ExtremeTech - Leaked nVidia Driver Doubles 3DMark03 Score
- Thread starter Hanners
- Start date
Heh....You beat me to it by about ten seconds...
Here's the link I had ready:
http://www.extremetech.com/article2/0,3973,1188167,00.asp
Obviously, since apparently nVidia is now encrypting its D3d drivers to avoid UW's anti-detection scripts, it's incumbent on FutureMark to quickly release another patch as they did with 330 so as to study nVidia's newest cheating tactics. IMO, you don't encrypt something you have no interest in hiding from public view...
I think this will be the litmus test we've all been waiting for from FutureMark so that we can determine where the loyalties of the company lie: to IHV's or to making honest benchmark software. I think everyone agrees with me that the picture is no longer clearly cut at all. If FM does not release such a patch (I think they will have to do so with every nVidia Detonator release until nVidia desists), then I think it will be obvious the company has decided that placating nVidia is more important than its software--and I guess that will be that. I am not privy to what's been going on behind the scenes at FM with respect to these issues, but I have a gut feeling we will see no more anti-cheat patches out of the FM company. I certainly hope I am wrong. I would be pleased to be wrong.
Here's the link I had ready:http://www.extremetech.com/article2/0,3973,1188167,00.asp
Obviously, since apparently nVidia is now encrypting its D3d drivers to avoid UW's anti-detection scripts, it's incumbent on FutureMark to quickly release another patch as they did with 330 so as to study nVidia's newest cheating tactics. IMO, you don't encrypt something you have no interest in hiding from public view...
I think this will be the litmus test we've all been waiting for from FutureMark so that we can determine where the loyalties of the company lie: to IHV's or to making honest benchmark software. I think everyone agrees with me that the picture is no longer clearly cut at all. If FM does not release such a patch (I think they will have to do so with every nVidia Detonator release until nVidia desists), then I think it will be obvious the company has decided that placating nVidia is more important than its software--and I guess that will be that. I am not privy to what's been going on behind the scenes at FM with respect to these issues, but I have a gut feeling we will see no more anti-cheat patches out of the FM company. I certainly hope I am wrong. I would be pleased to be wrong.
Freak'n Big Panda
Regular
I wonder why they would compare with 44.03... I mean everybody knows most if not all of the 'optimizations' were stopped by the 330 patch so of coarse your going to see a large performance increase because NV has detected build 330. I think it would have been more useful to compare to 44.65, I doubt they woul'd have seen any diffrences though. 44.67 seems to just add anti-antidetect.
Well, i don't know how they did it, but Nv's driver team is good, +98% in GT4 8)
http://www.extremetech.com/article2/0,3973,1188123,00.asp
Evildeus said:
And 0% performance improvements in any of the "real games" ET tested with the new drivers....
So much for nVidia's "real 3D games" comments, right?Heh...right now nobody knows what new tricks nVidia is up to in 3DMark, and if nVidia has its way, nobody ever will...
It's going to be very interesting indeed to see if FM does another anti-cheat patch...
WaltC said:
If they don't, they're going to lose a lot of what respect they have left. If an IHV releases a driver with a performance boost that significant in their benchmark and doesn't show any related performance boost in actual games, at this point the question of cheating vs. legit code development must be addressed. ESPECIALLY if Nvidia's going to be coming back into the FM beta program.
Well, not much game use PS2.0WaltC said:And 0% performance improvements in any of the "real games" ET tested with the new drivers....
Sure, that depends also on B3D/ET integrity also, and the real presence of cheats.
Needs the presence of the cheat, on first hand, and the possibility to cope with if there's one.
3rd choice: Don't allow any NV benches to be uploaded & published on the ORB.
FM has the authority to stop any & all published results of NV cards using 3DMarks. Since NV says the benchmark is not representational of 'real world' games & not to use the BM > FM should (IMO) disable the ability for any NV cards to upload & kill its usage (for NV cards) in reviews too.
If all the BM makers did the same (not allow results of their proggy to be published in NV reviews) NV would be out in the cold & would have to come up w/their own benchmark. Of course that would defeat apples-to-apples comparisons & only give consumers ATi, Matrox, Trident, SiS, etc. cards to compare evenly. 8)
FM shouldn't waste their coders time to defeat NV 'optimizations', they should just disable all uploads until the voice of consumers puts pressure on NV to clean up their act. If a known thief comes to your house: do you let him in after you nail everything down so he can't steal it or do you just not allow him into your house?
Close the door & lock it FM!
.02,
FM has the authority to stop any & all published results of NV cards using 3DMarks. Since NV says the benchmark is not representational of 'real world' games & not to use the BM > FM should (IMO) disable the ability for any NV cards to upload & kill its usage (for NV cards) in reviews too.
If all the BM makers did the same (not allow results of their proggy to be published in NV reviews) NV would be out in the cold & would have to come up w/their own benchmark. Of course that would defeat apples-to-apples comparisons & only give consumers ATi, Matrox, Trident, SiS, etc. cards to compare evenly. 8)
FM shouldn't waste their coders time to defeat NV 'optimizations', they should just disable all uploads until the voice of consumers puts pressure on NV to clean up their act. If a known thief comes to your house: do you let him in after you nail everything down so he can't steal it or do you just not allow him into your house?
