Starcraft II GPU performance/IQ

I'm pretty sure the whole point of bezel correction is exactly that you can have monitors with varying bezel thicknesses and correctly set those up.
Hence there's an automatic mode, which uses edid (and if I remember correctly, also doesn't work in the [H] video?) and a manual mode where you can adjust the bezel correction yourself.

That's in there since Cat 10.2 or 10.3

The 5% you are talking about is the pixel density or PPI which means you line up a 22" 1680x1050 with a 27" 2048x1152 because they have a very similar pixel density (90:93) whereas you can't line up a 24"1920x1200(94) with a 27"1920x1200(84) display.

But that's where you're already in a situation where your gameplay would be less than optimal.

 

Contemporary Graphics Cards in StarCraft II: Wings of Liberty

 

Interesting, contrary to what other sites seem to say, the GTX 4xx cards take a much higher hit when FSAA is enabled.

Although interestingly, while without FSAA the GTX 4xx cards drop fps much faster as resolution goes up, when you enable FSAA the fps drop as resolution increases remains similar to the 58xx cards.

Interesting that the 58xx does relatively better in GPU limited cases where the 4xx cards do relatively better in CPU limited cases.

Regards,
SB

 

All the tests are pretty different in terms of system load, as the captured gameplay or whatever is pretty different for each website, too.

Nice to see the differences in the terrain with the various detail levels, though, and of course the effect of the small light sources.

 

In case anyone's interested the AthlonX2 and X800 combo is running the game quite fluidly at 1650*1080.
The drawback is that except for texture and terrain detail at medium, everything else is at low settings It's quite ugly really, no shadows, shaders, effects at all. Oh and the loading times are quite slow as well, even though I've been trained on the X360 where 30-60 seconds are the norm.

I'll need some time to backup my data and settings, and then I can assemble my new computer. i5-760, GTX460, a fast hard drive - it should do far better.

 

Anything that moves the bottleneck away from geometry processing (higher resolution / AA) should theoretically cause a bigger % drop on Fermi cards because they're starting from a higher performance level. What you should look at are cases where Fermi and Cypress are close without AA and see what happens when AA is applied - that's where the belief of faster AA on Fermi came from.

 

Could you see yourself enjoying the single player experience like that? Or engaging in any competitive multiplayer? Honest question, not trying to be a smartass.

Yeah I think you'll notice a speedup in most games.

 

The game itself should work well, after all it looks better than SC1 and I think even the larger unit counts should still be manageable for at least 1 vs 1 games. I've only played the first mission but it had a very solid framerate. Any advanced lighting/shader model kills it though, the extra data traffic from the FP16 deferred rendering is definitely too much for this poor old X800.
So the game looks kinda like a more detailed Warcraft 3 which I've replayed about 1-2 years ago and had a lot of fun with it. If I had no option to buy a new machine, I would probably play through the SC2 campaign and look into multiplayer, yes.

Although, the in-engine cinematics do look very bad (no shaders and shadows at all, lower texture res, and so on) and it does detract from the enjoyment of those parts, as I've seen how much better it looks on a newer system.

I don't have any new PC games although I might get the HL2 episodes, should be dirt cheap on Steam by now. Maybe I'll look into Crysis too... but honestly I don't really feel compelled to. Have a bunch of X360 games to play as well and I find the couch and the controller much more inviting

 

Get the Crysis Maximum Edition. It's $40 on Steam and well worth it. You'll be rocking it on all very high settings and you'll be amazed. It's also one of the best FPSs in recent history.

 

SC2 AA Performance @Hardwarezone

http://www.hardwarezone.com/features/view/64848

pretty much confirming xbitlabs numbers

 

I tried my X800 XT (@ 500/600) with Supreme Commander when it came out. It was really too much for the card (I was surprised). Actually, that card is still powering my boss' kids' game machine and they use it every day.

I have a FX 5950 Ultra around. Maybe I should try that? I saw in Blizzard's presentation that the FX series was an intended target. (lol yeah right) Unfortunately I don't have SC2 yet.

 

I'd be surprised as while the Radeon 9800 (same generation as the FX 5950) is listed as a min, for Nvidia it doesn't start until the next gen 6600? Not sure how the older 5950 compares to the 6600. Would be interesting to see what kind of performance you could get out of it either way.

Regards,
SB

 

I seriously doubt it would be even remotely playable. 5950U can barely handle Half Life 2 in DX9 mode. It's about half as fast as when it's in the default DX8.1 mode that Valve specifies for it. It may be even slower than that actually. Generally if you try to run any PS2.0 shaders through a FX card, things go very badly. I think a Radeon 9800 Pro is well over twice as fast in this case. GF6600 is about the same speed as a 9800 Pro, I believe.

Yes I admit to have been playing with a 5950U rather recently. NV's worst oopsie series. They fascinate me in retrospect.

I have tried Oblivion on both 9800 and 5950 too. 9800 allows the game to be playable with generally high quality settings up to around 1280x800. The 5950 really can't run the game at a playable speed unless you crank every detail to the minimum. I should try that again. That was a few years back so it's a bit foggy at this point.

FarCry runs pretty good but according to the console the game uses a mix of optimized PS2.0, 2.a and 1.4 shaders.

 

Are these older cards really that weak or is this simply from using a "too high" resolution for that generation? It wasn't that long ago that you were simply crazy to go above 1024x768 in HL.

 

It's both. I've been able to play both Doom 3 and Half-Life 2 at 1680*1050 on this old X800 with full detail settings and maybe even 2xAA in HL2 when they were released (and with a single core AthlonXP instead of this X2 whatever).

The thing is that SC2 uses far, far higher resolution textures, so the 256 megs of VRAM are just not enough for a start, it'd probably take at least 512 for the full texture detail.

And the other problem is that advanced shaders require deferred rendering, which uses multiple render targets and all are FP16, instead of the single 32bit integer render target for the low detail settings. The fill rate, memory and bandwidth requirements are increased about 8 times and no framebuffer compression can help that.

So the old cards aren't quick enough at the pixel shading level, they don't have enough memory, and what little they got is just not fast enough either.

 

Works with Eyefinity now with a custom resolution launcher and should work with surround as well in landscape:


http://www.widescreengamingforum.com/forum/viewtopic.php?f=57&t=19928&start=15

 

"Works" looks a too strong word...

 

If you can't run it, perhaps this version is an option:

http://www.gametrailers.com/user-movie/lego-starcraft/348644

 

I wouldn't consider it a strong word. Rather a boolean expression, since in its context its either true or false.

If you find that wrong, it would be a bit more fruitful if you provide some arguments behind your objection of my use of the term "works". As it stands, it seems more like an emotional reaction then an argument.

The hack enables the game to run in common Eyefinity (and surround) resolutions with an increased field of view as result. By my definition, it works in Eyefinity when hack is applied.

 

But the game doesn't render at that ratio, right? I've seen some YT videos of the cutscenes and what basically happens you have a 16:5 view of a 4:3 movie,