AA KILLS q4 performance

[Corporate Thug]

before i start, rig is in sig

ok, i tried running Q4 @ 1600*1200 no aa/no af. ran smooth as butter

then i tried 1600*1200 w/ 0aa/8af. still pretty smooth. in heavy fights, FPS momentarily drops to 45-55 FPS

then i tried 1600*1200 4aa/8af and in the fight sequences, FPS drops to ~15-20

is this common?

all setting are on high

 

[imported_Bleh]

Yes especially if you have a Nvidia card because they dont have hardware help to render AA like ATI, thus Nvidia cards have a higher preformance hit when using AA.

 

[mOeeOm]

Originally posted by: Bleh
Yes especially if you have a Nvidia card because they dont have hardware help to render AA like ATI, thus Nvidia cards have a higher preformance hit when using AA.

Uh oh...you shouldn't have said that. Its like Rollo's call.

 

[imported_Bleh]

Originally posted by: mOeeOm

Originally posted by: Bleh
Yes especially if you have a Nvidia card because they dont have hardware help to render AA like ATI, thus Nvidia cards have a higher preformance hit when using AA.

Uh oh...you shouldn't have said that. Its like Rollo's call.

This was not ment to flame in anyway its just the facts.

 

[mOeeOm]

Originally posted by: Bleh

Originally posted by: mOeeOm

Originally posted by: Bleh
Yes especially if you have a Nvidia card because they dont have hardware help to render AA like ATI, thus Nvidia cards have a higher preformance hit when using AA.

Uh oh...you shouldn't have said that. Its like Rollo's call.

This was not ment to flame in anyway its just the facts.

 

[NokiaDude]

Originally posted by: Bleh
Yes especially if you have a Nvidia card because they dont have hardware help to render AA like ATI, thus Nvidia cards have a higher preformance hit when using AA.

Where did you pull that from? your a$$?


1600x1200 4aa/8af is alot for even a 7800GTX. Throw in another 7800GTX and you can dance all day. I get about 60fps solid in most newer games like Fear and FarCry @ 1280x1024 4aa/8af with my single 7800GTX.

 

[Acanthus]

I would love to see playable frames from any ati single card in quake 4 at those settings (playable being 30 min frames).

Its not magic, youre just setting the bar a little high for a single card system. Id step down to 1280x960 and give that a shot at 4xAA.

ATi does not have special hardware calls for AA either. It's implemented through drivers, and yes ATi cards do not take as much of a hit, however, it would not be playable frames either.

 

[Matt2]

Anand's 1600x1200 4xAA Q4 performance

According to Anand, u should be getting 55 fps avg.

Looks like that extra GB of RAM you keep swearing you dont need is finally needed

 

[Corporate Thug]

Originally posted by: Matt2
Anand's 1600x1200 4xAA Q4 performance

According to Anand, u should be getting 55 fps avg.

Looks like that extra GB of RAM you keep swearing you dont need is finally needed

possibly...

 

[yhelothar]

Originally posted by: Bleh

Originally posted by: mOeeOm

Originally posted by: Bleh
Yes especially if you have a Nvidia card because they dont have hardware help to render AA like ATI, thus Nvidia cards have a higher preformance hit when using AA.

Uh oh...you shouldn't have said that. Its like Rollo's call.

This was not ment to flame in anyway its just the facts.

Uh what facts? The performance difference depends on the AA algorithm used. SuperSampling AA, which is what nvidia uses with the GF4, besides quincunx, gets the biggest performance hit. However, it results in the highest quality AA. Super Sampling basically doubles the resolution for every 2xAA, and shrinks it back down. So if you're running 800x600 with 2xAA, that's like running 1152x864.
GFFX and above cards, AFAIK, all used multisampling, which just blurs the edges, resulting in a lower image quality, albiet MUCH lower performance hit.

 

[Acanthus]

Originally posted by: Matt2
Anand's 1600x1200 4xAA Q4 performance

According to Anand, u should be getting 55 fps avg.

Looks like that extra GB of RAM you keep swearing you dont need is finally needed

FX57

 

[DAPUNISHER]

Originally posted by: Matt2
Anand's 1600x1200 4xAA Q4 performance

According to Anand, u should be getting 55 fps avg.

Looks like that extra GB of RAM you keep swearing you dont need is finally needed

If he turned off AF he probably would average that, he is using AA+AF. And isn't he referring to minimum FPS in the first post, not the avg?

Originally posted by: AcanthusFX57

At the res&settings he uses, the difference between the 146@2.6ghz and the FX57 are insignificant.

 

[Matt2]

I dont think AF is going to cause a 30 fps decrease.

 

[Cookie Monster]

You guys are forgetting AT actually uses AF in the benchs. They just dont show that on the graphs, so its very misleading.

The Q4 bench was 16x12 4xAA 8xAF.

 

[blckgrffn]

Try the ram upgrade, it has worked for me. And the 85.91 drivers from tweaksrus are nice, gave me a very solid boost in FEAR.

Nat

 

[bjc112]

Originally posted by: Cookie Monster
You guys are forgetting AT actually uses AF in the benchs. They just dont show that on the graphs, so its very misleading.

The Q4 bench was 16x12 4xAA 8xAF.

You know that for sure?

 

[DAPUNISHER]

Originally posted by: Cookie Monster
You guys are forgetting AT actually uses AF in the benchs. They just dont show that on the graphs, so its very misleading.

