Hi guys! I have an high end Mac, it usually run everything at max in term of games,, even some PPC games under emulation

It has 4 core at 2,66Ghz with 4MB of cache for cpu. I have 2GB ram and a Radeon x1900XT 512MB. Anyway Nexuiz isn't that fast, naturally with all settings on/max. If i activate all options i got around 20-30 fps in action.


Maybe Mac version isn't optimized?

Do you plan to support multithread OpenGL? It could be double or more performance.


PS: i donated, i apprecciate your work and the fact you support both Mac and Linux too.

Hi.. If you enable abseloutly everything, I'm not sure that any existing hardware would cope. iirc someone once said "its never ment to be used" or somehing. However to go to that extremeties i think you have to enable lots of stuff not even in the menus.

However its earlier been complained about the 3D suport with mac, dont rememeber weither it was the HW or drivers that was problematic, and Nexuiz dont really run on more than one CPU iirc, unless something has changed lately.

However nexuiz is a project always being worked on so things might very well change, and might very well have changd in these regards withiut me knowing, so wait for a post by someone more enlightened than me.

cool that you like the game btw, and welcome to the forums

Thanks so much for your reply.

Imho multithread support would be fantastic (some games really double/quad fps).

ikir wrote:Imho multithread support would be fantastic (some games really double/quad fps).

Multithreading wouldn't give you any performance boost with Nexuiz. On any modern multi-core system, Nexuiz doesn't even use 100% of a single core, the bottleneck is usually the GPU. So multithreading would probably just create additional overhead and slow things down on single-core systems.

I don't know i'm not a coder, but World Of Warcraft and Quake 4 double their perfomance on Mac OS X (multithread OpenGL).

ikir wrote:I don't know i'm not a coder, but World Of Warcraft and Quake 4 double their perfomance on Mac OS X (multithread OpenGL).

Well, I have no idea how that multithreaded thingie is supposed to boost performance - perhaps it just hides the drivers slowness away (the app can continue although the driver is still pondering on the last frame).

Actually after reading http://developer.apple.com/technotes/tn2006/tn2085.html that seems to be sorta the case.

Our main problem on the Mac platform isn't speed (well, that too) but the horrible driver quality. On a Mac Nexuiz will look worse than on a normal PC thanks to the drivers forcing us to disable eye candy (otherwise it'll just crash or draw stuff wrong). If things work they're usually slower than on Windows or Linux.

Another problem is that not one Nexuiz developer has a recent Mac (only one old Mac mini IIRC). Currently we just compile Nexuiz for Macs and hope it'll run. That's far from optimizing performance.

I understand. As far i know new games, written with OSX in mind are fast as on XP, so much more faster than same game in Vista. Since OSX has a very good multicore/cpu support, games written for more cores or using multithread OpenGL gains a lot. indeed WOW gained double fps with last Blizzard update on low end machines like Macmini e MacBook.

If you need a tester i can help a little. And i'll make my friends donate, so you can buy a new Mac
Could you ask to an Mac developers to join your team? maybe you can find someone posting on some Mac big portals or developmente website.

If you want i could ask for a developer on some italian website (telling that the coder must know english)

Anyway, thanks for your fantastric work.

PS: with this machine i can play WOW with all to max setting at 1680x1050, Quake 4 with high quality setting and all effect/option to max at 1600x1200 (with 1680x1050 runs very good but on some area i get a slow down to 20-25 fps) and HL2 mods (Source engine) like dark Messiah to 1680x1050 here again max setting. So i think Nexuiz/Darkplaces could be optimized a little.

Hi,
i am a mac user, i have played at your wonderfoul game.
But, i have see that i have slow performance too on my MaBookPro

So i will ask if mac palyers in the future could have an optmized version which runs as fast as Win version

I would donate if it become more playable

Searh an OS X developer,

thanks

Actually, the main problem is that the engine developers don't have Macs. I am not one of the main engine developers, but I do have a Mac - a Mac mini, with a Radeon 9200. That is too slow for Nexuiz anyway (on Linux too). Also, it is a PPC Mac.

Actually now there's code to use the multithreaded OpenGL thingie in the engine (thanks LordHavoc!). Of course this is untested.

