NV vs ATI again CS:Source Bench

[Gemini][]

These benchmarks are obviously messed up, no doubt about that. The really sad part was in the thread follow up, where DriverHeaven forum members were trying to defend these benchmarks with blind fanboyism.

ATI's 6xAA looking better then NVidia's 8xSA? Rrrrrright....

Believe it or not, it's true.

 

[^Ben]

Ok guys those results are bullcrap.

I just tried on a 6800GT stock.

1600x1200 4xAA, 8xAF and just got 85FPS.

AMD 64 3000
6800GT stock
1 gig geil value ram

 

[blackeye]

It could be true but from what I read the people that were saying the benchmarks were biased were the people with 6800 ultras and gt's. There wasnt one person that had an x800 that posted anything to contradict (dont know spelling) the benchmarks. So its still kinda hard to tell who right.

Im concerned though now as I was concidering getting a 6800 ultra and overclocking it to ultra extreme levels. But now that I see these benchmarks I might just have to wait for my x800xt preorder to come in from quebec still. But there is something wrong with the benchmarks. Hopefully.

 

[Asus]

Don't be comparing the Stress Test (HL2) to CS: Source benchmarks. Don't get them mixed up either.
I expect ATI cards to be ahead more in the HL2 benchmark than CS: Source.

Also keep in mind that on the second part, Driver heaven set the AA/AF settings to max for both cards (X800XT PE:6XAA/16xAF and 6800U:8xSAA/16xAF). That's why it is falling behind so much. He set them both to max and the 6800 has the option where it does two types of patterns for AA.

 

[^Ben]

Erm no i'm talking about the VST.

 

[blackeye]

If you look at the cs source bencmarks (not to be confused with stress test) now people are saying that ATI cards would have performed faster if it wasnt for the 90FPS cap. But the ATI cards didnt even get near the 90FPS cap so that shouldnt play a part in that. I want to see hardocp do something with the half life 2 stress test like they did with doom 3.

 

[^Ben]

Theres not a 90fps cap :Z

I have to cap mine at 85 otherwise there is tearing.

 

[blackeye]

Theres not a 90fps cap :Z

I have to cap mine at 85 otherwise there is tearing.

In cs source ther is.

 

[^Ben]

Weird that is because when i try it goes to 130fps

 

[Asus]

Yep, I get above 100 FPS in CS:Source. No lock.
Unless of course you have your Refresh rate to 90 and Vsync on.

 

[crabcakes66]

These benchmarks are obviously messed up, no doubt about that. The really sad part was in the thread follow up, where DriverHeaven forum members were trying to defend these benchmarks with blind fanboyism.

ATI's 6xAA looking better then NVidia's 8xSA? Rrrrrright....

bit of a hypocrite there :thumbs:

 

[blackeye]

Yep, I get above 100 FPS in CS:Source. No lock.

I dont have cs source but that review on vr-zone states

So what do we make of our results? Well it seems that there is a cap of about 90 frames per second in the build of Counter-Strike: Source we have. 90 FPS is the fastest we were able to achieve throughout testing, but that score is not published in our benchmarks since we are using average FPS scores throughout our entire demo.

I guess they had vsync on

 

[Gemini][]

Well, it looks like DH scores are about right.
A wealthy friend of mine who has the same rig (ram/hds and monitor are different) scored exactly 1 FPS more than the current DH benchmark using a BFG 6800U, same settings.
I'll try it myself later if I'll ever get the coupon...

 

[^Ben]

Erm newp they are not right.

If you read what i have been saying i get more FPS then what that "review" puts out.

 

[blackeye]

link
That explains a lot.

61.77 - 20% slower than x800xtpe
65.62 - 10% slower than x800xtpe

 

[Arno]

Believe it or not, it's true.

Ok, let me explain it to you then.

ATI's 6xAA = 6x MultiSampling
NVidia's 8xSA = 4x MultiSampling + 2x SuperSampling
MultiSampling only puts Anti Aliasing on the edges of polygons, but SuperSampling renders the whole screen internally at double the resolution.

So basically, what DriverHeaven is comparing is this:
ATI with 1600x1200 @ 6xAA
NVidia with 1600x2400 @ 4xAA
An experienced gamer can spot some difference in the edges between 4xAA and 6xAA, but this is neglectable compared to the two-fold increase in resolution that NVidia is offering. Especially fences, barbwire and foliage look much sharper with 8xSA.

So there you have it. 8xSA offers the best image quality. The downside is of course the massive memory bandwidth that 8xSA consumes.

 

[^Ben]

omg LOLOLOL i just read it and towards the end they dont even use the same settings for the card!

What a bunch of idiots hahahahahahaahahahahahaha!