The Q4 bench was 16x12 4xAA 8xAF.

Could you point me to where they stated 8x AF is a default for testing now? TIA. I still think Shady is complaining about the dips for minimum FPS and not his avg though too. But he'll have to clearify on that. BTW, is 2gb of ram a default now too? They don't state that in the test setup.

 

[Acanthus]

Originally posted by: DAPUNISHER

Originally posted by: Matt2
Anand's 1600x1200 4xAA Q4 performance

According to Anand, u should be getting 55 fps avg.

Looks like that extra GB of RAM you keep swearing you dont need is finally needed

If he turned off AF he probably would average that, he is using AA+AF. And isn't he referring to minimum FPS in the first post, not the avg?

Originally posted by: AcanthusFX57

At the res&settings he uses, the difference between the 146@2.6ghz and the FX57 are insignificant.

I consider 5 fps insignificant too if youre looking at the game and not at a counter. But, we are looking at counters and graphics.

 

[Pete]

HQ mode in D3 and D4 enables 8xAF by default, I thought.

Bleh, you're a bit misleading with your post. Both nV and ATI have hardware to help with AA, it just seems that ATI takes a relatively smaller hit. It's been a general trend since the 9700, AFAIK, but it's no longer as obvious as when the 9700P had a 256-bit memory bus and the 5800 had a 128-bit one.

 

[VIAN]

Did you expect free AA or something. Some games' weak point is AA and some games are AF. Especially when you are playing at that resolution, I don't doubt you should have problems. I would try and take out the AF to free up some memory bandwidth for AA if you are really into AA. Don't forget, There is a lot of crap on screen that Q4 has to smooth out like shadows and just all the objects that are in Q4. Especially in a heavy firefight, which flashes of light everywhere, bullets, many enemies and more shadows.

Originally posted by: Pete
HQ mode in D3 and D4 enables 8xAF by default, I thought.

I couldn't find what setting they set the game to. From this, I have to assume that this is due to very poor technical writing. The author's barely tell you about the settings in the game. Sure, we can assume that it is done at High Quality, but everyone knows what assuming does.

According to HardOCP, which seems employ better technical writers and a better performance testing method, High Quality enables 4xAF. Ultra Quality enables 8xAF. They're benchmarks Claim that a GTX OC should be able to run Q4 at 1600x1200 with 2xTR MSAA and HQ enabled, which auto enables 4xAF. This gives and average of 54fps and a minimum of 22fps. And to avoid any confusion. The CPU is a FX55 coupled with 2GB of RAM.

 

[Corporate Thug]

Originally posted by: VIAN
Did you expect free AA or something. Some games' weak point is AA and some games are AF. Especially when you are playing at that resolution, I don't doubt you should have problems. I would try and take out the AF to free up some memory bandwidth for AA if you are really into AA. Don't forget, There is a lot of crap on screen that Q4 has to smooth out like shadows and just all the objects that are in Q4. Especially in a heavy firefight, which flashes of light everywhere, bullets, many enemies and more shadows.

didnt expect free AA but wasnt expecting to go from playable to not playable w/ 4x AA

 

[Cookie Monster]

Ive sent derek a email regarding benchmarks:

We did enable 8xAF in most tests these days. A lot of games incorporate
AF into options like "texture filtering" in bf2 and "high quality" mode
in doom3/quake4. In games with the in game setting we use the in game
setting and what ever mode the developer decided to choose as their
highest setting -- which is often 8x but sometimes higher. In fact,
there isn't a game we test in which we enable AF in the driver anymore.
If we have to pick the mode explicitly we do select 8xAF.

Aside from the fact that we've been using 8xAF for a while, I tend to
prefer it when gaming. With 16xAF, the angular dependency becomes much
more apparent in most implementations (ATI's high quality mode
definitely looks better, and we hope NV will follow suit).

Which brings us to the last question ... we are trying to compare apples
to apples as much as possible. At the highest image quality modes, ATI
and NVIDIA have different image quality. For example, it's possible to
get better AF out of ATI's x1000 series, but it's possible to get better
AA with NVIDIA's transparency AA than ATI's adaptive AA.

That being said, you've inspired me to work on an article that compares
the highest quality ATI modes to the highest quality NVIDIA modes in
both image quality and performance. It may take a while to get it done,
but I'll try to get it out before the end of the month.

For most other articles, we'll still stick to noaa/noaf and 4xaa/8xaf.
midrange and low end users won't be able to get much out of the higher
quality features anyway, so looking at them is really most useful on the
high end.

Thanks,
Derek Wilson

We are going to see the benchmark that we always wanted.

Thanks Derek. :thumbsup:

 

[VIAN]

I still think that should have been done with the card reviews. F*ck apple to apple performance. Are we trying to make it harder to decide what to buy or easier? Use the card to it's max potential, tell me what it does. Is Nv faster than ATI at the highest settings? Is that gain at a loss of IQ? Show me pictures. Let me compare. Give me prices and one will emerge. There is no apples to apples.

 

[Cookie Monster]

Originally posted by: VIAN
I still think that should have been done with the card reviews. F*ck apple to apple performance. Are we trying to make it harder to decide what to buy or easier? Use the card to it's max potential, tell me what it does. Is Nv faster than ATI at the highest settings? Is that gain at a loss of IQ? Show me pictures. Let me compare. Give me prices and one will emerge. There is no apples to apples.

True, life aint fair enough to have apple to apple comparisons.

 

[Cookie Monster]

bump