Anand has Farcry 1.2 patch and tests SM3.0
- Thread starter Helevitia
- Start date
http://www.anandtech.com/video/showdoc.html?i=2102&p=10 
Sometimes SM3.0 just makes X800 look like a 5800
Sometimes SM3.0 just makes X800 look like a 5800
Evildeus said:Is it the only question you have? I think it's not the place. But well, evga and Gainward are selling some parts if you are willing to pay the pricejvd said:
why not ?
I don't have any questions .
Well mabye one but i asked awhile ago and never got an answer. and its ati related so i don't want to ask it here .
I don't have any questions because the benchmarks that showed the most gain were recored by nvidia to be played back by anand .
THe obviously represent best case senarios . But the performance gain on the ones anand choose showed pretty good perfromance.
Yet once again he is using older ati drivers and brand new nvidia drivers
hovz said:
well they are up to 4.7 beta 2s all i can tell us is they are indeed faster than the offical 4.6s . Can't really tell how much because i only have a slow athlon xp at the moment.
Anyway we got new cats (4.6 offical) on the 6th . So we should get new offical drivers in a week or so
Its hard to compare to the old patch because anand never mentions the names of the demos used in the past.
But it doesn't apear that sm 3.0 is that big of a deal when you look at the big diffrence.
It seems the new drivers and the patch benfit 6800s no matter the shader model used and the gains from going from the patch at sm 2.0 to sm3.0 are small on average. Nice but not a deal breaker.
pocketmoon66
Newcomer
The CryTek boys gave a demo and talk at the recent Nvidia Dev conference in London (see my other post).
SM3 is used to implement Geometry Instancing (batching of identical meshes) which gives up 40% gain but more usually 0-20% gain. They used this headroom to push out the point at which they render sprites (imposters) rather than meshes. They stated that this made vegitation rendering look better.
The reason given for not being able to implement the multi-lights-in-single-pass shader in PS2 was instruction count, although ps_2_1 seems to offer an extened limit on shader length over ps2 so not sure here. Crytek quoted upto 20% gains but IIRC this was demo's with stencil shadows off. So perhaps in regions without stencil shadowing you see a bigger gain.
HDR required 16 bit fp blending. Not sure if this is something that the patch will contain ??
SM3 is used to implement Geometry Instancing (batching of identical meshes) which gives up 40% gain but more usually 0-20% gain. They used this headroom to push out the point at which they render sprites (imposters) rather than meshes. They stated that this made vegitation rendering look better.
The reason given for not being able to implement the multi-lights-in-single-pass shader in PS2 was instruction count, although ps_2_1 seems to offer an extened limit on shader length over ps2 so not sure here. Crytek quoted upto 20% gains but IIRC this was demo's with stencil shadows off. So perhaps in regions without stencil shadowing you see a bigger gain.
HDR required 16 bit fp blending. Not sure if this is something that the patch will contain ??
gokickrocks
Regular
i havent read anandtech or nvnews' review, so dont know if its been asked, but does the performance increase come from the patch (as in game optimizations and fixing bugs) or the use of sm3.0? i guess what im trying to say is, if we force the 6800 to use the ps2.0 path does it see a similar increase?
