Starcraft AA benchmarks: ATi vs nVidia

[akugami]

I find it funny the usual suspects were making a big stink because it was ATI not having something that nVidia did. Now that both sides are largely equal, in fact ATI is looking good, all of a sudden it's not a big deal anymore.

I said before that I didn't find AA in Starcraft 2 a big deal and I'll say it again. It's a twitch game where if you have time to notice the jaggies on individual units, you're playing it wrong.

 

[Athadeus]

Maybe they're benchmarking the menu lol.

 

[Cerb]

After reading the reviews and looking at the pictures and video, is the 40/50 percent hit in performance with AA even worth it visually?
It does not look like it to me. I'm talking about both brands.

The difference between the detail levels, without high-end CPUs, is huge. For me, w/ a C2D and GTX 460, AA is all but free, as my PC can't handle High settings at playable rates (it's not great anyway, but a 50+ unit battle, and it's a <10FPS slideshow), even if I turn AA off. GPU utilization is also lower at high than medium, further indicating a CPU/RAM limitation. So, currently, medium settings w/ high texture is as far as I can go, and I may end up having to drop some of those, too, i I get into big multiplayer fights (I'm still retraining myself, ATM)...all the while, my video card has plenty of headroom.

Even so, with a fast CPU, if that 40% can be regained by using lower detail levels, anyone that prefers AA will do it. It's perfectly fine at low settings. Are you there to lose sleep playing an RTS game that requires FPS reflexes, or are you there to go, "ooo, pretty..." ? Pretty as gravy, when your system has performance to spare, is great; but you aren't missing out on the game by not running at Ultra.

 

[Grooveriding]

Maybe they're benchmarking the menu lol.

Didn't you hear ? The menu is the hardest thing on your cards!

 

[happy medium]

No benchmark in game so far. Everytime I look at the numbers I laugh out loud. Very misleading and totally inaccurate. That PCP review should be deleted.
__________________

I found the way they benchmark at Pc perspective.


Quote:

'' For testing purposes we are using the SC2 replay system that allows you to save and playback an entire multiplayer match. I played through several dozen matches, slaving away for our readers, to find instances of performance lag, low frame rates and what I would call your "typical" scenarios. I narrowed my replays down to a pair that seemed pretty interesting, but each provided very different frame rates.

One consisted of a modest battle between lower grade forces early on in a match and frame rates ranging in the 60-100 mark. The second was a more mass-onslaught of Zerg and Protoss that brought our system further down in frame rate.My first reaction was to use this second replay as our testing scenario for benchmarking but I found some interesting results when looking at the CPU utilization."

http://www.pcper.com/article.php?aid=958&type=expert&pid=1

Seems ok with me.:hmm:

 

[Grooveriding]

No benchmark in game so far. Everytime I look at the numbers I laugh out loud. Very misleading and totally inaccurate. That PCP review should be deleted.

Yeah no doubt, it's a flawed and incorrect review.

My numbers on 480 SLI are not much different from their single 480 numbers :thumbsdown:

I don't think they tested enough of what actual gameplay demands of the cards.

At least the xbit review seems accurate :thumbsup:

 

[Jules]

i wonder how a gtx 460 1gb @ 850/2000 will do vs a 5850 in this game

 

[3DVagabond]

i wonder how a gtx 460 1gb @ 850/2000 will do vs a 5850 in this game

Whatever else it does, it'll be sucking more power. That's even if you run the 5850 at the same clocks, where it'll still be faster.

 

[Keysplayr]

i wonder how a gtx 460 1gb @ 850/2000 will do vs a 5850 in this game

At those speeds it would most likely overtake the 5850.

 

[Aristotelian]

i wonder how a gtx 460 1gb @ 850/2000 will do vs a 5850 in this game

Shouldn't you be more curious as to how an overclocked 460 would do versus an overclocked 5850? Apples to apples comparisons, please.

 

[Qbah]

I found the way they benchmark at Pc perspective.

(...)

http://www.pcper.com/article.php?aid=958&type=expert&pid=1

Seems ok with me.:hmm:

Happy, people are telling you it's not representative of actual gameplay. And they are right. The PCP results are silly. Did you see my actual gameplay results? I used a campaign mission that is very demanding and with MSAAx4 at 1920x1080 I got 32AVG over a 10 minute gaming run. And my rig is by no means slow (CPU@3.2GHz, GPU@850/1150).

