Anand has Farcry 1.2 patch and tests SM3.0
- Thread starter Helevitia
- Start date
Oooooh, I got wood reading that!Ante P said:
Nice benchs. BTW, can you explain the "Dual Depth Shadow Mapping with 3x3 Bilinear Percentage Closer Filter" +333 %Ante P said:
boost? (in english or even french )Mintmaster
Veteran
Ante P, you really should do a geometric mean for tests like these. The test with 650 fps gets 50+ times the weight of a 12fps test, making the increase from 12 to 50 almost meaningless.
Also, when you do an overall average, you should calculate that from the overall result (i.e. 303/290 = 4.5%, except with a geometric mean), not an average of the individual percentages.
Also, when you do an overall average, you should calculate that from the overall result (i.e. 303/290 = 4.5%, except with a geometric mean), not an average of the individual percentages.
Bouncing Zabaglione Bros.
Legend
jvd said:
If it follows current and past trends, Anandtech just screwed something up when they did the benchmarking.
Mintmaster
Veteran
Maybe, but as dig said, it could take months.RickCain said:So will Anand fix the results and give the real deal or just continue to mislead the readers?
In his NV35 article, this page used to have this graph on it.
"The GeForceFX 5900 Ultra does extremely well in Quake III Arena, to the point where it is CPU/platform bound at 1600x1200 with 4X AA/8X Anisotropic filtering enabled"
Yeah right. You'd think he'd at least check if the image output was correct.Pretty ridiculous for a journalist. I only fully trust his numbers when they match up with other web sites. Maybe we can get other web sites like Toms to make a news post of his blunders, or just flood him with email asking for a correction.
Mr. Travis
Newcomer
from x-bit
"NVIDIA seems to have a driver bug that results in very rough shadows in the game. This is unlikely to affect performance seriously as shadows in FarCry hardly require any significant computing power, though, this is something negative we can tell about the new set of drivers."
...
shadows are the most performance busting thing in FC, I'm not sure where they are deriving their opinion on it from...
"NVIDIA seems to have a driver bug that results in very rough shadows in the game. This is unlikely to affect performance seriously as shadows in FarCry hardly require any significant computing power, though, this is something negative we can tell about the new set of drivers."
...
shadows are the most performance busting thing in FC, I'm not sure where they are deriving their opinion on it from...
ellingsen1
Newcomer
jvd said:
Could it be possible that the ATI cards where using the NV path? I read somewhere that Far Cry uses device ID's and the ATI card was being treated in some ways as a NV card because it did not exist when the game was made. This was causing driver glitches when ever high quality or very high settings were used. I think someone said it affected a couple areas in the game. If IRC like in the tops of trees being orange and certain shadows missing. If that's true that might explain it.
caboosemoose
Regular
If anyone cares I found the following:
5950 ultra with patch 1.2 (but not dx9.0c) - low shader precision / quality is fixed, performance as folows:
Volcano level, indoor test, room full of shaders that are blocky in patch 1.1:
16x12x4x8 patch 1.1:
20fps, blocky shader quality
16x12x4x8 patch 1.2:
16fps, shader quality identical to ATI boards
Can confirm 6800 boards also have correct shader precision in 1.2 (that's running dx9.0b, interestingly. Have to finish general testing b4 risking borking my rig with sp2rc2.) with apparently no performance loss.
All my tests done with driver 61.34, FYI.
Also my results are in line with FS, anand have obviously made a total fuck up on this one, not a huge surprise I guess.
5950 ultra with patch 1.2 (but not dx9.0c) - low shader precision / quality is fixed, performance as folows:
Volcano level, indoor test, room full of shaders that are blocky in patch 1.1:
16x12x4x8 patch 1.1:
20fps, blocky shader quality
16x12x4x8 patch 1.2:
16fps, shader quality identical to ATI boards
Can confirm 6800 boards also have correct shader precision in 1.2 (that's running dx9.0b, interestingly. Have to finish general testing b4 risking borking my rig with sp2rc2.) with apparently no performance loss.
All my tests done with driver 61.34, FYI.
Also my results are in line with FS, anand have obviously made a total fuck up on this one, not a huge surprise I guess.
Thanks CabooseMoose.caboosemoose said:
Which sites Chal?Chalnoth said:
Speaking of using Fraps with Far Cry, I've done multiple manual playthroughs of the Research map to verify how reliable, and consistent, charting the game's performance would be with the utility and I had no problems getting repeatable data. No, it's not exactly scientific but IMO it's better than the demo route.
caboosemoose
Regular
Indeed - it's not the best approach for all circumstances, but simply loading up a game level without touch the mouse and keyboard in which the spawn point has all the relevant effects in view (I used precisely that method for my Volcano numbers) with fraps running and noting the number when it settles down gives extremely repeatable results. It's not really what you want to do to get an overall idea of game perfromance, but in my view its the best (and lets face it, quickest) way of testing the difference a change in IQ or detail seting makes.
http://www.xbitlabs.com/articles/video/display/farcry30.htmldigitalwanderer said:
The firingsquad results aren't that far off, either, but you just have to remember that they used their own timedemos, and we know that the NV40's advantage comes from specific scenarios, so performance should obviously depend on which timedemo you choose. So I say their results are also similar.
