MoP engine changes and hardware upgrades

Bani

Member
Apr 13, 2007
1,238
Berlin
Hi all,

5.0 introduced some new engine features which decrease your FPS: SSAO and new lighting.
SSAO is optional, but the new directional light sources are not (or integrated somewhere else, but I haven't found anything about it).

The changes to lighting look like this in extreme cases:
http://i.imgur.com/fc0Y2.jpg
http://i.imgur.com/dtYc4.jpg
Here's a video where I think the new lighting looks awesome:

SSAO adds pools of shadow in areas that should be naturally darker due to indirect light. The change is very very subtle, but it helps a lot in making the world look "real" instead of just collections of polygons with sharp shadows. It also causes what some people misinterpret as a "dark glow" around a character, which is why you should disable it if it bothers you :p

No SSAO: http://i.imgur.com/D12F3.jpg
SSAO: http://i.imgur.com/y9nfO.jpg

Now, Pandaria will be even more taxing on your system since they increased the polygon count and world detail yet again as they do every expansion.

Aaanyway, I wanted to open a little thread on this and the effects on hardware since there is so much confusion around.

The impact of SSAO: SSAO performance depends only on your screen resolution, NOT complexity. It will always take the same amount of milliseconds per frame whether you're looking at the ground or fighting Ultraxion.

How large the impact is depends on your gfx card. On my desktop 5850, it was only around 3-5 fps that I lost.

My notebook entry-level GT 620M gets -40% FPS with SSAO high and then some more since it gets so hot from all the calculating (much like we Germans) that it reduces the clockspeed.

I also have a shit-old Core2Duo entry-level CPU at 3.33 Ghz from 2008. To test the impact of the CPU, I went to Twilight highlands (which is a cata zone, so high poly count) to an area where I see the Horde base, lots of landscape and some water for the reflections.

With absolutely everything on Ultra, I get the following results:
Pentium Duo E5200 at 3.33 Ghz + Radeon 5850: 30 FPS
C2 Quad at 3.0 Ghz + Radeon 5850: 35 FPS (with some other optimizations regarding ram and bus speed)
C2 Quad at 3.7 Ghz + Radeon 5850: 35 FPS.

So basically, going from a shitty dualcore to a good but also outdated core2quad got me around 15% FPS even though I'm mostly GPU-limited. I checked some benchmarks around the net, and in a DX11 game with settings that are brutal on the GPU (e.g., Crysis), even a 4.5 Ghz Sandy Bridge i7 won't give you even one more fps compared to a core 2 quad. So don't upgrade any Core 2 Quad for MoP :p

Also, my C2Quad shows no FPS change between 3.0 and 3.7 Ghz.

Will report here what difference an nvidia 660 Ti that I just ordered will make to the 5850 :)

And thanks for reading my rambling post ><
 

Croga

Says funny things =)
Sep 9, 2008
892
The Hellmouth
Good info Bani. I'm very curious what my laptop is gonna do with Pandaria.... upgrading GPU isn't really an option anymore :(
SSAO definitely looks impressive. Hadn't seen such a good example of it yet.

About the CPU:
WoW is going to demand a lot more from your CPU as soon as you go into crowded areas or areas where there's a lot of spells being cast (like in a raid).

Your example is an area with high GPU intensity but low player interaction and therefore the GPU will be your bottleneck.
As soon as MoP launches you'll see a lot of players and a lot of activity from those players in the MoP starting areas. In that case you'll see the CPU becoming the bottleneck.
Run the same test you just ran during a raid and your results will be much, much different.
During our BoT/TotFW/BWD days I upgraded from a Core2Quad 9550 to a Corei5 and the results were no less than spectacular.
 
OP
Bani

Bani

Member
Apr 13, 2007
1,238
Berlin
hey, my "new" one is also a Q9550.
I'd love to test it in raids but haven't really found a way to do so. Especially since now the old CPU won't be put back in, I can only play with the clockspeed.

Maybe I'll try LFR ultraxion with low GPU-only settings, but then you'd only notice with differences >10% due to the variation....