GFFX Tests / Test Apps
- Thread starter Dave Baumann
- Start date
whenever I have need to bench 3D hardware, I have used as an old game test Need For Speed High Stakes modified version. basic test is fraps average framerate on defined replay.
(own track with the visibility set with editor as max as game engine allows: every trackblock all way to front clipping plane is drawed on every frame. game mode is Hot Pursuit and as a cars I am using Mini Cooper ST with 7756 polygons per car. no HUD and Automatic replay camera. d3da.dll rendering library replaced with newer one from NFS PU. average calulation starts from green light, stops to green light.)
only problem is, that with 9700 / GF FX, there won't be much difference because both are pushing fps to the max (which is 60 fps in this case.) with low resolution and without FSAA. with 6xFSAA and 16x AF on 1280x1024 and up there should be some differences. Game is CPU limited, due it doesn't support HW T&L and does all geometry setup on CPU. (just adding more high polycount (>8000 polys) cars on race affects directly on fps with lower CPUs.) so, it could be used as practical FSAA/AF performance test.
(own track with the visibility set with editor as max as game engine allows: every trackblock all way to front clipping plane is drawed on every frame. game mode is Hot Pursuit and as a cars I am using Mini Cooper ST with 7756 polygons per car. no HUD and Automatic replay camera. d3da.dll rendering library replaced with newer one from NFS PU. average calulation starts from green light, stops to green light.)
only problem is, that with 9700 / GF FX, there won't be much difference because both are pushing fps to the max (which is 60 fps in this case.) with low resolution and without FSAA. with 6xFSAA and 16x AF on 1280x1024 and up there should be some differences. Game is CPU limited, due it doesn't support HW T&L and does all geometry setup on CPU. (just adding more high polycount (>8000 polys) cars on race affects directly on fps with lower CPUs.) so, it could be used as practical FSAA/AF performance test.
i would also like to see results from codecreatures benchmark. here is a working link:
http://public.planetmirror.com/pub/majorgeeks/benchmark/codecreatures_benchmark_pro_setup.exe
an AF-quality comparison to R9700 and GF4Ti with samX's AF-tester and 3DMark03's AF-testscene would be cool too..
http://public.planetmirror.com/pub/majorgeeks/benchmark/codecreatures_benchmark_pro_setup.exe
an AF-quality comparison to R9700 and GF4Ti with samX's AF-tester and 3DMark03's AF-testscene would be cool too..
I hear Aquanox is very heavily shader based - and I believe Aquanox 2 is out now. Any chance of using these games engine (Krass engine I believe). I am not sure what level of shaders they are running - maybe contact massive development?
The head of the technical developmeny at Massive on the Krass engine project is Ingo Frick, frick@massive.de <frick@massive.de>
The head of the technical developmeny at Massive on the Krass engine project is Ingo Frick, frick@massive.de <frick@massive.de>
Luminescent
Veteran
There is some info here: http://www.massive.de/german/index_ger.html?technology_krass_ger.html on the Krass engine.
How detailed are the 3DMark03 tests? Did you try using frame-based rendering or the image quality tools? I'd like to see comparisons using every feature 3DMark03 has with the GF4. Hopefully you have the Pro version. And since you're calling this a preview I suspect you're not going to show benchmark comparisons with the R300? If you are going to compare the R300, then do the same 3DMark03 tests too.
Tommy McClain
Dave: 3DMark03 wil not be used in tests relating to a reference FX (not the case if we come across retail boards). Before anyone asks I had agreed to this as I thought it better the site gets the opportunity to test one without 3DM03 than not at all. I will keep my word on this.
Tommy McClain
Dave: 3DMark03 wil not be used in tests relating to a reference FX (not the case if we come across retail boards). Before anyone asks I had agreed to this as I thought it better the site gets the opportunity to test one without 3DM03 than not at all. I will keep my word on this.
Bastards! Hopefully you're going to state this up front in your preview, right? And please tell me you pre-ordered a GFFX Ultra? I seriously doubt you will be able to get a retail board from a board vendor.
Anyway, I was just about to edit my message to ask if you were going to use 3DMark2001 as well. Any chance of at least that? Not that it will help much.
Tommy McClain
Anyway, I was just about to edit my message to ask if you were going to use 3DMark2001 as well. Any chance of at least that? Not that it will help much.
Tommy McClain
Ok, a suggestion. Here's DX9 tutorial that includes source code / binary for example render :
http://www.dimension3.host.sk/web/project.php?pid=Project7&page=1&PFriend=0
Uses shaders 1.4, 2.0, 3.0 and Cook-Torrance & Oren-Nayar lighting. No bechmark itself, but maybe someone could add simple benchmark option or include the those lighting models for other shader bench (Ilfirin ? ).
http://www.dimension3.host.sk/web/project.php?pid=Project7&page=1&PFriend=0
Uses shaders 1.4, 2.0, 3.0 and Cook-Torrance & Oren-Nayar lighting. No bechmark itself, but maybe someone could add simple benchmark option or include the those lighting models for other shader bench (Ilfirin ? ).
AzBat said:
AzBat: there's no better word than "bastards!" to describe my feelings right now. It was ok, to announce that they do not support nor endorse using 3DMark03, but this (giving press samples only to sites that agrees not using 3DMark03 for benchmarking.) is something that crosses the line a bit too much.
Nappe1 said:
AzBat: there's no better word than "bastards!" to describe my feelings right now. It was ok, to announce that they do not support nor endorse using 3DMark03, but this (giving press samples only to sites that agrees not using 3DMark03 for benchmarking.) is something that crosses the line a bit too much.
