Kategra wrote:
The purpose for this particular test was to find the next best CPU performance/$ for users like myself that already have a computer which can play the whole R8+ benchmark song. So when I want to upgrade the i7 6800K computer , I don't want to second guess (like it happened with my purchase of Ryzen R1700X--- my fault for thinking that good Cinebench performance=good Reason perfomance), but to get the best deal for my money.
If someone else would make a better "benchmark" song that could be used to gauge the relative performance between older mainstream CPU (SandyBridges dual, quads, older AMDs) and present enthusiast/ server CPUs than that would be awesome, I would use it and I would not feel wronged if this thread was deleted or locked to not have confusion.
My point is that any CPU that can play the whole R8.1 song can also play at least 20 seconds in this new R9.5 song and there is still room left testing the new 16 core Ryzens and 12 to 18 core i9s that will be launched this year.
I may be repeating myself: I don't think that you can make a song with Reason Devices + stock REs + Players + Automation (a formula that I use, and I would take a guess that so does the majority of R9 users) that can accurately measure the ratio of Reason performance between Intel i5-2500k/AMD FX-8320 and the upcoming Intel i9 7980XE/AMD Ryzen 9 Threadripper 1998X.
IMHO, any benchmark song with various devices and automation that that will show "Computer too slow to play song" for the i9/R9 will not play even 2 seconds on the dual core/quad core sandybridge/vishera.
You've already made a better 'benchmark' than this, the R7 one!!! You just need to make it longer to the point even the best CPU's wont make it even if it takes 20 minutes to test. In fact the longer it takes the more accurate the comparisons you can draw because the load introduced on each bar doesn't increase in such massive steps.
Just take a look how DAWBench DSP is set up and meant to be used, it doesn't use a bunch of fancy plugins designed to overload the system at start-up just a well selected bunch of half-dozen or so that use resources in different ways that get added to incrementally so even from the humblest setup to the latest record breaking system can run it and comparisons can be drawn, between anything. It hasn't changed since 2014 and it will eventually wipe out even the best system today, ScanUK among others use it to test their latest 3XS systems and because the benchmark hasn't changed there's nearly 4 years worth of comparison data to draw on. Likewise the R7 benchmark is still sufficient.
Similarly the R7 bench test can be used to successfully test performance changes between versions from R6 - R9.5 if it was a constant. It can also show users how much improvement a hardware upgrade has made...again if it was a constant, and you'd be able to use it successfully on any machine. Certainly though you'd be able to get a picture of how different CPU's stacked up against each other in this context better than you could with Cinebench or any other non-Reason based benchmark.
What can't be done in Reason with any accuracy sadly is test like for like performance from any other DAW to Reason as Reason is a pretty unique animal in that it carries around a rack full of stuff before it even does anything and most other DAWs use plugin's for their main functionality but we will be able to chart improvements to VST performance from this point on.
I'm not trying to pee on anyone's fireworks here, and I really appreciate the amount of work that you put into this stuff, but changing the test each time a new version comes out negates much of it's usefulness and worse in this case when you are running a current CPU and your DSP is almost maxed out before you even press play it could cause a lot of unecessary anxiety to someone who's just spent the best part of a grand on a shiny new box just to watch their prized new machine limp to a little over a few seconds when the truth is they are going to be able to make lots of great music regardless.
Perhaps you could just add the old R7 version to the OP and call it ReasonBench 'Classic' after extending it a bit (or providing users instructions how to do so) so higher end CPU's don't defeat it, and call your latest one ReasonBench 'Extreme' then people could just state which test they are doing when they post results.