Rollo's Far Cry SM3.0 Benchmarks

[nRollo]

The control panel was used to set the antialiasing level, which doesn't work with FarCry unless set specifically in the FarCry profile (which was not done here)

I wondered about what would happen if you DID enable the Far Cry Profile in the nVidia control panel, and then control AA/AF within the Far Cry control panel.
System A64 3000+/768MB CAS 2.5 DDR 333/Ref GF6800 NU/Windows XP Home
Driver rev. 61.45
All Far Cry settings Very High, Trilinear Enabled, default rendering, vsynch off
All demos looped multiple times, not averaged as they were always within 1.5 fps of each other

10X7X32 0X 0X
Research
SM2- 80fps
SM3- 95fps

Volcano
SM2- 132fps
SM3- 139fps

10X7X32 2X 8X
Research
SM2- 48fps
SM3- 66fps

Volcano
SM2- 82fps
SM3- 90fps

10X7X32 4X 8X
Research
SM2- 36fps
SM3- 53fps

Volcano
SM2- 61fps
SM3- 67fps

 

[nRollo]

BTW- I'll be out of town 7/4, will answer questions (if any) 7/5

 

[vshah]

thanks....nice results. high scores too. hmmmmm

 

[Acanthus]

ROLLO CHEATS HAX0R OMGZ ATI AA AF NO CHEAT WTF PWNED.






Nice results.

What chipset rollo?

 

[Keysplayr]

So. That means that the FarCry 1.2 patch with some SM3.0 capability along with the DX9.0c update DOES increase performance when AA/AF is controlled throught the in game options. I am glad he did this on his own and posted it. Now, we really do have some facts and dont have to go by a review site.

Man I can't wait to see what happens when a 100% coded SM3.0 title comes out.

 

[nRollo]

Originally posted by: Acanthus
ROLLO CHEATS HAX0R OMGZ ATI AA AF NO CHEAT WTF PWNED.






Nice results.

What chipset rollo?

Snuck down here before we have to leave:

LOL at first, MSI K8TNeo with VIA K8T800 is my chipset.

 

[GeneralGrievous]

10X7X32 4X 8X
Research
SM2- 36fps
SM3- 53fps

Hmm. This looks really odd. Your results go even past the 30% theoratical maximum benefit that SM3 provides. At least I think its 30% I might be wrong.

 

[imported_Apophis]

Where did you get Farcry 1.2?

 

[nRollo]

Originally posted by: keysplayr2003
So. That means that the FarCry 1.2 patch with some SM3.0 capability along with the DX9.0c update DOES increase performance when AA/AF is controlled throught the in game options. I am glad he did this on his own and posted it. Now, we really do have some facts and dont have to go by a review site.

Man I can't wait to see what happens when a 100% coded SM3.0 title comes out.

Keys:
Yes it did have significant performance increases and I saw no visual anomalies running the "Far Cry" path in the Forceware drivers.

Whether or not there are differences in using the Far Cry path and using the straight path, I don't know. I didn't notice any mip bands, texture cracking or anything like that, but I didn't do screenshots and zoom either. I thought the game looked amazing, and you could actually see the speed jump when you switched from SM2 to SM3.

Of note to you, I did use the same demos everyone else did (Research and Volcano) they are bundled in as a benchmark tool in the 1.2 patch and I'm sure designed to show the places where SM3 does make a difference. I don't think this is a bad thing personally, what would be the point of showing where it didn't?

In any case, to me the SM3, with AA/AF app controlled on game specific path, looks like very promising technology. I'm anxious to see what a engine totally developped with SM3 can do. It looks like the implications for outdoor scenes with lots of repeated plants are a good increase in performance.

 

[nRollo]

Originally posted by: GeneralGrievous

10X7X32 4X 8X
Research
SM2- 36fps
SM3- 53fps

Hmm. This looks really odd. Your results go even past the 30% theoratical maximum benefit that SM3 provides. At least I think its 30% I might be wrong.

It struck me as odd as well, so I ran those settings multiple times twice. Set AA/AF in Far Cry, exited game, started again, checked to make sure it was on 4x8X, ran benches.

 

[nRollo]

Originally posted by: Apophis
Where did you get Farcry 1.2?

I have a buddy that works at nVidia, the patch will be out soon for everyone, and anyone here should be able to replicate my results.

 

[imported_Apophis]

Originally posted by: Rollo

Originally posted by: Apophis
Where did you get Farcry 1.2?

I have a buddy that works at nVidia, the patch will be out soon for everyone, and anyone here should be able to replicate my results.

Any idea when? They've been saying "soon" for a long time now

 

[nRollo]

Originally posted by: Apophis

Originally posted by: Rollo

Originally posted by: Apophis
Where did you get Farcry 1.2?

I have a buddy that works at nVidia, the patch will be out soon for everyone, and anyone here should be able to replicate my results.

Any idea when? They've been saying "soon" for a long time now

I can't remember if I'm supposed to give this out, but it might be posted at nVnews. It really will be soon though, I can guarantee you if I have it, it's not a months off beta. (I'm not exactly Anand or Wavey Dave, although it would be nice)

 

[imported_Apophis]

Ah yeah I looked on there and it said Monday

 

[Avalon]

Nice benches.

