PDA

View Full Version : Farcry is a "directx9" game?


Nv40
04-28-04, 01:50 AM
well.. it was very interesting this article..
its seems that the game is mostly 95% ps.1.1 in all video cards ,the water is done in ATi and NVidia with PS1.1 ,introduced by NVidia in the Geforce3.. 4 years ago..:D and looks all the hype about the "first" directx9 game.. and this is still a directx8 game. use a few couple of Ps2.0 shaders for the indoor light.
http://firingsquad.com/hardware/far_cry_nvidia/default.asp

there is almost no way for Nvidia to replace shaders there in the FX ,a get great benefits since there is no lower version than ~PS1.x, :D
the biggest performance drop in that game comes from the tons of Polygons the game have ,and for the use of FP32 precision in Fx cards. with all this ,there is no way for Nvidia or any company to compare PS2.0 vs PS3.0 in farcry since the game is done almost entirely in ps1.1. doh!! :) the only diference when using PS2.0 in the indoors lighting is that higher precision is required.

wow.. so seems that Hl2 and Stalker will be the only big titles in this year with decent use of PS2.0 and even PS3.0 support. its quitte funny that many times some says Farcry looks better than Hl2.. :)

LycosV
04-28-04, 03:17 AM
Just because it used PS1.x for the water and other effects doesn't mean it can't look better than HL2. I think it is a rediculous assumtion that everyone makes that just because PS 2/3 is used it automatically looks/is better.

I think it just goes to show that PS2/3 isn't neccesary to make things look good since the most talked about Far Cry graphic (the water) doesn't use PS2.

ChrisRay
04-28-04, 03:55 AM
Just because it used PS1.x for the water and other effects doesn't mean it can't look better than HL2. I think it is a rediculous assumtion that everyone makes that just because PS 2/3 is used it automatically looks/is better.

I think it just goes to show that PS2/3 isn't neccesary to make things look good since the most talked about Far Cry graphic (the water) doesn't use PS2.


Heh this is what I have been saying All along. I think Pixel Shader 2.0/3.0 and Precision make a huge deal with Lighting tho.

Edge
04-28-04, 04:45 AM
Funny thing is, the water effects in the "preview" version of HL2 look the same on a TI4200 card as they did in the official videos that were powered by a 9700. Perhaps even HL2 is using less PS 2.0 shaders than they're letting on?

hordaktheman
04-28-04, 10:17 AM
Still doesn't make sense...

They benched using an outdoor level (w/ minimal, if any, dx9 usage according to them), so why are nv3x cards so much slower than r3xx cards?

If it's mostly 1.1, then nv3x cards should be similarly fast as r3xx cards, right? Also, there's a large performance difference between version 1.0 and 1.1, which was ONLY supposed to affect 2.0 and not 1.1 shaders.

Nv40
04-28-04, 01:51 PM
Still doesn't make sense...

They benched using an outdoor level (w/ minimal, if any, dx9 usage according to them), so why are nv3x cards so much slower than r3xx cards?

If it's mostly 1.1, then nv3x cards should be similarly fast as r3xx cards, right? Also, there's a large performance difference between version 1.0 and 1.1, which was ONLY supposed to affect 2.0 and not 1.1 shaders.

most likely because of the Fp32. for ps1.1 even the precision of geforce4 should be more than enough. it seems to me that in some benchmarks i have seen the NV40 is not as fast with PS1.x ,than it is with PS2.0. so maybe that have something to do ,with the small diferences in Fps.

so to improve performance ,Nvidia can replace tons of ps1.1 ,with a few long PS2.0/ps3.0 shaders.. in the NV40. just like in GUnmetal benchmark where NVidia use the flexibility of Nv3x ,and use a very long VS2.x ++ shaders ,vs multipassing in ATi .in PS2.0 test ,the Nv40 is ~2x the r350 performance. is an irony that NVidia is very fast this time in PS2.0 and PS1.4.. but not that fast running PS1.1 :)

hordaktheman
04-28-04, 02:42 PM
But I thought 1.1 shaders don't use fp32? I thought only 2.0 shaders use fp32 or fp16, while 1.1 shaders were integer based. And even if the integer ALU's were altogether removed from NV3x and it used floating point ops instead, then surely the card should default to fp16 for 1.1 or 1.4 shaders, and NOT fp32?

Also, then where does the performance increase come from? According to the patch's readme-file, there is only an increase in ps2.0 shader performance, so there shouldn't really be any increase in that particular benchmark, especially if there are no (or few) 2.0 shaders in that benchmark runthrough (as is being claimed).

My point is, if it's mostly ps1.1, then ATi's cards shouldn't be performing THAT much faster than the FX cards.

Btw, I'm only referring to the NV3x line, not NV40.

It'd be easy to find out if FS were to drop down to ps1.1 only and rerun a single test.

ChrisRay
04-28-04, 06:15 PM
Still doesn't make sense...

They benched using an outdoor level (w/ minimal, if any, dx9 usage according to them), so why are nv3x cards so much slower than r3xx cards?

If it's mostly 1.1, then nv3x cards should be similarly fast as r3xx cards, right? Also, there's a large performance difference between version 1.0 and 1.1, which was ONLY supposed to affect 2.0 and not 1.1 shaders.


Because the r300 still has a higher pixel throughput the NV30 series, its just not nearly as big of a deal as Pixel Shader 2.0.

The r300 has a much higher single pixel fill rate and still twice as many shader units, So its not a big surprise. Its just the Nv30 series higher clocks really help it be competetive.

hordaktheman
04-28-04, 07:07 PM
I'm not talking about the performance diff between Nv and Ati cards.

Edit: Oops apparently I did. Still, continue reading below...

I'm talking about the difference between the unpatched and the patched version of the game. FX cards gain some 20% extra fps, while ATi cards do not. Release notes mention an increase in 2.0 shader performance on FX cards. If there were no 2.0 shaders in this particular benchmark run, then there wouldn't be any performance difference between version 1.0 and 1.1 of the game on FX cards. Yet the FS benchmarks show an increase of 20% more fps. So unless the 1.1 patch included optimizations for shaders 1.1, then there shouldn't be any difference in these benchmarks on FX cards, UNLESS shader 2.0 was being used.

-=DVS=-
04-28-04, 10:22 PM
Because the r300 still has a higher pixel throughput the NV30 series, its just not nearly as big of a deal as Pixel Shader 2.0.

The r300 has a much higher single pixel fill rate and still twice as many shader units, So its not a big surprise. Its just the Nv30 series higher clocks really help it be competetive.

If FarCry would be only pixel fillrate limited then 6800 should fly even useing R300 path , yet its only few FPS faster acording to that odd comparision in DH, interesting huh , to bad devs don't tell how things work only PR bull :retard:

All this guessing is frustrateing :eek2: but it does show there is some background fishy things going :eek2:

Nv40
04-28-04, 11:34 PM
this time the Nv40 weakness .. (and all R4xx)seems that will be extremely bandwidth limited benchmarks..so they will end behaving like 8x1 pipes cards in those situatons.. so i expect to se ultra/pros with similar scores in those situations if clocks are clocked the same.

Ninja Prime
04-29-04, 01:42 AM
its quitte funny that many times some says Farcry looks better than Hl2.. :)

Hmmm where did all these "some" get their retail copies of HL2? Yeah thats what I thought... :lame: