The Official Graphics Card and PC gaming Thread

Sorry for language.....And thanks for the advice. By the way how's 8600GT or GTS???

 

From a recent game title:

Performance scores
9400GT: 20
9500GT: 71
8600GT: 58
8600GTS: 71
HD5670: 140

That should be all you need to know right?

 

Yeah, thanks a lot man. I'm gonna get an HD5670, thanks again!

 

We'll so far the 5850 is very unimpressive to say the least. Minimum fps during Bad Company 2 has only increased a little bit to 28 fps, and manages to deliver a less enjoyable experience than my 8800GTS at the same settings. Smoke is blinding, the minigun on the helicopter causes the entire screen to flash white with every shot and a number of other graphical flaws have been rearing their ugly heads. Going to try and borrow a game from a friend so I can try it on another recent game before making the decision to send it back and get a better card.

 

try older drivers?

 

That made me LOL. It brings a tear to my eye when someone gets it that easily. Nvidia is fail. Their stuff is overpriced and not nearly fast enough to make up for the price difference. I mean, yes their top-end is faster, but not 30% faster like the 30% higher price.

 

I'll do that after trying another game

 

There's nothing wrong with the card. The HD5000 series has render issues with the game. The 10.4 drivers are supposed to fix that when they come out. Remember both the HD5000 series and Bad Company 2 are brand new. There are bound to be issues until drivers and patches mature.

Namely one issue for me is that in Russian vehicles, the bloom bugs out and it's very hard to see. And I'm still on my 4870s. The second 5850 wont be here for a few weeks, though I have confirmed it a working unit. Anyhoo, turning off bloom fixes the issue, but leaves the game looking very flat. Both the latest patch and the latest ATi drivers are supposed to have fixes addressing that and a few other odd graphical bugs on ATi.

Also of note, my friend with a single HD4890 1GB has zero graphical or performance issues with the game. He runs it Dx10 maxed with HBAO off at 1680 x 1050 w/ 2xAA and it runs 45-60FPS average with dips to the mid 30s. Not amazingly smooth, but I was able to play over an hour on it without any noticeable slowdown. IMO 30FPS is my lowest threshhold. Considering the graphical fidelity of the game, 30FPS is very reasonable for a minimum. My friend was happy with it to say the least.


My point being...

You should get more than that. There's nothing wrong with the card. Sounds like drivers to me as Bad Company 2 has more reasonable performance than that.

 

It was a Russian chopper that had the problem with the minigun lol.

I'm using the latest released drivers, but like I said I want to try it on another game before making a decision about the card. If it's just one game with a problem I won't care and I'll start looking for a second 5850 lol.

 

Just got done doing some very interesting benchmarks with Crysis. It seems the situation has flipped lately. DX10 runs faster for me than DX9.

Run at 1680 x 1050 with a single GPU to further differentiate results and remove the variable of Crossfire performance scaling.

Single HD4870, stock game

Dx9 High - 44.7FPS
Dx10 High - 48.3FPS

Single HD4870, my tweaked settings to emulate Very High with better performance.

Dx9 "Cheap" V.High - 41.1FPS
Dx10 "Cheap" V.High - 42.9FPS

Single HD4870, stock game

Dx10 Very High - 28.6

And for a reference.

Single HD5850, Stock game

Dx10 Very High - 37.8FPS

I just thought my results were interesting. Even if you don't plan to run very high, Dx10 seems to be significantly faster. Also of note, even while putting the RAM to better use, 64-bit runs slower overall and causes bad FPS dips, even on a comparable setup but with 8GB(2 x 4GB) of RAM.

I remember the Dx9 vs Dx10 debate being the opposite about a year ago. It seems the newer drivers and optimizations plus Windows 7 has turned the whole situation 180 degrees. Though all this doesn't surprise me too much. Far Cry 2 is MUCH faster in Dx10 as is Resident Evil 5(easily twice as fast). The rendering path isn't the issue. Dx10 in itself seems to do exactly what it says. Renders the same effects but faster. It's the effects they add with it that drop your performance.