ATI advertisments splashed all over the place, most of the posters have ATI cards, gawd damn that is funny.

Case closed this site can't be trusted to do a far review.

 

[Malfunction]

This is gonna be a fanboy pitfight! Woohoo!

3...
2...
1...

GO!

Let's get ready to ruuuuuuuuuuumble!!!!!!!


*ahem*

Anyway, I hope everyone realizes what the 6800 UE is anyway...all it is, is a beefy GT, which isn't much. Everyone knows the XT PE is the best card right now. It's obvious, but since many people can't afford a $500 card, and can't find the XT PE in stock, the real battle lies between the 6800 GT and x800 pro.

They have the same retail price, GT has 16 pixel pipelines, x800 pro has 12, GT has SM 3.0 support, x800 pro does not. Also, last time I checked the bechies, the GT, in 1024x768, could pretty much perform equally good as the x800 pro with AA and AF enabled.

If you ask me, I'd say that the x800s are better for those crazy people who play at those extremely high resolutions, like 1600x1200. I'm happy at 1024x768.

In conclusion, we all know that the XT PE will outperform anything right now, though, those benchies don't look right. I can't see how with no AA/AF the UE could be so far behind. I think that guy may be a fanATIc, but whatever. Believe what you want, but I still say the battle is GT vs Pro, and that the GT takes it home.

EDIT: Notice, whoever took those benchies didn't run a test at 1024x768. I now refuse to trust them, sorry. That's probably the most popular resolution anyway. And no, I'm not a fanboy. I'm like a spy in the GPU War, I take whoever I think is better at the time's side.

 

[^Ben]

Yeh i was seriously considering between a GT and pro and went GT in the end

 

[Malfunction]

Yeh i was seriously considering between a GT and pro and went GT in the end

As did I!

Oh, anyone who thinks Nvidia sucks now, read a little further into the thread, like page 2, and see why these benchies are faulty. A guy with an OCed 2500+ CPU, 6800-GT, and everything else the same as the test system for the benchmarks, he got 81 FPS with everything on high, same settings, everything.

Have fun.

 

[tranCendenZ]

Firingsquad is showing results similar to Gamer's Depot, ~20% difference between 6800U and x800xtpe with 61.77 drivers. This is opposed to the ridiculous 40-60% figures claimed by Driverheaven. Half of that will probably be made up by 65.62.

BTW, apparently the benchmark uses both 3dc and sm2.0b, but not sm3.0.

 

[blahblahblah]

I'm going to say this again (like I did with Doom 3). Any of the new high-end cards are awesome.

Also, no mention of 3Dc and SM2.0b is mentioned. Pure speculation.

I hope the firing squad article is representative of the VST and HL2. I want to be able to play 1280 by 1024 with 8 AA (thanks to temporal AA) with 16 AF and still be above 60 FPS.

 

[Darkknighttt]

and once again ati owns nvidia.

:bang:

oh my god! ATI owns NVIDIA in half-life2 and cs:s!!!!! that must mean its a better card!!!!! oh my god!! BTW they didnt own them, that benchmark is wrong.

 

[blackeye]

Firing squad review was much better.
To bad they didnt add the ultra extreme along with the X800 XT. I think they would have performed neck and neck. I was surpised when I looked at the nvidia FX series how much closer they are in perfmance to the radeon 9800's now. Overall the X800 XT is definatly the best card to run Half life 2 (if you have the money) untill we see some ultra extreme benchmarks. I think anandtech had an ultra extreme that they might use if they do a benchmark.
Does anyone know if the game runs at 1792x1344?

 

[Diablo2k]

oh my god! ATI owns NVIDIA in half-life2 and cs:s!!!!! that must mean its a better card!!!!! oh my god!! BTW they didnt own them, that benchmark is wrong.

Don't try talking sense into fanboys, it wont work anyways.

 

[Gemini][]

That doesn't mean anything. 4XAA from nVidia isn't equal to 4XAA from ATi. The same goes for the other modes.
We'll have to see if HL2 is using polygons (my guess) or alpha textures for transparency, making SS not useful at all.
Also, HL2 will probably use Centroid Sampling.

 

[Arno]

We'll have to see if HL2 is using polygons (my guess) or alpha textures for transparency, making SS not useful at all.

SuperSampling improves the looks of all textures. It's just that it's the most obvious on alpha textures.
However, I don't think 8xSA works well performance-wise on such a demanding game as HL2. It's better suited for older games. All the more reason not to use 8xSA in HL2 benchmarks and to just stick with the more popular 4xAA.

 

[^Ben]

Well if it's not usefull at all why put it in a benchmark and not test fairly against another card?