Big World performance

Slapshot1188

OK I had a chance to pop into the "Big World" last night and was really impressed with the responsiveness.   One of the issues I had with prior iterations was very poor FPS.  I didn't blame the game, but my system is a bit old so I was just hoping for something playable at release.  Well current performance is WAY beyond that.  I was getting a solid 60 FPS consistently.   My rig is below:

3570k (overclocked to 4.1Ghz)
16 GB RAM
Radeon 7950

Of course that will probably go down once there are big battles with effects going off all over the place, but the step change from what I previously experienced was unexpectedly great. 

MaxBacon

Yup it's looking better; more stuff keeps rolling in to fill that world and then we can start getting what will be the gameplay loop the whole thing will play as.

genaknosc

I feel like the initial pass addition of graphical settings options is helping some people with that. I don't see much difference in setting it to medium quality settings, and FPS is much improved.

Slapshot1188

genaknosc said:

I feel like the initial pass addition of graphical settings options is helping some people with that. I don't see much difference in setting it to medium quality settings, and FPS is much improved.

Yeah.. whatever they did made a HUGE difference for me.  I hope it stays like this!

JamesGoblin

Besides large scale fights (I mean really large, hundreds in small / battleground sized area), another critical test will be what happens during sieges. When they started testing (voxel) destruction some half a year ago, FPS went to hell and only partly recovered.

francis_baud

JamesGoblin said:

Besides large scale fights (I mean really large, hundreds in small / battleground sized area), another critical test will be what happens during sieges. When they started testing (voxel) destruction some half a year ago, FPS went to hell and only partly recovered.

Crowfall doesn't feature voxels afaik.

Volgore

Does the game actually have a "big world" or is it more like PvP areas/zones?

*curious*

francis_baud

From an awesome thread created by Kraahk

botrytis

francis_baud said:

From an awesome thread created by Kraahk

So pretty small parcels.

JamesGoblin

botrytis said:

francis_baud said:

From an awesome thread created by Kraahk

So pretty small parcels.

@botrytis If you put together tens of thousands of these, it becomes kinda sorta bigger.

francis_baud said:

JamesGoblin said:

Besides large scale fights (I mean really large, hundreds in small / battleground sized area), another critical test will be what happens during sieges. When they started testing (voxel) destruction some half a year ago, FPS went to hell and only partly recovered.

Crowfall doesn't feature voxels afaik.

Furthermore, it features no destruction in current tests.

What might happen is that both voxels (as described in the FAQ and as seen in front page video) and these ˝bloxels˝ you linked simply get removed and replaced with good old pre-arranged destruction, because of the bad influence on (already very critical) Unity performance. It seems like one of the easier parts to cut off, if things become really bad.

francis_baud

JamesGoblin said:

botrytis said:

francis_baud said:

From an awesome thread created by Kraahk

So pretty small parcels.

@botrytis If you put together tens of thousands of these, it becomes kinda sorta bigger.

francis_baud said:

JamesGoblin said:

Besides large scale fights (I mean really large, hundreds in small / battleground sized area), another critical test will be what happens during sieges. When they started testing (voxel) destruction some half a year ago, FPS went to hell and only partly recovered.

Crowfall doesn't feature voxels afaik.

Furthermore, it features no destruction in current tests.

What might happen is that both voxels (as described in the FAQ and as seen in front page video) and these ˝bloxels˝ you linked simply get removed and replaced with good old pre-arranged destruction, because of the bad influence on (already very critical) Unity performance. It seems like one of the easier parts to cut off, if things become really bad.

With all the time and effort they've spent on this particular field (first with the fully destructible world that Voxel Farm was supposed to provide, then with the different attempts to solution the wall destruction with acceptable performances), the most likely outcome imo is that they'll keep this tech they've been working on but modify it in a way that there is less calculations required, such as making the blocks bigger and fewer, removing all the unnecessary parts, and making only some walls destructible.

