Starcraft II GPU performance/IQ

I never said impossible, but some quite significant headache. And definitely not something I`d want to force down on gameplay-programmers (read scripters) which would need to be well aware of the problems and limitations.

Oh in a ideal world x86 would have died altogether long ago, Im pretty sure even in 10 years we will still have a fair share of mostly single-threaded game-logic.. and worse yet still x86 dominance.
I mean, its Blizzard dammit, you cant even zoom out primary because of "competitive gameplay" and possibly the existing Starcraft players losing a edge they have by having mastered zapping around the map. They set their lowest common denominator and if it runs well there it will run well on faster machines too.

 

Texturing performance seems to be one of the performance limits (blizzards SSAO implementation requires a lot of texture sampling)...

 

Why would scripters need to care about this? I already said that scripting engines are hard to parallelize.
At least to me flow field is relatively easy to parallelize and should be.

 

Cat 10.7 hotfix AA performance.

http://www.pcper.com/article.php?aid=958&type=expert&pid=4

 

Unfortunately those drivers are even less stable than the 10.7's on my system(4850), the display driver stops responding often(and hard crashes eventually) even when I'm just on the desktop, the 10.7's would only do so when watching a video. I wonder what's in the codebase that causes the 10.7 and the 10.7 AA hotfix to suck so badly on my system?
Btw, I was able to use 2x AA with quality AAA with the game maxed at 1680x1050, 4x fsaa proved to be too much for it, but luckily aliasing isn't that bad in SC2.

 

the AA part just makes me shrug, but does SC2 run with Eyefinity6 for that "gameplay advantage" ?

 

I like it that AMD has gone beyond itself and provided customers with the option to use AA at their discretion. This way, everyone can decide for himself what he find's the most acceptable compromise between resolution, details and antialiased edges.

Of course AA costs performance, always has, always will.

edit:
Our piece on antialiased SC2-performance is up also:
http://www.pcgameshardware.de/aid,7...-fuer-Geforce-und-Radeon/Strategiespiel/Test/
Interestingly, the radeons seem to take a much harder hit from the transparent (shadows of) protoss units cloaked by their mothership.

 

Here's Eyefinity 6:
http://www.youtube.com/watch?v=HwEgxtQLTsA

Its confirmed running with 3X1 also, but only in portrait mode. People haven't gotten it running in 3x1 landscape yet.

 

I tried running it using nVidia surround the first day I got it, but couldn't get it working. I'll have to investigate further on how. I was more worried about actually playing the game, so I just played at 1920x1080.

Currently, I'm running the game at 1920x1080 with 16xAF, 32x AA, 4xSupersample AA. I'm running 2 GTX 480s.

 

What's this, a GTX260 beating the 5970?

 

Is there a profile in the latest beta Crossfire profiler thing?

 

There's no 5970 in the graph I see...

 

It's in the text.

Once again, Crossfire/SLI demonstrating it's not always as simple as 1+1 = 2.

 

Probably not. First installed 10.7 WHQL pluss CFX application profiles, then 10.7a AA-driver. While we seem to have identified the main performance killer with AA actived for the Radeons in our benchmark sequence (cloaked units -> transparent shadows, turn off in variables.cfg), CrossfireX on the 5970 was a no go. That is why we didn't include it in our graphs.

 

Bah! Those tick bezels are game killers. I wouldn't stand it.

 

And bezel compensation doesn't seem to be enabled… It's really ugly.

 

unfortunately ATI's Bezel Compensation leaves a LOT to be desired,.. Hell Matrox has had working b.c. with differing size displays for some time and iirc Nv's surround vision is far far more flexible when it comes to different displays. ATI seems to use a monitor/display's pre-programmed edid info rather than registry/driver info and this leads to serious issues. The only way around this is for people to risk killing their monitors flashing edid info (for DVI) instead of just allowing user customized b.c. I think ATI/AMD has a very strict (<5% ) difference in area to be able to use b.c. Out of the four main multi-display options (SoftTH, MatroxTH, nV Surround and ATI Eyefinity), ATI's is the most prohibitive and as a result the biggest pita to set up, even with near identical displays, changing from DVI to DP or VGA can disable any possible bezel compensation. As I mentioned part of the problem lies with manufacturers however when other competing technologies can correct use compensation and eyefinity can not, it leaves users with very few options (suck it up and wait, bitch or switch).

 

Eh? There is a very nice tool integrated in CCC for bezel correction configuration

HardOCP has a video demonstrating it:
http://www.hardocp.com/article/2010/03/24/ati_radeon_5870_2gb_3x2_eyefinity_gaming_experience

 

Thanks for the transparent shadows tip/tweak!

 

yeah thats if/when bezel correction/compensation works correctly and more often then not such "reviews" are using pre-supplied hardware that usually includes 3-6 identical displays,.. the problem with ATI's bezel correction/compensation is that it is only enabled when the displays are nearly identical according to eid settings within the display's firmware and that the info can vary on the same monitor depending on whether one uses, DP, or DVI or VGA. Other solutions such as Matrox SoftTH etc aren't nearly as limited (again iirc, the display's edid info must be within 5% of one another or bezel correction won't be enabled). I haven't played with Nv's solution personally however before R3D blew up more than one person was able to use nv's b.c. with varying sizes of displays (21 23 21) where as ATI is incapable of doing so.