VR/Non-VR Benchmark tool

[Voyager]

The CPU wars are coming, and DCS (especially in VR) is CPU limited.

 

You guys should set up a benchmark tool so that all the hardware review sites are talking about how DCS with settings on uber crushes computers like tiny little bugs. Feature the Hornet doing insane things, like dropping walls of JDAMs at a dozen different targets, or whatever else crushed computers the most.

 

Not even sure it would specifically need to be integrated into the current build of DCS, but could simply be an independent standalone download containing a snapshot of the code in time. This could, for example, allow a demo of the Stennis advanced Carrier operation to be shown without requiring the carrier itself to be owned, or even, at the time of snapping the line, ready for prime time, but could involve all of the expected calculations one would see during deck handling a strike package of Hornets. That could be expanded to a psuedo run through of a strike mission, breaking it up into legs, that function and recorded flight snippets, (i.e. the actual engine, purges and starts a new flight track after each stage of the flight, while, cinematicly, appear to show stages of the same flight) allowing each snippet to focus on a specific stage of flight, and providing protection against individual unexpected behaviors from derailing the entire flight run.

 

Thoughts?

[SharpeXB]

I figure the live running of flight physics and AI prevent a recorded track from representing actual game performance. I had tested out scenarios in DCS and from what I can recall that was the case. Tracks give better performance than running the game live.

[Voyager]

That why I'm thinking a canned demo might work better: basically snap the line on a specific build and code could help ensure consistent results, even if the calculations are running live.

 

Or, one could just set up a mission that starts cat launching F/A-18's until a minimum of 20 or them are airborne and in the pattern. Do it at a 15 second interval, and that should get a lot of destructible elements going too...

[SharpeXB]

I usually do one of the canned missions and try to make identical runs.

[Voyager]

True, but you can only do one of those at a time. A good bench marking tool can be set up on a batch of different test beds and run multiple times for averaged results and a multiple settings for identifying bottlenecks. Think the sort of reviews that anandtech or linustechtips does.