My thoughts on "Optimization"

[Behemoth]

Originally posted by Hanners
Exactly - You can often view the same effect by using the 'fly' command in a game like Half-Life, and moving to a part of the level the game developers didn't expect you to that's totally outside the clipping planes.
can you post a screenshot?

[Hellbinder]

Joe.. You are addressing Behemoth like he actually even understands what is being discussed in this thread. He is in no position to even discuss what a Clip-plane is, or half the other technical subjects that get brought up around here.

Arguing with this guy or a handfull of others is pointless. They just type up contrary responses as fast as they can. Regardless of what the content of the discussion actually is.

This is truely what seperates Dyed in the wool *Fanboys* like these, from people who have a hardware/software *Preference*.

[Hanners]

Originally posted by Behemoth
can you post a screenshot?

I'll see if I can dig one up tonight.

[Behemoth]

Joe.. Dont talk to Hellbinder like he actually even understands what is being discussed in this thread. He is in no position to even discuss what a Clip-plane is, or half the other technical subjects that get brought up around here.

Arguing with this guy or a handfull of others is pointless. They just type up contrary responses as fast as they can. Regardless of what the content of the discussion actually is.

This is truely what seperates Dyed in the wool *Fanboys* like these, from people who have a hardware/software *Preference*.

:D

[Hanners]

I'll give Behemoth some credit - He knows how to copy and paste. :p

[Joe DeFuria]

Originally posted by Behemoth
possibly bug.
possibly nvidia cheat.
possibly futuremark cheat. :D

No, I mean specifically what TYPE of bug/cheat can cause 3D Mark anomolies (like the screenshot I posted). You claim it's not a "clipping plane" bug/cheat.

So what kind of bug/cheat could cause that behavior?

[Behemoth]

Originally posted by Joe DeFuria
No, I mean specifically what TYPE of bug/cheat can cause 3D Mark anomolies (like the screenshot I posted). You claim it's not a "clipping plane" bug/cheat.

So what kind of bug/cheat could cause that behavior?
well i cant tell what was going on from just a few screenshots, it could be anything a few screenshots cant prove anything, but it does not look like clip plane problem to me, so what is your opinion? cheat?

[Sazar]

this image basically shows the edge of the T spawn map in 'free look' mode in CS...

I have not gone over the clipping boundaries yet and note I can fly anywhere within the bounds without image corruption...

[Sazar]

here I have gone OVER the clipping planes... and notice the image corruption that occurs...

[Behemoth]

Originally posted by Sazar
here I have gone OVER the clipping planes... and notice the image corruption that occurs...
can you mess up the whole screen please? :)

[Hanners]

Originally posted by Behemoth
can you mess up the whole screen please? :)

What will that prove that the other screenshot doesn't? :confused:

[Sazar]

Originally posted by Behemoth
can you mess up the whole screen please? :)

?

what do you mean ?

I am showing you what is happening half inside the plane and half outside... I would think that it is fairly obvious that if you do indeed continue to exit the planes set you will suffer image corruption throughout the image and not only half as I have shown...

you asked for a screenie... I provided it...

[Ady]

Originally posted by Behemoth
can you mess up the whole screen please? :)

http://www.testeagles.net/cs/clip.jpg

[Behemoth]

Originally posted by Ady
http://www.testeagles.net/cs/clip.jpg
ok i am wrong, everyone happy :D
but thanks for your time :D

[Sazar]

Originally posted by Behemoth
ok i am wrong, everyone happy :D
but thanks for your time :D

everyone knew that from the first post you had in this thread :)

we were just waiting for you to figure it out... you know... saving the lost soul sorta thing...

but I have a feeling you were just mucking round anyways...

[Behemoth]

Originally posted by Sazar
everyone knew that from the first post you had in this thread :)

we were just waiting for you to figure it out... you know... saving the lost soul sorta thing...

but I have a feeling you were just mucking round anyways...
yeah thats why i am so popular :)
honestly i never seen those clip plane problems b4....

[Joe DeFuria]

"The truth is out there!"

Hellbinder said:
Joe.. You are addressing Behemoth like he actually even understands what is being discussed in this thread. He is in no position to even discuss what a Clip-plane is, or half the other technical subjects that get brought up around here.

Arguing with this guy or a handfull of others is pointless. They just type up contrary responses as fast as they can. Regardless of what the content of the discussion actually is.

Actually Hellbinder, though I know what you're saying, (it can be exceedingly frustrating) addressing Behemoth like we have is the best way to actually show he's wrong.

If we just ignored him, he, and anyone else believing what he was saying, might still be in that position right now. Obviously, at this point, there can be NO DOUBT by anyone, that his line of argument is simply wrong.

And if anyone else tries to use it in this or any other forum...just link back here. :)

[Sazar]

Originally posted by Joe DeFuria
"The truth is out there!"

Hellbinder said:


Actually Hellbinder, though I know what you're saying, (it can be exceedingly frustrating) addressing Behemoth like we have is the best way to actually show he's wrong.

If we just ignored him, he, and anyone else believing what he was saying, might still be in that position right now. Obviously, at this point, there can be NO DOUBT by anyone, that his line of argument is simply wrong.

And if anyone else tries to use it in this or any other forum...just link back here. :)

yup... but i had to fire up cs and NOT play for the sake of this argument :(

thats just misuse of computing power...

[Joe DeFuria]

Originally posted by Behemoth
well i cant tell what was going on from just a few screenshots, it could be anything a few screenshots cant prove anything, but it does not look like clip plane problem to me, so what is your opinion? cheat?

yes, cheat.

For two reasons:

1) It is an issue which can speed up rendering, so there is clear motivation for doing something like this.

2) It does not happen when you stay "on the rail".

If there were clipping issues during a normal benchmark run, then I could see this being some sort of bug. That they are attempting some legtimate optimization. But the fact that everything is rosy when on the rail, and it all goes to hell just by going off it....I can honestly not think of how that can occur unless it was purposely done that way.

[legion88]

Originally posted by Nv40
this tread go in circles again and again...
re-read my previous post.. you can only cheat if you agree with
the rules of the contest ,that your cards be used in that benchamrk.
ATI agreed that their cards be benchmarked in GAMES like QUAKE3.
If they agreed with the rules are latter break them they are cheating.

BUT SINCE NVIDIA HAVE ALREADY STATED THAT 3dMARK2003 DOES NOT REPRESNT THE REAL PERFORMANCE OF THEIR CARDS ,THEY WARNS WEBSITES TO NOT USE THAT BENCHMARK ,TO NOT TRUST IN THE SCORES .. :nono:

WHo is MISleading the public then? ;)

Your comments (especially those in caps) are ludicrous.

Just because NVIDIA 'warned' people not to trust 3DMark2003 results, it does not follow at all that NVIDIA did not attempt to cheat--assuming all this discussion on clipping planes, including the technique used to discover the anomaly, is valid.

One can argue that NVIDIA is deliberately trying to invalidate 3DMark2003 results altogether by employing this 'optimization' realizing that people will find out that the results are invalid. With their results questioned, the results from other cards would also be questioned.

As you already know, the technique employed by NVIDIA in boosting their 3DMark2003 scores requires knowledge like camera angles and such. Correct? Such knowledge is not afforded to drivers programmers when 'optimizing' for games during real-time usage. This was already mentioned but ignored by you.

Thus, NVIDIA is utilizing "optimizations" that can only be used in static benchmarks like 3DMark2003 and Quake 3's pre-recorded timedemos and not in real-time usage that is critical for the enjoyment of the users.

It is an optimization but it is also a cheat. "Opti-cheat" is a term I like to use.

[zakelwe]

Originally posted by Joe DeFuria
yes, cheat.

For two reasons:

1) It is an issue which can speed up rendering, so there is clear motivation for doing something like this.



Yes but does it Joe ?

Show us some figures to prove it.

Maybe you would like to quote Extremetechs original figures they quoted in their article ?

Maybe Beyond3D's figures as well , as they are also members of the beta program.

Actually has anyone got any figures ?

No, only this thread actually ....