And for me the IQ improvement is very much worth the performance hit. Can't see myself playing games without AA tbh.

 

[SlowSpyder]

i wonder how a gtx 460 1gb @ 850/2000 will do vs a 5850 in this game

I would imagine that it'd be a bit faster than the 5850, or at the very least match it. But 5850's are no slouch in the overclocking department, either. It seems like just about every 5850 is good for 950+MHz.

 

[Grooveriding]

Happy, people are telling you it's not representative of actual gameplay. And they are right. The PCP results are silly. Did you see my actual gameplay results? I used a campaign mission that is very demanding and with MSAAx4 at 1920x1080 I got 32AVG over a 10 minute gaming run. And my rig is by no means slow (CPU@3.2GHz, GPU@850/1150).

And for me the IQ improvement is very much worth the performance hit. Can't see myself playing games without AA tbh.

I used AA from the outset, but due to this thread and wanting to compare my frames to that horrible pcper review, I tried running it without AA.

Cut scenes look really bad without it, and there are a lot of cut scenes, tons of crawling and jaggies. Even in gameplay you notice improvements on structures and geography, I wouldn't call it necessary in game, but for the cut scenes, certainly.

It runs a whole lot faster without AA though. The forced implementation is really taxing.

 

[Wreckage]

http://www.pcper.com/article.php?aid=958&type=expert&pid=1

Seems ok with me.:hmm:

Wow at 1920x1200 with 4xAA the 460 keeps up with the 5870. In fact it has a better minimum frame rate. :thumbsup:

 

[happy medium]

Shouldn't you be more curious as to how an overclocked 460 would do versus an overclocked 5850? Apples to apples comparisons, please.

That 5850 Apple is 75$ more. Fair?

Cheap ass gtx 1gb factory overclocked to 800 core.
http://www.amazon.com/gp/product/B003WUXNPO/ref=oss_product

 

[happy medium]

Happy, people are telling you it's not representative of actual gameplay. And they are right. The PCP results are silly. Did you see my actual gameplay results? I used a campaign mission that is very demanding and with MSAAx4 at 1920x1080 I got 32AVG over a 10 minute gaming run. And my rig is by no means slow (CPU@3.2GHz, GPU@850/1150).

And for me the IQ improvement is very much worth the performance hit. Can't see myself playing games without AA tbh.

From what I read, they use real life situations, whats wrong with that?
I think the only thing wrong with it ,is that it gives a different result then Xbit labs.

 

[Voo]

I think the only thing wrong with it ,is that it gives a different result then Xbit labs.

..and anyone else who really plays the game.

If I compare what I get with my OCed 4870 when playing, compared to their 4850 numbers - just doesn't add up

 

[happy medium]

Wow at 1920x1200 with 4xAA the 460 keeps up with the 5870. In fact it has a better minimum frame rate. :thumbsup:

But now we will get the bad benchmarks arguement.

Simple fact anyone who pays 285$ for a 5850 to play this game, when you could pay 210$ for a factory overclocked 800 core gtx 460 for the same performance is just dumb.

To be honest, I'll say it again, the next drivers for both brands will probrobly bring big changes, we are arguing about reviews that mean nothing.

 

[happy medium]

And anyone else who really plays the game.

? I don't understand?

From what I read, they use real life situations, whats wrong with that?

You forgot to answer this part?

 

[Grooveriding]

..and anyone else who really plays the game.

If I compare what I get with my OCed 4870 when playing, compared to their 4850 numbers - just doesn't add up

Not worth the argument. That pcper review is inaccurate and does not represent real gameplay. The people arguing against, don't even own the game, and don't even own the hardware themselves, who knows why they think they can make any sort of claims. It's just common sense to see that in the xbit review, http://www.xbitlabs.com/articles/video/display/starcraft2-wings-of-liberty_6.html#sect1 , if they can produce a situation where those framerates are the result of a given bench, that the game is that demanding. Just because a flawed review can use a weak bench environment means nothing if you are going to encounter much more taxing gameplay in the game.

Good reviews use the most demanding conditions to give you a truthful perspective on the performance you'll get with given hardware.

Also, that review is not the subject of the thread, just report him for derailment.