 

[chsh1ca]

Originally posted by: Rollo

The control panel was used to set the antialiasing level, which doesn't work with FarCry unless set specifically in the FarCry profile (which was not done here)

I wondered about what would happen if you DID enable the Far Cry Profile in the nVidia control panel, and then control AA/AF within the Far Cry control panel.

I take it:

We appologize for the error, and have updated our graphs and analysis accordingly.

was not posted at the time you grabbed that quote?

 

[nRollo]

It was, but I misunderstood it.
Upon reading it again, I don't believe it.
Notice in my benchmarks how adding 4x8X cut my performance in half? If AA/AF were not being done, why would I lose half my performance? (besides the fact that it looked different?) If it were only controlling the AF, why wouldn't my framerate be the same at either level of AA at the same resolution/AF?

I'll try some "control the AA/AF in the nVidia control panel instead of Far Cry when I get a chance"

 

[ronnn]

Originally posted by: keysplayr2003
Now, we really do have some facts and dont have to go by a review site.

:roll: A good review by someone with the experience and hardware to bench properly. Yes it is positive towards Nvidia...... Also in my mind shows that Crytek is determined to use the advantages of all new cards - to promote the engine. :shocked:

 

[nRollo]

Originally posted by: ronnn

Originally posted by: keysplayr2003
Now, we really do have some facts and dont have to go by a review site.

:roll: A good review by someone with the experience and hardware to bench properly. Yes it is positive towards Nvidia...... Also in my mind shows that Crytek is determined to use the advantages of all new cards - to promote the engine. :shocked:

Ronn:
1. I don't bench cards for a living. In the Far Cry profile, it says "Application Controlled" for AA/AF, so I thought that would be worth trying. (setting it in the application)
2. No matter what the settings I ran were doing in regard to AA/AF, there was a big improvement in SM3, unless you think we can't trust the counter anymore?
3. It's not like the old days when you typed stuff like "timedemo map_demo1" at the console and off it goes. For example, in my follow up of this I found out I have to set 8X aniso in the nVidia control to use it in the app, but that I can set 4X in the app. When I set 4X aniso in the nVidia control, it set it automatically to 4X in Far Cry. When I put it in Far Cry profile, it disabled my ability to set AA in the nVidia control panel at all. So there are a few more variables to consider here than typing a command line. Pardon me for trying to get you guys some info at different settings than the reviews.
4. "hardware to bench properly"? What is wrong with my hardware? Seems like a fairly reasonable mid range pc to me? Do you have something against A64 3000s and 6800NUs?


Anyway, I reran some benches, this is what I found:

nVidia Control on "Global" (std) 10X7 4X4X set in application:
Research SM2 40fps SM3 55fps
Volcano SM2 67fps SM3 72fps

nVidia Control on "Global" (std) 10X7 0X4X in application only:
Research SM2 60fps SM3 76fps

nVidia Control on Far Cry 10X7, set to 4X AA in app (disabled in nVidia control) and 8X AF in both nVidia and Far Cry
Research SM2 39 SM3 51
Volcano SM2 56 SM3 63

Anyway, think it's pretty clear on any setting that I got some decent increases in performance using SM3, especially on the Research demo.

 

[nRollo]

Woot! Vindicated!
THG findings equal Rollo findings
Research demo
THG 10X7 4X4X 40/57fps
Rollo 4X4X 40/55fps
(although this may make my 4x8X 36/53 suspect)


THG 10X7 0X0X 80/98fps
Rollo 10X7 0X0X 80/95fps

What was someone saying about me not having any experience or the right hardware? :beer:

http://graphics.tomshardware.com/graphic/20040705/farcry-12.html

At least in FarCry, the GeForce 6800 cards show more or less pronounced performance gains in those levels that use special SM 3.0 shaders. In other levels, on the other hand, we couldn't spot much of a difference, aside from an general performance advantage when the flashlight is turned on

This conclusion flies in the face of what ATi has been preaching for the last few months. In many of our FarCry tests, the NVIDIA cards can now pull ahead of their rivals from ATi.

 

[Jeff7181]

Unlike some, I appreciate your efforts, Rollo. Very interesting to say the least.

 

[nRollo]

Originally posted by: Jeff7181
Unlike some, I appreciate your efforts, Rollo. Very interesting to say the least.

Thanks Jeff, I'm trying. More benches to come, hopefully will give people an idea of what kind of performance they can expect with nVidia's $300 MSRP solution.

 

[GeneralGrievous]

The vanilla 6800 is performing extremely poorly on those benches SM3 or not.

 

[VisableAssassin]

Originally posted by: Rollo

Originally posted by: Jeff7181
Unlike some, I appreciate your efforts, Rollo. Very interesting to say the least.

Thanks Jeff, I'm trying. More benches to come, hopefully will give people an idea of what kind of performance they can expect with nVidia's $300 MSRP solution.

Im with Jeff any more I look to actual users for input on cards not review sites.
So Rollo whenever you get a GT or ultra... bench it again please

 

[nealh]

Rollo is the 6800NU very loud?..does it overclock wiith any artifacts with Far Cry?