benchmarks found on benchwell for various processors and get a formula that converts someone else's benchmark to what it would be on your processor +- some percent error.

That said, I'm not sure if displacement test correlates to the standard test scene in terms of benchmark scores, maybe if there was a standard displacement scene for everyone to test.

Moreover, there could be a built in machine benchmark that tests how much time it takes for your processor to do a given amount of work, say standard scene to sl 10, and compares that to some standard. Thereafter benchmarks can be displayed as the [standard number] +- [your constant].