Official HardOCP HL2 benchmarks

[Razor1]

My point was from the begining that nV had been working much closer with ID on doom from the very beginning. Everyone else here knows that fact. Its the same thing what is happing with ATI/Valve and again thats the only point I was trying to make.


Please your point is not there, because you left out the fact Carmack didn't make those optimizations, Nvidia did


Until some one post the code I dont think you or I should try to second guess valve. They know more about making games than your or I :)


You're the one that said it uses 90 instructions maybe you want to read over how instruction sets and limitations on different code paths work..... oh yeah we know that the sm 3.0 code path isn't in there, what path is the geforces running on then, not ATI's path because who said the geforce 6 paths are not as optimized as the radeon x800 paths?

I do know alot more then you are giving me credit for, I don't know valve's code but I know the way my code works and how to optimize for different graphics hardware.


Oh yes they are. The big man himself as "said" so. Granted anytime JC says something its always crypitc but yes NV is doing shader replacement in Doom3. Its not that big of a deal. If the game runs faster and the out put is the same then great. However in benchmarks its kind of a cheat...


He never said they were doing shader replacement, so thats not concerete. He said the drivers are doing something funny, most likly something with the AA and AF, not shader replacement. If you remember if nvidia was doing shader replacment they would have a much more pronounced lead with the fx over the 9800 not tied. Its remarkably very similiar to ALL OGL games.



Oh good greif. You have no idea but are just guessing at the reason. Why not wait for some more proof... If you think its just that then trick the device ID and run the becnhmark again. If you get 20 more fps you might be on to something. If you dont..well there ya go...but if its about the same...then there goes your different path theory...


I'm not guessing if its using 90 instructions for that shader it will take one more pass on a native dx9 path. Dave was guessing more then I was. Its not a hardware issue if Valve purposefully did that, and from your own mouth you said it was a 90 instruction shader. And that you reported from where?

[NoWayDude]

I just dont see why people think that Valve on purpose is Making NV run slower than it can...its makes no sense for them as a business move, as a gaming move or any move on vavle's part....
It does not make sense to Valve, or ID or any other dev, but if you have, lets say, 6 million reasons (or any other number we could put in here) then it could make sense for someone else part :smoking:

PS: I tried to run the game with 3Danalyze but it wont let me.Any ideas?

[Razor1]

It does not make sense to Valve, or ID or any other dev, but if you have, lets say, 6 million reasons (or any other number we could put in here) then it could make sense for someone else part :smoking:

PS: I tried to run the game with 3Danalyze but it wont let me.Any ideas?


tried running it too, not working

[NoWayDude]

Does not like Benchemal also.Weird.Anyone know out to bench this game with a third party app?

[Lars]

I recently purchased a Gainward 2400Ultra Golden Sampl GLH (a 6800GT with 1.6ns memory clocked @ 400/1200 Mhz). I bought this card because I really wanted to try out a Nvidia card once again, haven't had one since ATI put the 9700pro out... At any rate, I already own a ATI x800xt videocard, so I've been eager to try out gaming on both cards and get a feel for the differences.

After watching driverheavens benchmark results I felt puzzled, as I've been playing quite a few hours of HL2 ( I'm @ the prison right now, don't know exatly how far that is in the game actually, but I'm enjoying the game, investigating everything, so I'm not trying to finish as fast as possible). What puzzled me were the reported framerates as I've been playing @ 1600x1200x32 with 4xFSAA and 8xAF the whole time with a feel of much better fps than reported by Driverheaven and a few other places. It's been absolutely fluid all along.

I did the timedemos Anandtech used, mainly because they were easily available for download, and this is what I get @ 1600x1200x32 with 4xFSAA and 8xAF, same quallity settings as Anandtech used (and as recommended by HL2)

at_c17_12.dem, 62.5 fps, 5.8, fps variability

at_canals_08.dem, 68.4 fps, 6.0 fps variability

at_coast_05.dem, 93.8 fps, 6.2 fps variability

at_coast_12.dem, 70.5 fps, 5.3 fps variability

at_prison_05.dem, 65.4 fps, 3.2 fps variability

Haven't benched on the x800xt yet, as I'm rather liking this 6800GT card. On the other hand Doom3 isn't feeling any different on the 6800GT then it did on the x800xt card either...

[jbirney]

Please your point is not there, because you left out the fact Carmack didn't make those optimizations, Nvidia did

Who wrote the NV30 path for doom and spent more than a year developing it only to drop it at the end? It wasn't NV. Again everyone knows that NV and ID worked together on Doom3.


You're the one that said it uses 90 instructions maybe you want to read over how instruction sets and limitations on different code paths work..... oh yeah we know that the sm 3.0 code path isn't in there, what path is the geforces running on then, not ATI's path because who said the geforce 6 paths are not as optimized as the radeon x800 paths?

I realize what paths that are used will effect the perfromance. And I dont know what path the ATI cards are using. Since 3Dc is also not being used you could guess that the other ATI features are also not being used...but I dont know so thats just a wild guess..

I do know alot more then you are giving me credit for, I don't know valve's code but I know the way my code works and how to optimize for different graphics hardware.

No I give you credit. Just you can not expect to know all of the details about somone elses shader code when you dont have it in front of you. You have no idea what they are doing with it. Thus your guess to the cause is just as bad/good as the next persons.


He never said they were doing shader replacement, so thats not concerete. He said the drivers are doing something funny, most likly something with the AA and AF, not shader replacement.

From JC himself: (http://www.hardocp.com/article.html?art=NjQyLDE=)

. On the other hand, the Nvidia drivers have been tuned for Doom's primary light/surface interaction fragment program and innocuous code changes can "fall off the fast path" ,and cause significant performance impacts, especially on NV30 class

I took the liberty of Bolding the part we he says NV has tuned their drivers (IE shader replacement). And please note I am not saying this is a bad thing. Just poiting out that both sides are doing it in Doom3.



I'm not guessing if its using 90 instructions for that shader it will take one more pass on a native dx9 path. Dave was guessing more then I was. Its not a hardware issue if Valve purposefully did that, and from your own mouth you said it was a 90 instruction shader. And that you reported from where?

From Brandon last HL2 perfromance review over at FS.


It does not make sense to Valve, or ID or any other dev, but if you have, lets say, 6 million reasons (or any other number we could put in here) then it could make sense for someone else part

Your forgetting about the inclusion of HL2 in ATI cards. How many HL2 vouchers got shipped with ATI cards? I think that would be neat to know. I dont think valve let ATI just have them. Part of that 6 million figure had to cover that. I know I have been playing HL2 free thanks to that 9600 I bought last year :) I mean why should I pay 50 bucks for the game when ATI will do it for me :)

BTW you do know that nV paid about 4 million to Id/Activision for Doom3 "rights" right? I suppose we can run wild with theories there as well.....

[MUYA]

yeah great..can u we get back to HL2 benchies from HardOCP (although it has runit's course) and not what IHV does or what for eh?

thanks

[jbirney]

Edited by MUYA:Please read my post jbirney. Thanks. That's one warning

[muygar]

where are your results ,

mine = 80.00 FPS

hardocptimedemo_canals01


System info: CPU P4 2.8E @ 3.4 GHZ
GPU BFG 6800 GT OC 370/1000
512 MB PC3200 RAM

Settings = 1024 x 768 4 x AA, 8 X AF Details = All High,
Reflect World ,V-sync off

[lmetza]

the timedmeos no logner work, i get some "protocol version 6 missing , current version 7" tyoe of error

[jbirney]

the timedmeos no logner work, i get some "protocol version 6 missing , current version 7" tyoe of error


Usally happens when they update HL2 game itself. And not just HL2, most time demos seem to run under a certian version of that game..update the game and you usally have to make new time demos. Gives Brent over at H something to do :)

[Lars]

the timedmeos no logner work, i get some "protocol version 6 missing , current version 7" tyoe of error

Yes I get the same error. I would have liked to test the same demos on mt other systems as well.