So from my newest conclusions, Crysis should be run in Dx10 32-bit. Make of that what you will. I imagine there are people out there who will still find the opposite is true. Nvidia and ATi in particular make it very hard to compare platforms. The cards calculate Dx9 and Dx10 very differently.

 

I don't have any real performance issues with BBC2 other than that the game is really demanding, but there are numerous graphical glitches I get, not all of which are exclusive to ATIs, I've seen some happen on geforces as well.
I agree with you on DX10, I've seen several titles that run better in DX10 than DX9, FarCry2 is the most famous example.

 

Funny enough, even fully maxed Dx10 on my rig performs better than Crysis. What 3 years now and nothing has really topped it? Talk about ahead of its time

All-in-all I find the game runs smoothly enough and I don't feel like I've sacrificed to get that performance either. It looks fantastic at my chosen settings and plays more than well enough to be considered "smooth". As previously stated, 60FPS minimum is unrealistic. 60FPS average is realistic with decent hardware. As long as the game never drops below 30FPS I would call it smooth. Of course the Frostbite engine is a bit more forgiving than say the CoD4 engine in that FPS spikes don't instantly equal terrible lag.

If you could list them please? The only glitch I've currently found I'm not even sure is a glitch. In Russian vehicles, the HDR and bloom are way over-exposed. Makes it very hard to see a target unless someone marks it for you. Just for reference, American vehicles don't have this issue, just a green tinted filter. So I'm not sure if this is a glitch or just a small mess-up on the part of the devs...

Resident Evil 5 is the one that shocked me. Dx10 literally runs more than twice as fast on my ATi cards. On my 8800GTS both modes run about equal with Dx9 having a slight edge. I have been able to confirm the ATi example on other rigs as well. Have you noticed this? It might not be Dx10 running faster in this case but Dx9 running slower due to the TWIMTBP branding.... really all an enigma to me.

Notice I tend not to come to concrete decisions on my stance about performance. Games have very dynamic performance characteristics, and ATi and Nvidia cards seem to behave entirely differently given the on-screen situation. The differences are more than just the manufacturer of the chips. The actual hardware is structured differently.

Case in point. On Nvidia cards, I always had to watch my level of AA. Many games can only have so much AA before the performance drops off. On ATi cards, no matter the amount of AA, there is a smooth gradation in performance between the levels. This illustrates a lot about the differences and helps to further my curiosity.

 

Well, I agree, BBC2 isn't quite as demanding as Crysis, but it's not far off, and most importantly it requires higher frame rates to run smoothly, so it effectively is just as demanding. As for lagging at a lower frame rate compared to COD, it's not really visible behind microstutter, since BBC2 will lag below about 75fps due to this.
Also, plenty of stuff has topped Crysis. Crysis Warhead, Metro 2033, Cryostasis, STALKER Call of Pripyat, Shattered Horizon, to name a few.

As for the glitches, it's texture flickering. All water surfaces flicker badly, as do a large number of wall textures. It can be stopped by changing AA modes to refresh the graphical output, once per map change.
I didn't play RE5 on the PC, so I wouldn't notice that there.

As for AA, this is mostly due to memory performance. nvidia cards manage memory badly, but typically have more of it. When on equals or superior, i.e. 1GB HD4870 vs GTX260 216 896MB, the Radeons come out on top. When the geforces have more memory, it's not as clear cut - the GTX480 with 1.5GB of RAM destroys the HD5870 with 1GB once you use more than 1GB worth (e.g. Crysis Warhead at 2560x1600 with 4xAA)

 

As far as saying "but it's not far off", is a bit misleading. BC2 easily gets twice the performance of Crysis in Dx10 if both are maxed w/o AA. There is no contest. Not noticed any microstutter at all playing 12+ hours at ~60FPS. Sure, there's little load stutters here and there, but nowhere near the terrible jittering like FEAR 1 with Crossfire. Runs MUCH better than Crysis. Also don't forget, every system is unique and will have its own issues. A game that might have microstutter for you might be buttery smooth for me. And yes I am VERY sensitive to it.

