Doomtrooper
Veteran
nggalai said:
As STALKER developers put it to me, they had no choice BUT to use CG for Nvidia cards, as they were having all kinds of issues with HLSL...including speed.
How Cg favors NVIDIA products (at the expense of others)
- Thread starter RussSchultz
- Start date
As STALKER developers put it to me, they had no choice BUT to use CG for Nvidia cards, as they were having all kinds of issues with HLSL...including speed.
Doomtrooper
Veteran
DemoCoder said:
Hi,
In fact, the graphics engine is developed on Radeon9700
We demand special support from FX driver guys, 'cause in pure/standart
DX9 it's impossible so run the engine on FX, at the moment.
Yes, FX will be slightly faster, but in the margin of several
percentages...
So, don't worry, your card will be excelent performer in
S.T.A.L.K.E.R.
--
Best regards,
Oles V. Shishkovtsov
GSC-Game World
oles@gsc-game.kiev.ua
Going by what I'm told Democoder
Doomtrooper
Veteran
Date: Friday, February 14, 2003, 5:48:38 PM
First some broken thoughts, then a logical examination into the endless possibilities.
First let’s consider something that resembles the current situation in shader performance (very roughly). Let left column be standard DX9 performance, right CG.
Clearly this benefits NVIDIA; however, it also benefits the game developer as he now has a game that runs better on a potentially significantly greater number of machines. We will guess that if ATI's potential performance increase of 10% which while giving them back the performance crown is hardly worth their time if the performance threshold of the game is already being met (say 120fps+ through all settings).
The previous example might be more closely related to overall performance than that of shaders... just to try and keep complaints down:
Here, there is even less motivation for ATI to develop a back end as they are winning by a large margin either way. Let's say in the next generation the picture changes a bit though....
Now, it is certainly more valuable to ATI to acknowledge CG. (If your wondering why the shader performance didn't increase for the R4XX, I am letting 1 represent the score of the performance leader and displaying the lesser's as a percentage of that score). This works out well for all parties concerned at the developer and ATI get increased performance (retains competitiveness vs. losing it) and NV cuts the ATI lead in half. Perhaps, as some fear NV would try to limit the implementation of CG so that competitors’ products can only use the default profile. However, if ATI develops a backend for CG, there is really nothing NV can do to stop game developers from using it (as suing the people that are going to make the games that will [or wont!] run on your hardware does not seem like a viable plan if NV wants to stay in business for any length of time.)
Game genre FPS
Performance threshold (60fps):
Here, we can see another possibility with regards to performance in that CG here provides NV with the valuable service of boosting the performance of their card to acceptable levels despite giving an even greater increase to ATI.
In thinking about the situation analytically, CG in my eyes does not directly benefit NV but rather all those companies who trail the performance leader. Thus, CG currently benefits NV more than ATI; however, the results of CG do seem odd if NV is indeed intent on reclaiming the performance crown. Somewhat ironically, NV may be counting on CG to aid them in this quest, but at the same time, it might aid their adversaries enough to push them across the competitive threshold.
So what we have in CG is a proposal to:
1. Improve overall performance
2. Increase capabilities
3. maintain an equal or reduced workload
In general, I have no fear of CG succeeding or failing as there are really only two possibilities: the industry advances faster because of cg, or at the same rate as it would have without cg. From my viewpoint, I consider it a win/win situation. Of course, I probably just wasted a large amount of time in trying to understand something that will only have one result. Well, future reference I suppose......
development costs, company resources, time, effort, producing code marginally faster than dxhlsl in some cases actually slower in others, games that use cg right now? don't really see how CG is that beneficial to NVIDIA right now
If this is true, CG will most likely become the standard unless it has competition which produces a greater net gain in performance.
If the previous statement was true, then this hardly matters. ATI wouldn't have much to fault in CG because I’m pretty sure they stated that DXHLSL is already performing optimally on the R3XX. The only complaint they could have is that it does increase the performance of their competition, but this is hardly valid unless companies are expected to role over for their competition in a capitalistic system. That's not right is it?
code once: overall increase in performance = better gaming experience = more games sold = greater developer support... At least that is how I see it.
This is pretty much true... let us examine a few cases.[/quote]
First let’s consider something that resembles the current situation in shader performance (very roughly). Let left column be standard DX9 performance, right CG.
Code:
NV3X R3XX
150% 100%
0.70 | 1.05 1.00 | 1.00 (max 1.10)The previous example might be more closely related to overall performance than that of shaders... just to try and keep complaints down:
Code:
NV3X R3XX
150% 100%
0.50 | 0.75 1.00 | 1.00 (max 1.10)
Code:
NV4X R4XX
150% 100%
0.90 | 1.35 1.00 | 1.00 (max 1.40)Game genre FPS
Performance threshold (60fps):
Code:
NV4X R4XX
150% 155%
0.75 | 1.125 1.00 | 1.55 (of 60)
45 | 67.5 60 | 93In thinking about the situation analytically, CG in my eyes does not directly benefit NV but rather all those companies who trail the performance leader. Thus, CG currently benefits NV more than ATI; however, the results of CG do seem odd if NV is indeed intent on reclaiming the performance crown. Somewhat ironically, NV may be counting on CG to aid them in this quest, but at the same time, it might aid their adversaries enough to push them across the competitive threshold.
So what we have in CG is a proposal to:
1. Improve overall performance
2. Increase capabilities
3. maintain an equal or reduced workload
In general, I have no fear of CG succeeding or failing as there are really only two possibilities: the industry advances faster because of cg, or at the same rate as it would have without cg. From my viewpoint, I consider it a win/win situation. Of course, I probably just wasted a large amount of time in trying to understand something that will only have one result. Well, future reference I suppose......
Doomtrooper
Veteran
Evildeus said:Hmmm, shouldn't this date make you wonder why
No, as the same issues affect the entire FX line-up with exception to the NV35. I assume you feel these hardware issues exposed here and many other sites are just magically disappearing ?
Doomtrooper
Veteran
They switched to a Nv35 about a 2 months ago to start 'optimizing' for being a technical partner and all.
Bottom line is, not everyone will have a 5900 to play Stalker with, in fact approx 3% of the people buying the title will have one.
Bottom line is, not everyone will have a 5900 to play Stalker with, in fact approx 3% of the people buying the title will have one.
StealthHawk
Regular
Tridam said:
No, AFAIK I am correct. Someone ran 3dmark03 with a gfFX5800 with PS1.4 vs PS1.1. There is no reason why his results should be inaccurate, and unless someone more trustworthy contradicts said results I'll believe them.
http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=13945
StealthHawk said:
Actually you're right and I'm right
In fact I've checked it up with FX5200 and I've assumed it was true for FX5800 and FX5600. I was wrong. I've just finished some more tests (not with 3dmark):
FX5600 shading power:
PS 1.1 : 1300 Mops/s (285 MPix/s)
PS 1.4 : 1300 Mops/s (281 MPix/s)
PS 2.0 : 650 Mops – texturing ops (81 MPix/s)
FX5200 Ultra shading power:
PS 1.1 : 1300 Mops/s (283 MPix/s)
PS 1.4 : 650 Mops/s – texturing ops (105 MPix/s)
PS 2.0 : 650 Mops/s – texturing ops (80 MPix/s)
With FX5200, ps_1_4 is really slow but it seems good with FX5600.
ninelven said:
I rather thought the "proposal" of Cg was to utilize a high-level language coupled with an efficient compiler in order to avoid the mind-numbing tedium of hand-coding shaders in assembly.....?
I have no fear of Cg whatever from any perspective. I agree that since we're getting HLSL from M$ and for OpenGL that Cg isn't going to impact the industry to much of any degree. Those that find it useful will no doubt use it, just as those who don't will use the alternatives.
Doomtrooper
Veteran
ninelven said:
Hardly, HLSL is outperforming CG on a 5600, so how can that above statement be correct. A developer has a much better compiler to use that has already been optimized for all hardware, vs. one.
The evidence is in this thread, and even on Nvidia hardware.
Radeon 9800 Pro HLSL : 125 MPix/s
Radeon 9800 Pro Cg : 100 MPix/s
GeForce FX 5600 HLSL : 11.2 MPix/s
GeForce FX 5600 Cg : 12.4 MPix/s
GeForce FX 5600 HLSL_pp : 14.8 MPix/s
GeForce FX 5600 Cg_pp : 13.8 MPix/s
GeForce FX 5600 HLSL AA/AF : 7.0 MPix/s
GeForce FX 5600 Cg AA/AF : 6.1 MPix/s
http://www.beyond3d.com/forum/viewtopic.php?t=6864&postdays=0&postorder=asc&start=20
HLSL is faster in 2 out of 3 tests on Nvidia Hardware, so what is the incentive again
That statement is wrong.
I was under the impression that CG was a work in progress (i.e. improving). Please forgive me if I was mistaken.
This would be an advantage of Cg over assembly sure. However, I was adressing some of the potential issues regarding why a person might choose one hlsl over another.
This would be an advantage of Cg over assembly sure. However, I was adressing some of the potential issues regarding why a person might choose one hlsl over another.
Doomtrooper said:
Doomtrooper, there is no way to be optimal for all hardware. You wouldn't want a single device driver for all video cards and require register level compatability to make it work, and you don't want a single compiler for all video cards.
You want the compiler to be part of the device driver, and you want each vendor to ship their own optimizer that optimizes it at runtime.
Your crusade against Nvidia is blinding you to the real issue which is to have multiple backend optimizers be part of DX9. You won't admit it, because you perceive it as an Nvidia originated idea and no one can admit that Nvidia has any good ideas.
All Microsoft's compiler can do is perform generic optimizations like common subexpression elimination, copy propagation, and dead code elimination. It cannot schedule instructions (reorder them to take advantage of each card's architecture) nor alter it's register allocation strategy based on card architecture.
Some people expect the device driver to do this on the shader assembly, but that is essentially trying to have the device driver reverse engineer the full semantics of the source code out of compiled binary at runtime, and you won't get the same level of optimizations as if you start with the original source.
Since both HLSL and GLSLANG contain semantics which are not expressible in shader languages (e.g. loops, branching, etc) some of that information will be erased by the time it reaches the device (e.g. branches turned into predictation) which means if the hardware actually contains *real branches* it will have to "infer" this intent somehow from a batch of predicated instructions.
I say the device driver should contain the backend of the compiler, and the front end merely does generic optimizations and stores the result as a serialization of the internal representation of the compiler.