Go away and do some benchmarking yourself, turn yourself into Sir Isac Newton and away from being Jerry Springer.

Regards

Andy

[Gar]

Originally posted by zakelwe
Yes but does it Joe ?

Show us some figures to prove it.

Maybe you would like to quote Extremetechs original figures they quoted in their article ?

Maybe Beyond3D's figures as well , as they are also members of the beta program.

Actually has anyone got any figures ?

No, only this thread actually ....


Go away and do some benchmarking yourself, turn yourself into Sir Isac Newton and away from being Jerry Springer.

Regards



Andy

I don't see anywhere in his posts that states "It absolutely speeds up benchmarks by 10^50%". If you would properly read his post it says "can" speed up benchmarks. I think the majority of people that know a few things about video cards would say the same statements Joe made. To dismiss covert tactics by a large company/gov't for the purposes of increasing profit/support is a completely narrow minded way of looking at life, something I call blind favoritism/patriotism and in no way corrects shady practices made by these entities.

[zakelwe]

Originally posted by Gar
I don't see anywhere in his posts that states "It absolutely speeds up benchmarks by 10^50%". If you would properly read his post it says "can" speed up benchmarks. I think the majority of people that know a few things about video cards would say the same statements Joe made. To dismiss covert tactics by a large company/gov't for the purposes of increasing profit/support is a completely narrow minded way of looking at life, something I call blind favoritism/patriotism and in no way corrects shady practices made by these entities.

He says can but means DOES, or else there is nothing to argue against. Nvidia cheat and give no extra points, no really .. hardly what you are all thinking ?

It must be measurable for it to be effective so Joe or Extremetech can show the difference this makes.

Have they ?

Actually, you were so busy trying to avoid any benchmarks you should show the effective increase or be labelled the same way yourself.

1) Nvidia cheat to put the scores up so selling more cards
2) The scores are higher
3) You can measure this
4) You prove it.

Either do that or just keep avoiding giving a quantitive demonstration by warbling on.

Regards

Andy

[jjjayb]

Yes but does it Joe ?

Show us some figures to prove it.

Maybe you would like to quote Extremetechs original figures they quoted in their article ?

Maybe Beyond3D's figures as well , as they are also members of the beta program.

Actually has anyone got any figures ?

No, only this thread actually ....

And how are they supposed to show the figures without the cheat? Hack Nvidia's drivers to remove the cheat?
Of course it is going to make a difference in the score when you aren't rendering everything. How much of a difference only Nvidia knows.

[Joe DeFuria]

Originally posted by zakelwe
Yes but does it Joe ?

Show us some figures to prove it.

Well, Andy, it's called "common sense". Apologies if this type of thing seems like rocket science to you. I'll try and explain with a hypothetical:

Situation 1) Card A renders 100,000 pixels.
Situation 2) Card A renders 120,000 pixels, a superset of the pixels rendered in situation 1).

Common sense dictatates that situation 1 will be, if anything faster than situation 2, because less work is being done.

How much faster is obviously an open question, and one that I did not claim to know or address. (Nor does ExtremeTech, which is why they are waiting for drivers that don't have this issue to make any such claims.) It all depends on exactly how many extra pixels there are, and how "difficult" it is to render those extra pixels. The end result is, it might be a significant increase in time...it might be very slight.

We do know that the SKY in particular (which the clip planes seem to specifically address), is pixel shader intensive, and these types of pixels (ones with pixel shading) are among the most time consuming to produce.

Now, I realize that likely went over your head, and you probably don't understand the significance of the sky being pixel shaded. But hey, I can only lead a horse to water...

Maybe you would like to quote Extremetechs original figures they quoted in their article ?

Maybe you would actually like to re-read MY statement where I said it CAN increase the performance, not that it did, or by how much. The point is, ignoring pixels certainly won't DECREASE performance.

Go away and do some benchmarking yourself, turn yourself into Sir Isac Newton and away from being Jerry Springer.

Regards

Andy

Apparently, "Jerry Springer" seems to be about the level of comprehension and common sense you can handle. Go back to your trailer, get bombed, and start beating your wife's boyfriend, who happens to also be her brother... :rolleyes: