Disaapointed in FPS after upgrading video card

Joël

Volunteer Moderator
The thing is alot of us including me upgraded a few years ago to already a strong CPU mainly for this game. I bought a 6700k for about 400 euros 2 years ago and remember being disappointed when the fps didn't improve at al in big parks. I'm not gonna upgrade again to a 9900k which costs about 550 euros for something that probably won't make much of a difference either.

We are ignoring the problem. The problem is no DX12 support. You can upgrade all you want, run software all you want, install drivers all you want but if you don't grab the problem by the source it won't ever improve.
Upgrading the game with DirectX 12 support is not going to magically increase the FPS. It will most likely result in performance improvements, but it probably won't be the huge improvement that some people seem to think.

However, DirectX 11 is indeed the main bottleneck of the current performance issues when it comes to FPS; DirectX 11 can only use a single thread for its DirectDraw calls and since Planet Coaster has an enormous load of DirectDraw calls to execute it is slowing down overall performance and FPS of the game. DirectX 12 is capable of multithreading its DirectDraw calls and this should in theory result in a performance improvement.

Because of this the general advice is to use a CPU with a single core clock speed as high as possible, since higher speeds will benefit the single thread used by DirectX 11 DirectDraw calls.
 
I bought a 6700k for about 400 euros 2 years ago and remember being disappointed when the fps didn't improve at al in big parks.
Explicitly speaking, for this game, the 4770K out-performs the 6700K.. check any benchmark website and you'll see that the 6700K is actually quite a bad one..

And again, Explicitly speaking for this game, you'd be better off with an i7 7740K, i7 9700K or the i9 9900K as these are the fastest single cores available (be careful of the 9900K though because as it utilises more cores the base frequency reduces in steps.. where-as the 7700K doesn't)


Adding to the DX12 debate.. It's not beyond the realms of possibility that Planco can be released to retro-fit the game.. Cities Skylines was released as a Windows 10 version before hitting consoles - you could update to that for free if you owned the old version of the game, but Joel is correct - you don't get the performance uplift you crave from going up to DX12. It's not a miracle solution that makes your 16,000 guest park run at 60fps.
 
Explicitly speaking, for this game, the 4770K out-performs the 6700K.. check any benchmark website and you'll see that the 6700K is actually quite a bad one..

And again, Explicitly speaking for this game, you'd be better off with an i7 7740K, i7 9700K or the i9 9900K as these are the fastest single cores available (be careful of the 9900K though because as it utilises more cores the base frequency reduces in steps.. where-as the 7700K doesn't)


Adding to the DX12 debate.. It's not beyond the realms of possibility that Planco can be released to retro-fit the game.. Cities Skylines was released as a Windows 10 version before hitting consoles - you could update to that for free if you owned the old version of the game, but Joel is correct - you don't get the performance uplift you crave from going up to DX12. It's not a miracle solution that makes your 16,000 guest park run at 60fps.
When I bought the 6700k it was the best one at that moment. Didn't know the 4770k would've outperformed it back then but still I think for other games and tasks the 6700k was the wise decision.
 
We are ignoring the problem. The problem is no DX12 support. You can upgrade all you want, run software all you want, install drivers all you want but if you don't grab the problem by the source it won't ever improve.
Since the OP refuses to upgrade to Windows 10 which is a requirement for DX12 he is left with the only option of trying to get the best hardware supported for DX11. His i5 is not top of the line (i5 never has been) not as good as an i9 at 5GHz. For him, it would not matter if they go to DX12 because he would stay where he is now.
 
I do have a very good single core CPU. All my cores are locked to 4.2GHz, and this CPU was out when they would of been designing the game and when the game released the CPU I am using was still just as good as what was being release by Intel. When this game came out and was being developed, quad core CPU's were common and what was being released, so there is no reason why this game shouldn't be able to utilize 2-4 cores easily.
Doesn't matter how many cores you have if you still can only use one for DriectDraw calls. That is the biggest limitation with DX11. You can get a CPU with 16 cores, would not make any difference.