I totally agree and had they pulled this shit with me when I was doing benchmark testing for JPA I would have made sure it was known to the readers and customers. I never once had a hardware vendor refuse to send me hardware that I requested. Especially if the hardware had already been announced and was within a couple of weeks of shipping. Nor did they give me any requirements for accepting such hardware.
I sure hope NVIDIA reconsiders this stance. If they don't then I sure hope to hell they get nailed for it. And I sure don't want to see them allowing 3DMark03 testing in the future if they decide their newer hardware will provide results substantially better than their competition. Would be very hypocritical to say the least.
Tommy McClain
Tommy, they didn't refuse to send me hardware, I just thought it would increase my chances of seeing a board at all without running 3DM, given the number of reference boards made available and the fact that I had resigned myself to not getting one. I still feel this puts B3D, and the readers of B3D, in a better position for seeing one - does the preculsion of this bechmark mean that we still won't do as fair an accurate job as we have done on prior occasions? For that matter, have any of our (p)reviews been deficient without this benchmark, seeing as its new?
Note: I am using 2001SE.
Note: I am using 2001SE.
Archaeolept
Newcomer
Did Nvidia suggest that 3DMark2K3 not be used? Did they broach the subject first?
I mean, I certainly don't think this will harm the review. I have never been let down by any of your reviews, but I'd still like to know whether you just pre-emptively suggested this (angleing to cadge a card), or else whether Nvidia was dropping hints.
If you feel you can't say anything because you don't want to risk your sources over at Nvidia, that's fine.
On topic: I wouldn't mind some in-game screenshots as far as anisotropic quality goes. There still seems to be a bit of confusion about what exactly the FX's settings do, and how exactly they compare w/ ATI's
I mean, I certainly don't think this will harm the review. I have never been let down by any of your reviews, but I'd still like to know whether you just pre-emptively suggested this (angleing to cadge a card), or else whether Nvidia was dropping hints.
If you feel you can't say anything because you don't want to risk your sources over at Nvidia, that's fine.
On topic: I wouldn't mind some in-game screenshots as far as anisotropic quality goes. There still seems to be a bit of confusion about what exactly the FX's settings do, and how exactly they compare w/ ATI's
DaveBaumann said:
With the way you phrased your words, it definitely seemed that you were ASKED to not run 3DMark03. You did say you "agreed to this". If that's not the case, then it's not as bad as I initially thought. Though I would like to know if that concession helped your chances of getting hardware?
DaveBaumann said:
No, but had NVIDIA actually made the requirement then it could have shown your review as not being as impartial as we might have thought.
Although I understand why you did it. I think it would have been better if you had not volunteered to do that, but instead waited to see if they would have asked you to do so anyway. If they hadn't and you still thought it was better for your readers to make a few concessions in your testing, then by all means do what you think is best. As long as you don't jeopardize your journalistic integrity in the meantime.
DaveBaumann said:
No, but that's not point. It was the idea that NVIDIA would even consider make testing requirements as precursor to receiving one of their products. No matter if it was 3DMark03, Doom3, or even Deer Hunter 5 for that matter. If you ask me, that's a very distasteful idea, if in fact it were true. If they haven't done this, then great.
DaveBaumann said:
That's good to hear. Were any other concessions made?
Tommy McClain
Typedef Enum
Regular
I guess that's the way it is now...If you plan to run benchmarks that expose their latest/greatest in a not-so-great light, you won't get a board.
Seriously...What's next? I've lost track already...
And to think nVidia was encouraging the use of TreeMark just a few years ago.
Seriously...What's next? I've lost track already...
And to think nVidia was encouraging the use of TreeMark just a few years ago.
DaveBaumann said:
I think it would be nice if you could find an OEM who'd be more forthcoming. nVIdia's already gotten tons of negative publicity from its strange stance on something that's always been a synthetic benchmark anyway--3D Mark. Now they've had their say it simply should be a non-issue to them, thus you should not have to walk on ice around them because of it. Always remember your position: you can promote them--they cannot promote you. You are in the driver's seat and always have been.
Whatever small, worthless quantity of Ultras trickle out into the market, it still won't be worthwhile to put one through any kind of paces since people won't be able to buy it--hence at best you'd be doing little more than propagandizing a company over a product not available on the open market. Until the FX Ultra's ship in quantity to store shelves I think the attention should be paid to the GF FX Standard model (400/800) that will, presumably, ship in quantity. Besides, you don't need an Ultra to solve these puzzles, right?
Hi Dave,
I must admit that I'm a bit surprised about this turn of events. Well, I guess that's a call you've made and it's late to whine about it.
Given that a few people have been mildly surprised by the dramatic increase of the results with different drivers when running 3DMark03 I was really looking forward to see your in-depth analysis regarding what's going on with these drivers. There's few people around that would have the capacity and knowledge to understand the issue regarding to driver optimizations and it would have been interesting to see your take on that topic.
I'm afraid well have to just wait for now. Hopefully that's going to be something to look forward in your future (p)reviews anyway?
Cheers,
AJ
I must admit that I'm a bit surprised about this turn of events. Well, I guess that's a call you've made and it's late to whine about it.
Given that a few people have been mildly surprised by the dramatic increase of the results with different drivers when running 3DMark03 I was really looking forward to see your in-depth analysis regarding what's going on with these drivers. There's few people around that would have the capacity and knowledge to understand the issue regarding to driver optimizations and it would have been interesting to see your take on that topic.
I'm afraid well have to just wait for now. Hopefully that's going to be something to look forward in your future (p)reviews anyway?
Cheers,
AJ
