On 3DMark Issues
- Thread starter Dave Baumann
- Start date
For those of you wondering about all the (extra) bolding, in PMs with ED I've said I would add extra emphasis to help illustrate the meaning of my sentences, and I'm continuing that here in an attempt to draw his attention to important phrases.
ED, I was simply comparing something new to my prior criticism of Kyle's statements, and stating why. You then responded to that, to oppose my viewpoint again with nothing new. The repetition began then.
Eh? Since 3dchipset's comments were new, yes I was. That was the only thing I ever attempted to address as new here, and pointing to Kyle's comments recognized the similarity of the criticism I proposed. Your reply, however, was not anything new...it was simply a restatement of your belief with no substantation whatsoever.
You have startlingly selective memory, as that reflects exactly the stance Futuremark communicated prior to this. You responded to it the same way then as well.
Hmm....there comes a time for some people when you can only consider them a flat out liar. For me, as far as you're concerned, this is that time.
The statement you say is "new":
Let's go back in time a bit to 17 May 2003, 12 minutes after the hour:
One would presume that this, on 17 May 20003, 40 minutes after the same hour, signifies that you actually read what was stated:
Or maybe I give you too much credit.
What it looks like to me is that you are using the pretext of "new" here to pretend that your repetition of viewpoint is validated, and that there is significance in GRilla's statement beyond repetition because you do so. I say that, because it is in reply to that assertion that you decided to disagree, and then use a clearly and completely false statement to support that disagreement.
Well, what do you know, you're claiming your viewpoint is validated by repetition.
Yes, I do view it as enlightening for you to so clearly illustrate what you view as support for your viewpoint: Falsities and Innuendo.
You sure showed me.
Evildeus said:
ED, I was simply comparing something new to my prior criticism of Kyle's statements, and stating why. You then responded to that, to oppose my viewpoint again with nothing new. The repetition began then.
Eh? Since 3dchipset's comments were new, yes I was. That was the only thing I ever attempted to address as new here, and pointing to Kyle's comments recognized the similarity of the criticism I proposed. Your reply, however, was not anything new...it was simply a restatement of your belief with no substantation whatsoever.
You have startlingly selective memory, as that reflects exactly the stance Futuremark communicated prior to this. You responded to it the same way then as well.
Hmm....there comes a time for some people when you can only consider them a flat out liar. For me, as far as you're concerned, this is that time.
The statement you say is "new":
AJ said:
Let's go back in time a bit to 17 May 2003, 12 minutes after the hour:
[url=http://www.beyond3d.com/forum/viewtopic.php?p=117441&#117441 said:
One would presume that this, on 17 May 20003, 40 minutes after the same hour, signifies that you actually read what was stated:
[url=http://www.beyond3d.com/forum/viewtopic.php?p=117454&#117454 said:
Or maybe I give you too much credit.
What it looks like to me is that you are using the pretext of "new" here to pretend that your repetition of viewpoint is validated, and that there is significance in GRilla's statement beyond repetition because you do so. I say that, because it is in reply to that assertion that you decided to disagree, and then use a clearly and completely false statement to support that disagreement.
Well, what do you know, you're claiming your viewpoint is validated by repetition.
Yes, I do view it as enlightening for you to so clearly illustrate what you view as support for your viewpoint: Falsities and Innuendo.
You sure showed me.
Evildeus said:
As I said, this removes one of the effects that can be seen in the screenshots. the difference isn't great (a few % in GT1 and nothing in the other tests for a GF4, 1-2% all the tests for a faster board such and NV30) but this makes the 'bug' argument harder to believe - why would this only occur when the loading screen are there?
Doomtrooper
Veteran
Dave there is other ways also, does the Nvidia 42.63 drivers exhibit the camera rail, splash screen anomoly.
Randell said:
It's amazing, and a little scary, to think that there are still folks missing the fundamental issue. Yes, 3D Mark is a benchmark and not a 3D game, and it follows a fixed camera path when run normally.
However, the Detonators are not rendering 3D Mark in the same way that they render a real 3D game, while the Catalysts are rendering 3D Mark in the same way they'd render an actual 3D game.
This means that while running the benchmark the Detonators are requiring much less work from the nv3x gpus than the Catalysts are requiring from the R3xx vpus, because while the Catalysts are rendering off-screen elements at the same time as on-screen elements, just as they would when running an actual 3D game, the Detonators are rendering only the on-screen elements congruent with the camera track. The Detonators do not render 3D games the same way they are rendering 3D Mark--if they did, you couldn't play any of those 3D games because you'd get the same kind of mess that happens with 3D Mark when the camera is moved off track...
(Aside from the fact that there aren't any fixed-track 3D games out there that I'm aware of.)So, the nVidia Dets do not actually increase rendering performance at all, but rather merely run the benchmark faster by coding the Detonators to ignore the rendering of any elements not congruent with the camera track. Again, such camera tracks do not exist in 3D games, and so the Detonators must render all of the elements, both on-screen and off. This means that performance in real 3D games will not parallel, even loosely, performance in 3D Mark--it will be a good deal slower in real 3D games which exercise the same parts of the gpu that the 3D Mark benchmark does. In addition, it also means you cannot compare 3D Mark scores between the R3xx and nv3x, because whereas the Catalysts are rendering the scenes in their entirety, the Detonators are not.
As far as FutureMark's comments go--simply by verifying the results of ET and B3d independently--they've said all they need to say on the issue, IMO. The issue itself is self-evident, there need be no "confirmation" of it apart from the confirmation that has already occurred--which is that everybody agrees that when the camera is pulled off track in 3D Mark the Dets stop rendering, and the Cats render normally. Such evidence speaks for itself with an unmistakable voice. Again, IMO, the definitive answer as to whether nVidia's "driver bug" answer can be believed is already apparent. The proof is in the pudding, in other words...
WaltC said:
it's amazing and a little scary that people dont read and understand what I wrote and in what context.
I'm talking about 3DMark2001SE! Do not try and reply with another too long a paragraph to bother reading post as I wont read it as I agree what happens in 3Dmark03 is cheating
OK!?
