New article using Unwinder's anti-cheat script at EB

[digitalwanderer]

"EB" as in Elite Bastards, not the world's greatest online game store.

Our newest bastard Hanners has been playing around with his 9800 Pro & the Catalyst 3.5 set on a number of different benchmarks & tests. He posts his results & findings.

(Just a quick and shameless site pimp? Sort of....but it's topical too. )

 

[Hanners]

Thanks dig.

I'd just like to add that any comments, be they good, bad or indifferent, will be taken on board and much appreciated.

 

[Geo]

Nice job. Would I be correct in concluding from these results that the mere presence of the anti-detector script does NOT impact performance? I.e. they have to find something to detect before they go into action and cause a drop? I have seen elsewhere persons yattering on about the lack of a "control" to validate the results --wouldn't your non-affected results being statistically the same both with and without the script provide that "control"?

 

[nelg]

Good read Hanners. I, for one though, think it is safe to assume that ATI would not have any cheats in OpenGL games. Nvidia on the other hand most likley cheats in both.

 

[Quitch]

Considering I have an ATI card, I found the article very reassuring. Actually, I was impressed that they haven't any kind of specific optimisation for a game as big as UT2003, yet STILL outperform nVidia.

Impressive stuff.

 

[Hanners]

Nice job. Would I be correct in concluding from these results that the mere presence of the anti-detector script does NOT impact performance? I.e. they have to find something to detect before they go into action and cause a drop?

As I understand it, the changes to block the detection mechanism are small and don't actually require any 'extra' code per se, therefore it doesn't impact the performance or CPU load at all. I guess the similarities between my results with and without the AntiDetector script go some way to prove that.

 

[K.I.L.E.R]

Hanners, doesn't 3dmark have a 5% error ratio?

First score w/o script: 10391
2nd score w/ script: 9973

10391*.05
519.55
answer+9973
10492.55
-------------------
9973*.05
498.65
answer+9973
10471.65

These are acceptable max scores.
I'm trying to say, that with and without the script it is well within the error ratio.

Nature is the only score that proves there are cheats included for that one test.

 

[Tim]

These are acceptable max scores.
I'm trying to say, that with and without the script it is well within the error ratio..

Even if the total score is within the error ratio a 72% increase in performance in GT4 is way beyond the error ratio.

 

[K.I.L.E.R]

These are acceptable max scores.
I'm trying to say, that with and without the script it is well within the error ratio..

Even if the total score is within the error ratio a 72% increase in performance in GT4 is way beyond the error ratio.

Yes, you are right. I have pointed that out.

I should also point out that it was very sneaky of Ati to increase the Nature framerate and keep the onion score within the 5% error ratio.

I don't have 3dmark01 installed so if someone would kindly post the formula so I could MANUALLY check it out to see if there are any other manipulations going on.

 

[Hanners]

I should also point out that it was very sneaky of Ati to increase the Nature framerate and keep the onion score within the 5% error ratio.

I don't think it was sneakiness so much as the fact that altering scores on that test doesn't have that large an impact on the final score.

Here's the formula for calculating the final 3DMark 2001 score by the way.

3DMark score = (Game1LowDetail + Game2LowDetail + Game3LowDetail) * 10 + (Game1HighDetail + Game2HighDetail + Game3HighDetail + Game4) * 20

 

[K.I.L.E.R]

First manual no script test:

( 149+192.7+137.2 )*10+( 54.9+109.7+67.4+48.2 )*20=10393

Noted result: 10391

2 points less, where is my error?

----------------------------------

2nd manual script run test:

( 147.5+193.1+137.5 )*10+( 55+109.3+67.3+28 )*20=9973

Noted result: 9973

Thanks Hanners.

 

[Jima13]

I thought the error ratio was +/- 3%

 

[K.I.L.E.R]

I thought the error ratio was +/- 3%

I read it was 5%.

 

[Neeyik]

As far as I'm aware, there is no "official" statement on accepted variance in final 3DMark scores but around 3% is the one that Worm, myself and others will quote the most. The actual variance is really dependent on the score itself as a 20k system is unlikely to change by nearly a 1000 marks from test to test.