Doom3 as benchmark...
I made this: http://www. driverheaven. net/showthread. php?s=&threadid=17491 post over at DriverHeaven. net about using Doom3 as a benchmark. . . lets see if we can bring attantion to the issue eh? ;P for those non DriverHeaven users: http://www.
This topic was started by El_Coyote,
I made this: http://www.driverheaven.net/showthread.php?s=&threadid=17491 post over at DriverHeaven.net about using Doom3 as a benchmark... lets see if we can bring attantion to the issue eh? ;P
for those non DriverHeaven users:
http://www.rage3d.com/board/showthread.php...readid=33690672
for those non DriverHeaven users:
http://www.rage3d.com/board/showthread.php...readid=33690672
Participate on our website and join the conversation
This topic is archived. New comments cannot be posted and votes cannot be cast.
Responses to this topic
Should I feel cheated buying an ATi card or will it still handle Doom ?
(I bought a Radeon9700 PRO)
(I bought a Radeon9700 PRO)
"FX 5900 is gonna be the best for Doom3 because of the stencil shadow z-culling the UltraShadows implement... sadly, thats why i see it as a flawed program for benchmarking."
Yeah, It would be unfair to give an advantage to an accelerator that accelerates those things that need accelerating. Dammit. Where is my Voodoo 5? (Best IQ and FSAA yaknow...)
Yeah, It would be unfair to give an advantage to an accelerator that accelerates those things that need accelerating. Dammit. Where is my Voodoo 5? (Best IQ and FSAA yaknow...)
I don't see the logic in your argument. Benchmarking with games is not a synthetic test. Most cards do some things slightly different producing slightly different rendering results. Just because you do not notice them does not mean they are not there. This is what I would call good trade-offs or optimizations. Basically it would come down to "will the user see a signigicant degradation of rendering quality?" when tweaks/optimizations/cheats are used to improve rendering speed. In some scenarios it will be beneficial to go for speed and in others to scarifice it for better rendering quality.
If ATI cannot render Doom 3's shadows in hardware assisted mode, then this would be cosidered a flaw/lack of feature in that product. This is a negative aspect to said product and positive feature to the product which can do it. Very simple. Both products will show shadows. One will be able to do it faster because the hardware you paid for is accelerating it as an accelerator should.
Perhaps if ATI had a different implementation of shadowing and Carmack refused to support it, it would be a more interesting issue. As it stands now, ATI cannot hardware assist for shadowing and Nvidia can. This is not an excuse to say "Oh, that makes the comparison unfair" as long as the rendered result looks the same.
Imagine, if you will, if an "accelerator" did not have PS/VS/T&L as was the case not very long ago. Would you think it logical to say "Well, this benchmark is not fair because Voodoo 5 does not have TnL and the Geforce series do" ? Rubbish. You would say "With TnL you get what you pay for. It greatly improves performance in those titles that are coded to use it." At least I hope you would. This is the job of a specialized hardware renderer: to off-load rendering functions from the CPU.
The bottom line here is that Doom 3 will look the same on a Radeon 9700/9800 and a Geforce FX 5900, but the Geforce FX will render it much faster because it has more features and they are being used. Any conclusion as things stand now would be "The Geforce FX 5900 smears the Radeon 9800 in Doom 3." Hiding this truth is like saying "we cannot determine which car is faster because one runs on nitrous and the other doesn't."
The Radeon 9700/9800 Pro will probably run Doom 3 "fine". Just do not expect it to outperform a card with more funtions and a code path which is optimized to use those functions to best advantage.
If ATI cannot render Doom 3's shadows in hardware assisted mode, then this would be cosidered a flaw/lack of feature in that product. This is a negative aspect to said product and positive feature to the product which can do it. Very simple. Both products will show shadows. One will be able to do it faster because the hardware you paid for is accelerating it as an accelerator should.
Perhaps if ATI had a different implementation of shadowing and Carmack refused to support it, it would be a more interesting issue. As it stands now, ATI cannot hardware assist for shadowing and Nvidia can. This is not an excuse to say "Oh, that makes the comparison unfair" as long as the rendered result looks the same.
Imagine, if you will, if an "accelerator" did not have PS/VS/T&L as was the case not very long ago. Would you think it logical to say "Well, this benchmark is not fair because Voodoo 5 does not have TnL and the Geforce series do" ? Rubbish. You would say "With TnL you get what you pay for. It greatly improves performance in those titles that are coded to use it." At least I hope you would. This is the job of a specialized hardware renderer: to off-load rendering functions from the CPU.
The bottom line here is that Doom 3 will look the same on a Radeon 9700/9800 and a Geforce FX 5900, but the Geforce FX will render it much faster because it has more features and they are being used. Any conclusion as things stand now would be "The Geforce FX 5900 smears the Radeon 9800 in Doom 3." Hiding this truth is like saying "we cannot determine which car is faster because one runs on nitrous and the other doesn't."
The Radeon 9700/9800 Pro will probably run Doom 3 "fine". Just do not expect it to outperform a card with more funtions and a code path which is optimized to use those functions to best advantage.
Yes but I haven't seen Carmack do anything concerning ATi optimising...
been tossing this in my mind for a while..
why did Id run their DoomIII on a R9700pro last
E3, yet only optimize it for Nvidia? doesnt make any
sense to me..
id say its pretty optimized for both, or not at all..
On the other hand, using unfinished and unreleased
software as benchmark is questionable..
optimizations is prolly not final, and prolly not even
either, one path might be more close to finished then
the other..
and it wouldnt be a exact picture of how the games
perform with the cards that are tested with it..
But then again, why would Id send out a special test
demo if the techpart wasnt finished?
Since from what i gather the code used for benchmarks
was a special edition given out by ID, and not teh old
leak..
as stated here for instance:
http://www.anandtech.com/video/showdoc.html?i=1821&p=21
and they have said that what remains on the game is
content development..
i might add that Carmack has backends for both ATi and Nvidia
cards, and before 5900U came into the picture 9700pro was the
fastest card to run doom3 on, and that was why they did use it
at last years E3..
first time they showed the engine they used a Geforce3, since
that was the most highend at that time..
and Nvidia has admited that 5900U was made more or less to
run Doom3.. wich is pretty obvious with the shadow techs they
put in.. Its only a good thing, Doom3 isnt going to be the last
game with volumeshadows...
why did Id run their DoomIII on a R9700pro last
E3, yet only optimize it for Nvidia? doesnt make any
sense to me..
id say its pretty optimized for both, or not at all..
On the other hand, using unfinished and unreleased
software as benchmark is questionable..
optimizations is prolly not final, and prolly not even
either, one path might be more close to finished then
the other..
and it wouldnt be a exact picture of how the games
perform with the cards that are tested with it..
But then again, why would Id send out a special test
demo if the techpart wasnt finished?
Since from what i gather the code used for benchmarks
was a special edition given out by ID, and not teh old
leak..
as stated here for instance:
http://www.anandtech.com/video/showdoc.html?i=1821&p=21
and they have said that what remains on the game is
content development..
i might add that Carmack has backends for both ATi and Nvidia
cards, and before 5900U came into the picture 9700pro was the
fastest card to run doom3 on, and that was why they did use it
at last years E3..
first time they showed the engine they used a Geforce3, since
that was the most highend at that time..
and Nvidia has admited that 5900U was made more or less to
run Doom3.. wich is pretty obvious with the shadow techs they
put in.. Its only a good thing, Doom3 isnt going to be the last
game with volumeshadows...
ATi aren't at fault for that are they?
ATi aren't at fault for that are they?
its said the leak was traced to them..
and they were supposedly the only ones
with the techdemo i think..
but its not really in carmacks intrest to make
it work less good on ATi cards, they have alot
of cards out in the gamers computers and Id
will only loose in doing so..
Besides what sais its less optimized for ATi cards?
I mean they ran the E3 demo in the 9700 pro so
they must have done some work on it..
the 3.4 Cat had a bug that made the 9800 run Doom3
bad, but they prolly just didnt optimize it for doom3 as
they will release new drivers when the game is out..
and optimizing drivers for games that arent done yet
is pretty much a waste of time..
Id might have snuck the benchmark version out without
telling ATi about it tho, as revenge perhaps, knowing that
they wouldnt adapt their drivers to it yet..
and give Nvidia a chance to tweak their drivers to run it..
altho the NV35 is basicly made for doom3 engine with its
ultrashadow techs.. so it could prolly winn in a fair fight aswell..
Nice one - you just contradicted yourself...
but you then say :
explain yourself :P
and optimizing drivers for games that arent done yet
is pretty much a waste of time..
but you then say :
and give Nvidia a chance to tweak their drivers to run it..
explain yourself :P
Nice one - you just contradicted yourself...
but you then say :
explain yourself
generally you dont release a special edition of a game
to use as benchmark before its released..
But in this case Id acctually did...
so generally you dont really need to optimize the driver
until the game is nearing release, but now Id sent out
benchmark version, and maybe they did it to get back
at ATi.. by not telling em.. i dunno..
Not telling them is a bit weird since the fact there is a special edition for Nvidia is all over the net...so I'm not sure either :P
yeah, the benchmark version is prolly equally optimzed
for each card, but Id might have given Nvidia the chance
to optimize their drivers for it, and leaving ATi unknowing
about it..
so the fault could lie in the driver optimizations and not
the doom3 code itself..
And that could be a way of Id to get even hehe...
well its just speculations...
for each card, but Id might have given Nvidia the chance
to optimize their drivers for it, and leaving ATi unknowing
about it..
so the fault could lie in the driver optimizations and not
the doom3 code itself..
And that could be a way of Id to get even hehe...
well its just speculations...
FX 5900 is gonna be the best for Doom3 because of the stencil shadow z-culling the UltraShadows implement... sadly, thats why i see it as a flawed program for benchmarking.
hl2.. i dont know.. but 9800 and fx5900 will propably be so close that it doesnt really matter
"Gabe Newell: If you look at the surveys we've done of the hardware our customers have (on Speakeasy.net), there are two companies that dominate - NVIDIA and ATI. We've spent a lot of time working with both companies. Right now for the DX9 generation, ATI is significantly ahead in terms of absolute performance and price/performance. We are working pretty hard with NVIDIA to try to optimize the situation with regards to getting their performance up to snuff. We spend more of our time with ATI working on new features like high-dynamic range rendering (HDR) which is a new technique to create more life-like lighting." :P
