CBJ wrote: ↑Sat, 13. Oct 18, 14:19
Maybe a little. As I've explained before, the CPU is mainly busy with simulation
We can test that though easily by going to busy and empty region. Pause button also should stop simulation without interrupting graphics (thought it might stop animations but that should be minimal impact)
- 100fps empty region - 122fps paused (70% gpu used)
- 30fps very busy region - 35fps paused (30% gpu used)
Rule of thumb is that the more objects on screen the less performance which shouldn't be really any surprise.
What surprising is small impact of economy simulation and GPU low use.
The more things on screen the lower GPU usage and the lower frame-rate.
We can further test that with changing resolution like i did before.
- @1080p - 30fps in busy sector - 30% gpu use
- @4k - 30fps in busy sector - 70% gpu use
So the limiting factor here is obviously CPU like you mentioned.
In this case changing resolution doesn't add draw calls which means GPU% increases with res.
Which pretty much means CPU is hammered by mentioned draw calls.
Simulation does have impact but not that much.
And this is where i don't quite understand second part about Vulcan which is why i asked about it.
Main reason low level API like Vulcan or DX12 was made was to solve draw calls issue + other things.
Pretty much any FPS improvements for games that changed to those new low level APIs was precisely draw calls batching which basically reduced massively draw calls in various ways like batching and basically made removed need for fast CPUs to run GPU fast.
Obviously engine developement is ongoing thing and just by moving from DX9 to Vulcan doesn't mean anything improved as that move right now could be just move to start working on new API and then piece by piece improve engine.
Either way. I hope you will manage in future resolve draw calls issues
Cheers.
CBJ wrote: ↑Sat, 13. Oct 18, 14:19
Perkel wrote: ↑Sat, 13. Oct 18, 11:59
Worth noting is that a lot of initial bad reviews for X:rebirth was from people
That's a relief. I was beginning to think that they were all from bots.
lol. I didn't finish sentence for some reason.
My point was that by reading your and other ego posts i get that performance won't be improved much from X:R and i think a lot of early reviews you got were about very poor performance on top end rigs or that performance contributed a lot to initial feelings people got from other flaws.
So imho as developers you should inform clearly people about expected performance because it would have a lot to do with reviews you will get just like in case of X:R where a lot of people expected at minimum average performance like 60fps on top end rigs and got 20-30fps with dips to 15-20fps for period of time in busy sectors. I myself back then had q9300 which was at the time pretty good cpu + hd7970 also v.good gpu and i basically had to play at 30fps average.
So by setting up expectations properly you might lose some buyers but you will get better reviews.
For example adding to specs:
minimal spec: 900p- medium - 30fps average
recommended 1080p- ultra - 60fps average (or whatever your prediction is)
Would give you a lot of goodwill when people will have game in their hands and when they will start to write reviews.
Other idea is to create FPS profiles. If you for example expect similar performance to X:R then maybe setting up 30FPS limit cap as default setting (which could be changed in options) would limit people bed experience with varied framerate at 60fps cap or Vsync cap which could give people a lot of bad time with very varied framerate.