OpenMW Performance

General discussion regarding the OpenMW project.
For technical support, please use the Support subforum.
User avatar
psi29a
Posts: 5361
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

OpenMW Performance

Post by psi29a »

I've been doing tests across multiple platforms (all Intel). At this point, it was all heavily dependent on the video card and not the CPU. Feel free to do your own benchmarking, below are also my settings and how I got the results.

NoAA, FBO, 640x480, measure starting location interior of bashara, then move up to door leading to exterior and measure

Nvidia G98 [GeForce 8400 GS] / i7 quad core
| branch | interior | exterior |
| 0.14 | 42 | 22 |
| 0.15 | 42 | 22 |
| 0.16 | 60 | 33 |
| 0.17-pre | 60 | 45 |

ATI RV635 [Mobility Radeon HD 3670] / Core2 Duo
| branch | interior | exterior |
| 0.14 | 97 | 38 |
| 0.15 | 94 | 37 |
| 0.16 | 102 | 39 |
| 0.17-pre | 105 | 39 |

Intel GM965/GL960 / Core2 Duo
| branch | interior | exterior |
| 0.14 | 17 | 8 |
| 0.15 | 17 | 8 |
| 0.16 | 17 | 10 |
| 0.17-pre | 16 | 9 |

Note: 0.17-pre is based on this

Code: Select all

commit 483b125aad586b3165537be7a6c55470503fc47c
Author: Marc Zinnschlag <[email protected]>
Date:   Tue Jul 17 18:37:20 2012 +0200
Tarius
Posts: 574
Joined: 24 Oct 2011, 19:29

Re: OpenMW Performance

Post by Tarius »

I also get a performance boost.

http://forum.openmw.org/viewtopic.php?f ... 1&start=10
I last tested 0.14.0. Link shows my system stats and previous test.(windows, 1440x900 windowed)

For .16
Interior, it fluxuated depending on where I was looking, but 70-80+ wasnt uncommon.
Average was probably about 70. (compared to .14, this is about 10-20fps more)
Looking up at nothing in the interior I maxed at almost 250, compared to .14, thats almost 35fps more.

Exterior had an increase of about 5-10fps(for .14, it was 30-40, for .16, its now 35-45) This was only while raining though. When sunny, there was no increase and in fact there was a bit of a decrease. This may have mostly to do with draw distance though.

Around the floating moon in vivec, I got low 20s for fps which is in fact lower than I have seen before.

The major difference between this and my previous test is that I had no AA this time(I had it set to 2 last time)

I think I may have found something that kills performance. It was actually raining when I went looking one time. Everytime a lightning bolt flash happened, my FPS dropped by like 50-90% at best, sometimes it just froze for a moment.

Interesting observation here. I know people have complained how armor attached to the chest was black or really dark. I have the exact opposite thing happening. NPCs are entirely black except for the armor attached to the chest. Activators are also completely black for me. Surely I am not the only one who has this problem? No, I dont care to try and fix it unless this some how fits a key to a puzzle or something like that.

Another interesting thing, I tried 1024x768(as opposed to 1440x900) and got 40+% fps boost on interior cells from .14. However, comparing exterior results, there wasnt any boost at all, it didnt seem to matter much.
HiPhish
Posts: 323
Joined: 02 Jul 2012, 08:36

Re: OpenMW Performance

Post by HiPhish »

I get ≤30 FPS outdoors and indoors on my early 2009 iMac (2.66GHz Core2Duo, Nvidia 9400M, 4GB RAM), no reflections, lowest view distance and any graphics filtering. This would be playable enough, but unfortunately it slows the cursor down as well, making menus rather tedious. The biggest performance hugs are reflections (just turning reflections on without the other ubsettings eats ~10FPS) .When I first tried OpenMW 0.16 it was raining, which dragged performance down even more.
Myckel
Posts: 72
Joined: 16 Jun 2012, 09:48

Re: OpenMW Performance

Post by Myckel »

Inside ~15FPS, outside ~7FPS (viewing distance on minimum), about ~15FPS when I turn off the texture filtering. Everything else is off as well.

Radeon HD 3450 (Xorg open source ati drivers), i3 4 core cpu, 4GB memory.
User avatar
raevol
Posts: 3093
Joined: 07 Aug 2011, 01:12
Location: Caldera

Re: OpenMW Performance

Post by raevol »

Myckel wrote:(Xorg open source ati drivers)
There's your problem.
Chris
Posts: 1626
Joined: 04 Sep 2011, 08:33

Re: OpenMW Performance

Post by Chris »

raevol wrote:
Myckel wrote:(Xorg open source ati drivers)
There's your problem.
I use the same thing with worse hardware, and get better performance (10~15 FPS outside and 20~30 FPS inside in the general case; sometimes worse, sometimes better, depending on the actual scene). The biggest hit seems to come from the shaders.
Myckel
Posts: 72
Joined: 16 Jun 2012, 09:48

Re: OpenMW Performance

Post by Myckel »

raevol wrote:
Myckel wrote:(Xorg open source ati drivers)
There's your problem.
Nope, these days the FOSS drivers are quite good (I've used games which require more graphic power than OpenMW, and they run fine). They should hold up for something like this.

There was an update last week for these drivers for my distro, but people reported segfaults when switching to them (including me). Last evening I had an other look at it. It seems the new drivers required KMS support in the kernel (which I hadn't activated yet due to issues with it in the past). Spend a few hours, rebuilding the kernel, xorg and the xorg drivers, but finally everything worked.

So I launched openmw and got a huge performance boost. Now 35 - 40 inside, and 20 - 25 outside with viewing distance on 4/5th and bilinear texture filtering. Activating the water shaders is still doable, but still has quite some impact.
User avatar
psi29a
Posts: 5361
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

Re: OpenMW Performance

Post by psi29a »

Myckel wrote:Nope, these days the FOSS drivers are quite good (I've used games which require more graphic power than OpenMW, and they run fine). They should hold up for something like this.
You've been reading too much phoronix. ;) My results are based on amd's and nvidia's binary blob. The intel result is the only one that is FOSS. If I used foss for amd and nvidia, I get about a 80% drop in performance. You be the judge.
Last edited by psi29a on 20 Jul 2012, 11:54, edited 1 time in total.
veepee
Posts: 12
Joined: 10 Aug 2011, 20:17

Re: OpenMW Performance

Post by veepee »

Speaking of Phoronix, afaik the Phoronix test suite is free to use in a project like this? I've never used it and i think it might be slightly beyond my skills, but possibly someone else could set up an automated environment for future performance testing?

Possibly Michael Larabel might even be interested enough in this project that he'd do some testing himself?
Myckel
Posts: 72
Joined: 16 Jun 2012, 09:48

Re: OpenMW Performance

Post by Myckel »

BrotherBrick wrote: You've been reading too much phoronix. ;) My results are based on amd's and nvidia's binary blob. The intel result is the only one that is FOSS. If I used foss for amd and nvidia, I get about a 80% drop in performance. You be the judge.
I don't read/visit phoronix. I also never managed to get the ATI proprietary drivers working, so my only reference is the FOSS ones. 2 years ago performance was bad, these days it is doable, but I believe you that they aren't performing as good as the proprietary ones do.
Post Reply