It's far from what was pitched during the Kickstarter (everything is destructible: the ground, the structures, etc.) but with Unity it's not surprising that Voxel Farm or other demanding tech kill the performances.

JamesGoblin

If things get rough, the criteria won't be "we invested so many hours in this..." but rather "we either remove it or we fail"

I hope that it won't happen, of course. But some kind of toning down seems mandatory, we agree on that.

nurso

JamesGoblin said:

because of the bad influence on (already very critical) Unity performance

francis_baud said:

but with Unity it's not surprising that Voxel Farm or other demanding tech kill the performances

So, would it be possible (theoretically, ofc!) to replace Unity with, let's say, Amazon Lumberyard without losing to much work? And would this be a good idea?

Relampago

Not really.

JamesGoblin

nurso said:

JamesGoblin said:

because of the bad influence on (already very critical) Unity performance

francis_baud said:

but with Unity it's not surprising that Voxel Farm or other demanding tech kill the performances

So, would it be possible (theoretically, ofc!) to replace Unity with, let's say, Amazon Lumberyard without losing to much work? And would this be a good idea?

@nurso I'm not an expert, but I guess that would cost them quite some time (= money), and they don't have Star Citizen budget. Whether Lumberyard is the right engine for a MMO is additional problem.

esc-joconnor

Star Citizen could pull off the switch because they were using the Cry Engine, which is what Lumberyard is based on. No easy road from Unity to Lumberyard.

Dullahan

If they are having problems with those graphics using Unity, the problem is them, not Unity.

nurso

esc-joconnor said:

Star Citizen could pull off the switch because they were using the Cry Engine, which is what Lumberyard is based on. No easy road from Unity to Lumberyard.

Oh, I see ... and that's the reason why I've asked, because of SC successful switch ... didn't know that Lumberyard is related to Cry Engine.

DMKano said:
The voxel data overhead is unavoidable, just look at Trove and how much bandwidth heavy it is with 50 players on screen building/destroying the world simulataneously. Ive seen my client hit 5+ Mb when in DAoC even with 300 player battle I wouldnt even hit 300Kb

Camelot Unchained does also feature destructible buildings, but the physics are processed on the servers ... would that reduce the overhead?

Kunai_Vax

DMKano said:

JamesGoblin said:

Besides large scale fights (I mean really large, hundreds in small / battleground sized area), another critical test will be what happens during sieges. When they started testing (voxel) destruction some half a year ago, FPS went to hell and only partly recovered.

This is due to voxel stream overhead.

non voxel games only deal with player actions and movement between clients and server, the world doesnt change so no need to exchange that data.

Once you introduce voxels and suddenly the server has to send geometry/voxel data update each time anything ia built or destroyed, and this has to be sent to all clients in the vicinity. This is in addition to player movement, actions and skill usage etc...

The voxel data overhead is unavoidable, just look at Trove and how much bandwidth heavy it is with 50 players on screen building/destroying the world simulataneously. Ive seen my client hit 5+ Mb when in DAoC even with 300 player battle I wouldnt even hit 300Kb

Aside from increased client bandwidth, the server load is also increased due to having to process voxel data - no matter how much they opimize it, the performance can never be as good as if they removed the voxels from the game.

Bottom line, voxels have a huge detrimental impact on performance, especially when you scale up the number of players on screen.

Which makes me wonder why they went with voxels in the first place. 
I'd choose smooth gameplay over a gimic any day of the week.

nurso

Kunai_Vax said:
Which makes me wonder why they went with voxels in the first place.

Maybe because of the already existing tools&plugins for Unity, like Voxel Farm?

Count_Dirkoff

Kunai_Vax said:

which makes me wonder why they went with voxels in the first place. 
I'd choose smooth gameplay over a gimic any day of the week.

ACE are trying to e innovative and push the boundary in certain areas. Some ideas may work others may not. But you got to give them credit for trying.