Really you are very performance critical. One small issue and it "runs like crap". I understand being discerning, but that's just being anal retentive. It's PC gaming, you're not supposed to run everything absolutely perfectly. Hence the forward march of technology.

Crysis Warhead - This is a given, IMO this is just Crysis v1.1 figuratively speaking. It's practically an expansion.

Metro 2033 - Only interiors and then only certain parts. Half the game is one giant dark sewer tunnel. All this while having worse minimums than Crysis ever had. IMO does not top it at all. Maybe close but not beating it. Also, Crysis' models and animation stomp the hell out of it hardcore.

Cryostasis - LOL no, not even remotely close. It looks good, but not even within the same ballpark as the other games you named. I was surprised to see it on your list.

STALKER Call of Pripyat - LOL Not if it's still on the original engine. Both Stalker and Clear Sky looked like vomit to me even maxed with the enhanced lighting model. I mean WTF did they get their gun and monster models from Quake 3? Call of Pripyat doesn't look much different from the first two. Maybe a slight texture improvement and even more demanding lighting. Also, the lighting isn't too pretty when you can see the damn lightbeams look like they're solid. Not to mention the ungodly amount of bugs in all the Stalker games. Way more than Crysis ever had. Very shoddily made game series.

Shattered Horizon - I'll give you that. But it's not really a game is it? More like an extended tech demo you have to pay for. BTW it runs worse too.

IMO Crysis has not been topped for sheer eyecandy yet. Sure, maybe in certain categories. But as a whole graphical experience, Crysis still looks more realistic, natural, and just plain shiny. Don't forget many small parts make a whole. Crysis just happens to have way more small parts than the rest and THAT'S what they need to beat IMO. It's the overall quality with which everything was brought together.

Well aware of this fact. Just using the example to illustrate an abstract concept.


Just an IMO list of a few games that stand with Crysis:

CoD4 Modern Warfare/MW2 - Matches it in models, better textures, but still crap water and no real "lighting" model to speak of. Everything is imitated with specular map.

Far Cry 2 - Matches in textures more or less, matches in lighting, but the models and physics are very "plastic" feeling. Does do foliage better though.

Resident Evil 5 - MINDBLOWINGLY FANTASTIC TEXTURES AND IMAGE QUALITY. But very bland, uninspired environments, very little color, contrast, or visual clutter(much like your opinion on Gears of War).


Bad Company 2 - Actually about even. Could have put more work into the overexposed lighting though and less into the useless "darker" shadows.

Not seen much else approach it. Maybe Shattered Horizon and Source 2007-Current
.

 

Haha don't mind me if I'm a bit abrasive today. Lack of sleep

 

i always thought that 60 minimum was insane to get

as for microsutter, never ever had it before, so i guess if i had 75FPS avg i would be VERY happy! mmmm time to pull out the crysis demo/benchamrk and see what i get

 

Finally settled down and played some GTA IV for a change. Was noticing some choppy edges. In certain areas. Turned up the Texture filter quality, and was astounded by the resulting effect. The following example says it all:

I don't believe there is a way to change "AA" in GTA IV. However, texture settings seem to effect it in that way Sure did effect my CPU/GPU. I saw one point where it dropped to 25Fps. I've never seen it go that low before. Usually I average 40+ frames easy. Even at stock on the 965. Which it's currently running at LOL! Overclocking unfortunately is not my current priority. Rewiring is. I'd like to rewire with a 750W corsair though Which would also set me up for dual crossfire wink wink LOL!

 

dual CF, as in 4 cards? wow lol.