New versions of Nexuiz will ship with that support, although it may be you have to enable it manually - we have no idea if it is safe to enable it by default. This needs testing.

Well as others have said: Other operating systems that that kind of stuff automatically. When apple does not it might because its not safe to enable it for all applications.
But maybe Nexuiz is lucky and does not get problems when enabling it

Thanks! I will test the new release when it comes out! Thanks you Do you have an ETA?

Ah Mac OS X 10.4.9 is out, 3D games are faster... and Nexuiz is not an exception. Indeed it gains few fps. Now with all turn ON i go from 60 to 15 fps it depens from the area... it isn't playable yet at max settings but still good.

Ah, exellent news, Im willing to beta test this multithread script with my 1.8ghz mac mini.

Ikir, I didnt see any such thing on the updates page, anyone bother link me?

well, probably divVerent could temporarily create a current release (which includes the mac binary too) and give it to the people who are already willing to test it, especially since LordHavoc included a cvar to use that multithread opengl feature now.

Indeed i'm waiting too

Speed increase was referring to Mac OS X 10.4.9

PS: how i can make a "timedemo" with Nexuiz?

just enter tdem demo1 or any other demo name in the console. but demo1 and demo2 are included in Nexuiz so everyone can use those to compare and demo1 has more action in it and is longer so its the better one for a benchmark..
(Open the console with shift-escape and close it again with escape)

You should run that at least twice as the first run might compile some stuff on the graphic card and is slower then the following runs.

For example my nv 6600gt gives me on default settings:
2.2.3: result 1909 frames 35.0140000 seconds 54.5210487 fps min/avg/max: 17.2413793/54.5210487/250.0000000
upcoming 2.3: result 1909 frames 29.6720000 seconds 64.3367484 fps, one-second min/avg/max: 37 64 95

The min and max in older versions are a bit meaning less as those are the absolute min/max fps but people want to know the one second average.. But still thats a improvement from 54 to 64 fps on average = 18%

At my usual play settings its:
2.2.3: result 1909 frames 20.3980000 seconds 93.5876066 fps min/avg/max: 41.6666667/93.5876066/500.0000000
upcomming 2.3: result 1909 frames 16.8920000 seconds 113.0120767 fps, one-second min/avg/max: 92 114 149
improvement from 93 fps to 114 on average = 22%

Hmm not sure what wrong with my sytem today, maybe its beryl, yesterday i got:
result 1909 frames 15.6910000 seconds 121.6620993 fps, one-second min/avg/max: 100 123 157
But thats just a bit faster anyway..

so is the cvar for multi-threaded OpenGL enabled by default in 2.3 test builds?

how can I display every cvar?

yes multithreaded opengl is enabled by default in 2.3.. i'm not sure which 2.3 test release had it enabled by default but we'll do a FINAL test release later today where its enabled for sure!

You can list all variables with cvarlist in the console..

"tdem demo1"

Didn't know that, been using: timedemo demos/demo1.dem

Is there a way to loop this timedemo? As a log is generated it could be useful for testing overclock stability

Thanks.

Hmm not from inside the game.. but the command line option -benchmark demos/demos1 will do a timedemo and close afterwards. If you start it from a shell like 20 times that should be similar enough. You can also use the option +forceqmenu 1 to make it NOT load the menu which is a huge speedup for such tests.
BTW yesterday we released the LAST test candidate for 2.3 which has multithreaded opengl active by default so this is a good thing to test NOW and report back

Thanks, I will look into it & also try out the new test release

Nex 2.3 is about 20% faster for me, I've added shadows & still get better performance than before

I've created a little script to run the timedemo for a dozen passes. That should be enough to prove reasonable stability of an oc'd vga. Here it is if anyone's interested:

Code: Select all

#!/bin/bash

i=0
while [ $i -lt 12 ]; do
    /opt/Nexuiz/nexuiz-linux-glx.sh -benchmark demos/demo1 +forceqmenu 1;
    i=$((i + 1))
done



Thats good and still sad though, i have much higher gains in 2.3.. i guess its either your graphic card or the bad apple drivers.. ( if you would enable opengl 2.0 shaders it should run a bit faster but the apple drivers screw up some of the eyecandy then..)