Close the door & lock it FM!
.02,
gokickrocks
Regular
Ostsol said:
i have...but i used a fx5200
i didnt use the anti-detect script because, well, i forgot :?
Code:
3dmark03 330
---------------------------------
driver - fps (vs, ps2, ragtroll)
---------------------------------
44.03 - 3.2 , 4.3 , 3.4
44.67 - 4.4 , 7.0 , 3.9
+37.5% , +62.7% , +14.7% * 44.03 - 44.67 *
44.71 - 4.4 , 7.0 , 3.9
D3D Rightmark 1.0.1.6
---------------------------------
driver - fps (geometry proc. spd. [vs2,ps2], pixel shading [ps2], point sprites [vs2])
---------------------------------
44.03 - 39.35 , 19.79 , 118.46
44.67 - 39.34 , 21.46 , 116.04
-0.1% , +8.4% , -2.1% * 44.03 - 44.67 *
44.71 - 39.05 , 21.07 , 116.99
-0.8% , +6.5% , -1.3% * 44.03 - 44.71 *
-0.8% , -1.9% , +0.8% * 44.67 - 44.71 *some of my math may be wrong, cause i dont remember if i rounded or not when doing the calculations
Evildeus said:Well, not much game use PS2.0
Of course...and how convenient....
Seriously, nVidia's all over the map on this one--"real games", indeed....I have to say that with 98% performance increases and driver encryption in the mix, the probabilities of cheating present are well beyond 50-50....
The reasoning here may be somewhat inductive, but I think it has merit.I'm not sure that FM need do anything except what they did for 330, which is to slightly alter the recompile for tiny differences in the memory configuration the program uses--just enough to throw off any driver recognition code keyed to somewhat more technical identifiers than the name of the executable. They wouldn't even need to see a cheat to be able to do this effectively, IMO. Of course, I also think it wouldn't hurt for 3D Mk to randomly change the name of its executable in each patch release--say to 3DMarkA.exe, and then with a succeeding patch 3DMkA32.exe, 3DMkA04.exe, etc.
An advantage for honest IHVs in something like this is that drivers that don't cheat would be as apparent as those that do. I'd like to see FM issue such a patch with each new Detonator release. Not very much work for FM--but gobs of frustration to an IHV's time invested in manipulating the benchmark. Main thing about this approach is it would defeat nVidia's encryption scheme handily--whether nVidia cheats or not--which IMO would be a good thing.
How could Futuremark ever justify issuing more patches when Nvidia is paying them to be in the beta program and have access to all the code?
I think Futuremark is in an impossible position now, and they will not issue more patches unless forced to.
I think the ball is now in the court of the 3rd party here, namely ATI. Nvidia has responded to Futuremark's move and made their decision to cheat and be secretive, and it remains to be seen what ATI intends to do about it. For myself, I hope they choose to place maximum pressure on Futuremark to issue anti-cheat patches, and publically slam Nvidia's cheats in their advertising, while keeping to the (relative) highroad in their own drivers. They've made initial moves in this direction, I hope it continues.
rms
I think Futuremark is in an impossible position now, and they will not issue more patches unless forced to.
I think the ball is now in the court of the 3rd party here, namely ATI. Nvidia has responded to Futuremark's move and made their decision to cheat and be secretive, and it remains to be seen what ATI intends to do about it. For myself, I hope they choose to place maximum pressure on Futuremark to issue anti-cheat patches, and publically slam Nvidia's cheats in their advertising, while keeping to the (relative) highroad in their own drivers. They've made initial moves in this direction, I hope it continues.
rms
rms said:
Well, remember that it's only a rumor that nVidia intends to rejoin the FM program at some future date...and also that nVidia was a paying member of the FM beta program for around ~14-15 months of the development cycle of 3DMk03 and had "access to the code" all that time (which presumably is what has allowed them to manipulate it as they have.) IMO, FM could still issue such patches routinely without specifically targeting nVidia because the patches are aimed at everyone, not just nVidia. And of course if nVidia does rejoin the program hopefully they will have pledged to behave themselves and so should not be concerned with such patches.
As to whether FM's position is now "impossible" I think that depends on how much of their autonomy they have sold, or bargained away, to nVidia (if any.)
StealthHawk
Regular
Ok, why should Futuremark release a new patch just for one unofficial driver? NVIDIA already leaked 44.65 which restored their 3dmark03 scores, what's new with 44.67? I've heard some speculation that 44.67 will be the next official release, but for now it is "just another leaked driver," correct?
This might be a stupid question, but has anyone tried these 44.65 or 67 drivers with Build 320 of 3DMark 2003?
I'm just wondering if using that version would:
a. Show up all the old cheats in there if they are still present.
b. Actually do the same job as releasing a new patch, and disable all the new patches as a result of the slight changes made between the two revisions.
I'm sure nVidia aren't that stupid, but... Well, you never know!
I'm just wondering if using that version would:
a. Show up all the old cheats in there if they are still present.
b. Actually do the same job as releasing a new patch, and disable all the new patches as a result of the slight changes made between the two revisions.
I'm sure nVidia aren't that stupid, but... Well, you never know!