I think it's great to see an accurate review though, with all the fud going around, it's good to see some light shone on what different cards are really capable of.

Short story, if you want to play Starcraft 2 with AA at 1680x1050 or below, a 460 is your best bet, but it dies and can't perform at a higher resolution.

Want to play at 1920x1200 or above, get a 5870 or a 470 or 480.

I'll bench a playback of a busy 3v3 game on my system and post some benches at 4xAA on Ultra to give a realistic context for performance.

 

[happy medium]

But now we will get the bad benchmarks arguement.

Simple fact anyone who pays 285$ for a 5850 to play this game, when you could pay 210$ for a factory overclocked 800 core gtx 460 for the same performance is just dumb.

To be honest, I'll say it again, the next drivers for both brands will probrobly bring big changes, we are arguing about reviews that mean nothing.

Again for those who missed it.

 

[Jules]

Shouldn't you be more curious as to how an overclocked 460 would do versus an overclocked 5850? Apples to apples comparisons, please.

Nope thats why i asked.

 

[Qbah]

From what I read, they use real life situations, whats wrong with that?
I think the only thing wrong with it ,is that it gives a different result then Xbit labs.

You even quoted the part that explains why the results are meaningless for playing... Those are replays, not gameplay. I am telling you, you won't be getting those numbers in-game with the components they test. Not even close to that. So you can see how the cards compare in replays, but why does that matter? Well, it doesn't. You play the game and not watch replays (for the most part anyway).

 

[Aristotelian]

That 5850 Apple is 75$ more. Fair?

Cheap ass gtx 1gb factory overclocked to 800 core.
http://www.amazon.com/gp/product/B00...ef=oss_product

That wasn't my point. My point was stated in more detail by SlowSpyder, so let's have a quick look at what he said:

I would imagine that it'd be a bit faster than the 5850, or at the very least match it. But 5850's are no slouch in the overclocking department, either. It seems like just about every 5850 is good for 950+MHz.

So, if one is going to compare the performance of an overclocked 460 to a stock clocked 5850, one is not obtaining a proper basis for comparing the two cards. Overclocked cards ought to be compared with overclocked cards, unless someone is looking for a skewed vision of the performance of the cards.

Nope thats why i asked.

That you are not curious raises the question: what could examining the overclocked performance of a 460 prove over the stock performance of a 5850? Surely not that the 460 is a better buy, for the 5850 can be overclocked as well. Once the overclocking potential is ascertained for both cards then cost effectiveness can be brought into the picture. But unfortunately for the Nvidia crowd, benchmarks posted in this very thread show that ATI cards dominate Nvidia cards in SC2 performance. As has been posted many times, before you point to minimum framerates, a graph denoting the occurrence/duration and so on of these minimum framerate spikes is needed to understand the impact that these spikes may have on the gaming experience. So clutching to the minimum framerate statistic of the 5850 will not help save Nvidia cards in this regard.

From the very first post in this thread:

The 5870 is pretty close to the GTX485, while the 5970 is miles ahead.

The GTX460 is also ahead of the 5830, but is slower than the 5850.

The 5870 is very close in performance the 480. This was posted much earlier in this thread:

looking at that review, the 480 needs to come down in price to $400 to match the 5870

Yet barely anyone seems to have taken Outrage's lesson to heart. It's disappointing that all of these debates turn out to be easily interpreted as a few forum members essentially advertising Nvidia cards (and providing helpful links to purchase the cards directly) and others attempting to explain that more information is needed to make an informed purchase.

So to answer your question, happy medium, what would be fair is for you to say something like: At the ~200 price level, the 460 gives the best performance, and at the ~400 price level, the 5870 gives the best performance. I merely posted initially to say that overclocked should be compared with overclocked: apples to apples, not apples to oranges.

 

[Wreckage]

But now we will get the bad benchmarks arguement.

Simple fact anyone who pays 285$ for a 5850 to play this game, when you could pay 210$ for a factory overclocked 800 core gtx 460 for the same performance is just dumb.

To be honest, I'll say it again, the next drivers for both brands will probrobly bring big changes, we are arguing about reviews that mean nothing.

That pcper review shows the 460 doing very well against even the 5870 @19/12. I'm guessing NVIDIA will get back their marketshare no problem.