 

Jeff: It's not far off, because playing Crysis at 2560x1600 with all max settings on four 4870s isn't doable, and neither is it with BBC2. You can turn one or two settings down and Crysis becomes playable at that res, as does BBC2 if you turn HBAO off. They're not that far apart.
I dispute the 'one small issue and it runs like crap' statement. I don't recall saying BBC2 runs like crap, I said it doesn't run perfectly smoothly, which is the truth. It runs mostly smoothly, but not all the time.
Also, 'you're not supposed to be able to run everything perfectly' sounds as much as jealousy as it does anything else. Who decided that? There are several games out there that would run pretty badly even with the most powerful graphics setup you can buy, no matter how much money you throw at it, and that doesn't need to be the case at all. You shouldn't have to play a game and be like 'well, in 3-4 years time we can play it with extra detail!' By that time, games that are coded properly will likely easily surpass that graphical quality without maxing out 2014's hardware.
As for Metro 2033:

You tell me that in Crysis you only get 11fps minimum and 20 average on your 5850 with the lowest AA setting in the game, being honest. This is 1920x1080 too, so you you should take 7-8% off that figure as well.
Cryostasis is bar none, the most demanding game I have ever seen. You need a GTX285 to play it at 1680x1050 on all minimum settings.
Look at this crap:
http://www.hardocp.com/article/2009/06/01/cryostasis_gameplay_performance_iq/4
yes Crysis at that res brings the 295 to its knees, but not at minimum detail it doesn't.
STALKER Call of Pripyat: You seem to forget how demanding Clear Sky was.

Those frame rates are decent, until you see what resolution they're run at. I'm willing to bet a 5870 will beat that in Crysis. Perhaps I'm wrong, Crysis is very rarely tested at very high any more.
Shattered Horizon: Granted, the game is rubbish, but it's still a game that's more demanding than Crysis.
For sheer eyecandy, Crysis still does well. Not too well, because of the appalling quality of the rock textures, but reasonable. I'd certainly give Warhead the accolade of not being topped, but the original Crysis probably not. No doubt about it though, BBC2 is hot competition there.
COD4/6 does not match Crysis, not even close. The textures are nice in some areas, but very primitive in others, and the lack of any proper occlusion and even basic lighting makes it feel like an older game with shiny textures on top. That's no bad thing, in fact that's how I'd like most games to be, as it maximises the visual detail / performance tradeoff, but it does still mean it can't compete with Crysis at all for video quality.
FarCry2 has some very nice textures as well, and unlike COD actually has some decent mapping going on. It's just a shame I found so little to enjoy about the game, the fact that it doesn't support QuadCF was also a bit of a downer.
Resident Evil 5 PC I haven't actually seen yet. I know it runs extremely well on HD5 series setups, and the xbox game is as good looking as xbox games get, so that's hopeful. I don't think you can criticise the game's environments too much, FEAR was worse
BBC2 As I said before, I do agree with. The overexposed lighting can be a bit annoying, but it doesn't seem too unrealistic as to make me annoyed. It's just the graphics issues (presumably drivers, I'm not on 10.3 yet)
The Source Engine isn't really that close to Crysis for the same reason as COD4, everything's a bit flat. Performance / Visual quality though, Source is very impressive.


Shaff: There's a reason why 60fps minimum only sits in the 'Extreme' category for games performance I created. Wanting a minimal 60fps is nuts from the most demanding games out there, but there's no denying from anyone, a game that plays at that sort of frame rate is so much nicer and smoother to play, without any of that minor, but very noticeable 45-50fps area that doesn't lag, but yet doesn't feel quite right.
As for average frame rates, I stopped using them a long time ago when I realised that games with an average frame rate of 90 and a minimum of 17 are stupid to measure by average fps.

Omega: Resolution has a similar effect to AA in that, the higher the resolution, the smaller the 'steps' on the 'staircase' become, because there's more of them, more detail, more pixels in the textures. The clue is though, with high res textures, the 'stairs' will still be there if you look. With AA they never will be.
Textures, needless to say, stress graphics, not the CPU, GTA4 with high detail is very demanding.

Maw: I don't think that's what he meant, dual crossfire (2 cards) as opposed to quad crossfire (4 cards).




Addendum: re: Metro 2033, check this out:
http://www.hardocp.com/image.html?image=MTI3MDU3NTA0NVJwenVVTjROekdfMV8xX2wucG5n

nvidia up to their old tricks again?