I will upgrade to Win 10 when H**L freezes over. I hate that OS with a passion. I have to use it at work and do not like it. I will run Windows 7 until at least next year when security updates stop. Windows 7 is a far superior OS and I have been using it since it came out. At work with windows updates, they have had nothing but problems, a couple systems they lost data, and a couple others they had to do complete reinstalls. Never had those problems with Windows 7.
With this stance you will limit yourself more than the games going forward. Doesn't understand this hate against Windows 10. It is one of the best, if not the best, OS Microsoft has released. If the issues you are describing woul dbe common you would see a lot more fuzz about it all over the place. Why is Windows 7 far superior? Maybe it is the IT department at your work that you should hate with passion?

A game that I can compare this to to Two Point Hospital. That game has a lot going on in it and I can easily get 60FPS at all times, SimCity 4, Cities:Skylines is another one I compare it to and I easily get over 30FPS in my biggest city at max detail. as well as OpenRCT2.
You are comparing this to Two Point Hospital? Seriously? I enjoy the game and reminds me of the great game Theme Hospital but the requirements of the computer hardware are nowhere near the same as for PlanCo. In Two point you have maybe a few hundreds patients at the same time where PlanCo is counted in the 1000s or more. You may also have objects in the 100s, maybe 1000s if you go crazy with equipping the rooms. PlanCo, for parks that starting to slow down computers, you have 100,000s or even millions of objects. Two Point also have a very limited size where you can build, extremely limited in comparison.

City Skylines, another great game, but the graphics is very limited and low res compared to PlanCo. Don't get me wrong, it does look good and I truly enjoy the game, but the graphics are not the same. The simulation model is not on the same level either with everything being "fixed". No decisions "on the fly" etc.

I am not even going to comment on OpenRCT2 as that is a ridiculous game to bring up in a discussion like this. Of course a game that originally came out in 2002 will run fast on today's computers. The OpenRCT version won't change that.

We need to compare apples to apples for this to be a meaningful conversion.

I don't think anyone here don't want to see performance improvements. Everyone would support going to DX12 etc. The finances (and time which are closely connected to each other) are the issue. Frontier is a small company (in comparison to Rockstar with their GTA series that was brought up as an example). The install base for PlanCo simply is just too small for big risk (financially) re-writes like this.
 
City Skylines, another great game, but the graphics is very limited and low res compared to PlanCo. Don't get me wrong, it does look good and I truly enjoy the game, but the graphics are not the same. The simulation model is not on the same level either with everything being "fixed". No decisions "on the fly" etc.
Well, to start with graphics, that´s not an excuse. It´s developer´s decision to make the game on certain graphics quality and at the same time, keep it playable. Cities Skylines succeed, PC failed (althought I think the graphics itself is not the major issue here). Nobody forced Frontier´s to make this good graphics and sacrifice the performace, which is really not very smart move.

About the simulation, I think PC doesn´t rly have very complex simulation. Most of the stuff is pre-calculated and the major thing is the AI here, which is using agents, so if you have 10000 guests, game bscly counts with like 2 500 guests. Before the release, Frontier also said the game will scale (probably the AI) according to our HW and that is not happening at all.

In the end, I don´t get why not make the game so it´s playable on at least on the hardware that is considered high-end at the time of the release. Why should I pay for developer´s not having reasonable and realistic expectations? I wouldn´t mind if the game did some compromise in graphics or simulation, if it was playable.

I bought Early Bird for quite a lot of money and I will be happy if I will be able to play the game in like... 5 years (in the mean time, it will be on sales for 5 EUR every once in a while).
 
I'd love to know how much performance would improve if they let guests just clip right through each other and weren't constantly calculating pathing congestion. If it's a decent performance increase, I'd likely use it if there was a "Guest Clip" option under Graphics.
 
I'd love to know how much performance would improve if they let guests just clip right through each other and weren't constantly calculating pathing congestion. If it's a decent performance increase, I'd likely use it if there was a "Guest Clip" option under Graphics.
That would reflect badly on their game though if people were streaming/videos were uploaded with this enabled having the guests just as they were in RCT3. Whilst it may help with performance, the game's realism/quality would take a hit and I can't imagine Frontier would want that. This isn't RCTW remember that [mouth shut]
 
That would reflect badly on their game though if people were streaming/videos were uploaded with this enabled having the guests just as they were in RCT3. Whilst it may help with performance, the game's realism/quality would take a hit and I can't imagine Frontier would want that. This isn't RCTW remember that [mouth shut]
Yes exactly, they don´t care how the game plays, but how it looks. [down][down][down][down]
 
I'd love to know how much performance would improve if they let guests just clip right through each other and weren't constantly calculating pathing congestion. If it's a decent performance increase, I'd likely use it if there was a "Guest Clip" option under Graphics.
So do I. I'm also interested in "dumb guests", as in guests that just walk around randomly doing stuff randomly instead of constantly calculating what they wanna do, how to get there and what their needs are.

It would be my expectation that this would probably improve the most in performance if it were possible. They could just make it an option in the menu to "disable smart AI behaviour" or something.

Hopefully a dev could shine some light on this.
 
To begin with, it would be enough to make calculations less frequent. Just have a look at how "not smart" guests are. They go somewhere, but decide to do something else even before they get to do the thing they originaly planned to do. So they change their route again (they prefer to go back where they came from or at least the far opposite of the park :D) and it´s not that rare they they don´t make it again and in the mean time, they change their goal again... Just reducing this by half would save a lot of performance in my opinion...
 
To begin with, it would be enough to make calculations less frequent. Just have a look at how "not smart" guests are. They go somewhere, but decide to do something else even before they get to do the thing they originaly planned to do. So they change their route again (they prefer to go back where they came from or at least the far opposite of the park :D) and it´s not that rare they they don´t make it again and in the mean time, they change their goal again... Just reducing this by half would save a lot of performance in my opinion...
Yeah making the calculations less frequent could possibly help, but why all the advanced calculations to begin with? I guess 3 years ago when we first heard about Planet Coaster and guests AI we were all very excited how advanced every single guest is, but we didn't know the cost back then.

Why would they even need to have a "will" anyway? I'm guessing it adds for realism, but does anyone really ever notice? I never really care about the bladder or hunger of an individual guest, all I care about is having a nice, full theme park.

So if the guests were to just wander around randomly like ants and occasionally visit rides and shops they walk passed by, I would probably not even notice, and I think the same goes for alot of people.

If it could save alot of calculations and thus performance, this has got to be the most important request I could come up with.

So hopefully a dev could respond and tell us why "disable smart AI behaviour" as an option in the menu would be a good or a bad idea performance-wise.
 
Yeah making the calculations less frequent could possibly help, but why all the advanced calculations to begin with? I guess 3 years ago when we first heard about Planet Coaster and guests AI we were all very excited how advanced every single guest is, but we didn't know the cost back then.

Why would they even need to have a "will" anyway? I'm guessing it adds for realism, but does anyone really ever notice? I never really care about the bladder or hunger of an individual guest, all I care about is having a nice, full theme park.

So if the guests were to just wander around randomly like ants and occasionally visit rides and shops they walk passed by, I would probably not even notice, and I think the same goes for alot of people.

If it could save alot of calculations and thus performance, this has got to be the most important request I could come up with.

So hopefully a dev could respond and tell us why "disable smart AI behaviour" as an option in the menu would be a good or a bad idea performance-wise.

I would notice it, because it would probably feel a bit more realistic than these completely random calculations....
 
I think if you do that, it eliminates every aspect of it being a game.

If they walk around as mindless zombies it wouldn't matter how much you decorated the coaster, or how much the EFN values are of rides.
Or how many bins, or benches or all that other stuff you place, that has a meaning in the game.

I think there is to much stuff going on to just eliminate it.

Maybe "to much stuff going on" is the same reason some people have performance issues that's true, but I don't know.

A dev clarifying this could be helpfull (Andy C used to have good comments on these kind of things).
 

Joël

Volunteer Moderator
Regarding the guests of Planet Coaster, and specifically the 'Guest Brain', there are some nice video's about that subject, such as these Dev Diaries:

Definitely worth watching these if you want to learn more about how guests in the game are designed and how they work.

[video=youtube_share;7tyWUT0UsZE]https://youtu.be/7tyWUT0UsZE[/video]

[video=youtube_share;cuQh7IJCrs4]https://youtu.be/cuQh7IJCrs4[/video]
 
The principle of a guest heading some-where to change their mind is relatively realistic in approach.. People do that in parks all the time - but ultimately they end up going to their originally intended destination. For example, you head to a coaster but on your way decide to get a drink then carry on heading to the coaster.

I've noticed in the game however that guests bounce between thoughts and yes, they change their mind far too frequently - to the point that it cripples them into indecision. Watch if a ride breaks down and the guests are all at the exit - They will flip rapidly between finding a bench, filling a food / drink / toilet need and heading to a new ride. I've seen them change their mind 20 times in a minute before - and they do this in packs, so you can see a visible and very noticeable wave of indecision. Now, this creates a few questions as it seems that the game is already trying to handle all of these decisions in an efficient token based manner, which would mean that the only solution to this is to reduce the frequency in which guests change their minds.

You can kind of counter this by setting up a scenario with the trait of "Single Minded" set to high and "indecisive" set to None. But.. the issue still persists.

I posted about this a while back in a thread called "I wish the AI was more Intelligent" where I discussed the need for a guest to be aware of things before doing them. For example, the guest should be aware that a queue is full or too long before they get there and have to change their mind - or know that they will need food or drink between queue and exit of a ride and so will sort this need before heading to the ride etc. Whether this would positively or negatively impact performance is only known to the Devs though. It seems that a single guest need is calculated and then applied to the group.. So, a group of 6 SEEMS to have 6 calculations which influence the actions of the group - SO, what happens if the system is changed from individual to group needs? You may have 10,000 groups in a park of 25,000 guests. Would that reduce the load on the CPUs and have a positively noticeable affect? Instead of 25,000 individual calculations, you now have (say) 10,000 instead. Surely that's more efficient?
 
I agree that sometimes it seems a little excessive when the guests changes their minds halfway through a walk to a destination or that they go to a ride they can't afford, but not discovering this until they have arrived. Maybe something can be done to tweak that. But, at least I , don't want the feeling that a zombie or robot visiting my park. The guests today feels like "humans" more so than any game in this genera before (not saying they are perfect, far from it).

As been said above, taking away decision making would mean that everything else in the game means nothing and it will become even more a "building game" than what it is today. Ride, stalls, amenities etc. placement would not matter. Decorating rides and queues would not matter, keeping people happy would not matter etc. It would take away the simulation that actually is in the game (you can argue if it is a lot or a little but that is a different discussion).

Part of the performance issues could easily be handled with "restrictions" that many other games has (even Cities" Skylines has limitations built in by default on how many trees, objects etc. you can place. Mods have been created to remove such restrictions, but at your own risk and there are players that have experienced slideshow playing with that game after some time or after enough mods and assets being enabled). In PlanCo you can add as many items as you like to the game. Of course, with such limitless possibility comes the drawback of hitting a ceiling when your computer can't handle it anymore. It will happen regardless of if the game is perfectly optimized in a perfect world. It is inevitable. But putting limits on items, guests, rides etc. would create huge outcry among the majority of the players and not a good solution.
 
I think if you do that, it eliminates every aspect of it being a game.

If they walk around as mindless zombies it wouldn't matter how much you decorated the coaster, or how much the EFN values are of rides.
Or how many bins, or benches or all that other stuff you place, that has a meaning in the game.

I think there is to much stuff going on to just eliminate it.

Maybe "to much stuff going on" is the same reason some people have performance issues that's true, but I don't know.

A dev clarifying this could be helpfull (Andy C used to have good comments on these kind of things).
It's false that it eliminates every aspect of it being a game. As far as I know guests in RCT1, RCT2 etc also walk around pretty randomly and aren't nearly as advanced as the "smart guests" in Planet Coaster and that still feels like a game too, right?

Also it is completely optional. So it wouldn't be turned on by default. It would just be optionable to help improve performance. For a sandbox game on PC you should always give the players options to make the game as playable as possible and let them play it how they can enjoy it the most.



Regarding the guests of Planet Coaster, and specifically the 'Guest Brain', there are some nice video's about that subject, such as these Dev Diaries:

Definitely worth watching these if you want to learn more about how guests in the game are designed and how they work.

https://youtu.be/7tyWUT0UsZE

https://youtu.be/cuQh7IJCrs4
Yeah but the point is that this might not have been a good idea since it is apparently so taxing for your CPU to the point to which the game becomes unplayable. When I first saw those videos 3 years ago of course I got as hyped as everyone else, but back then you didn't know what it would do to your actual gameplay experience when the game was released and full of content.
 
Top Bottom