ExtremeTech Article Up

for SSSE

ogl_iFinish:
-
0 = do not force API to finish rendering (performance)
1 = finish all rendering before begining of next frame (optimal)
2 = force finish of rendering before swapping of buffers (safe)
3 = force finish of rendering before matrix change
and after swapping of buffers (paranoid)
-
for SSFE

gap_iFinish:
-
0 = do not force API to finish rendering (performance)
1 = finish all rendering before begining of next frame (optimal)
2 = force finish of rendering before swapping of buffers (safe)
3 = force finish of rendering before matrix change
and after swapping of buffers (paranoid)

 

Thanks tEd

 

Someone posted this at the Nvnews forums:

It was posted by vacs here: http://www.nvnews.net/vbulletin/showthread.php?s=&postid=123105#post123105

Any truth to this? Worm?

***off topic****
Doom, check your pm's.

 

I suspect an overzealous geometry/soft cull driver optimization. Dare I say it - akin to the infamous predictive HSR algorithms. The inclusion of static clip planes may throw off the "intelligence" of the system.

 

... ... I was actually roughly correct about what that flag does?

I'm afraid I'm not quite understanding you... with 3dMark it's very clearly not an optimization as much as a hack to remove what wouldn't normally be rendered due to the locked nature of the camera path, though, I don't think you're saying that's not the case... I'm just not quite understanding the context of the first sentence.

 

I think some other NVIDIA owners need to realize that we were hosed with 3dmark03 expectations. First with image cheats, then this. Come on, our scores almost doubled from what they were getting when things looked right.

I don't care anymore though, I just want my card to be better than my Ti4200. Which for the most part it is, except I still don't think AF is fixed in some games (like UT 2K3) even though Anand and Tom say it is.

Anyone else with an FX and Geforce 4 try 44.03 and check if the FX 8AF still looks washed out (smeared). It sure looks like this compared to when I used my old Geforce 4 card with the same drivers using 8AF.

Thx

Larry

 

Switching it to 1 fixed the problem.

 

personally I believe that we should get quality rendering all the time not just some of the time...........

 

Some people over here can easily confirme that.

 

I don't who vacs is, where is is coming from or from where/whom he has supposedly "heard" this kind of information, but what he posted is false and has no truth in it.

*edit: In cases like this (someone claims something about us or our products) it's best to simply email me or AJ and ask what's going on.

 

Pretty definitive answer. Thanks, worm

 

Can you say what is the actual official position of FM right now? Cos i didn't see any. Thx.

 

I can't really comment on that as we are still looking into things, and gathering more info on the subject.

 

Ok, thx you, so no definitive conclusion at that time. Hope to see some conclusive information on this issue soon. Do you think we will be able to have that before the end of the month?

 

Thanks for the answer.

Follow up, does the change seem to have impacted the speed? Seeing as how the description above seems to indicate that the 0 setting is for performance and the 1 setting is "optimal," it would be very interesting to see how that actually affects the speed.

 

Hi guys,

I did post following to our forums, but didn't post it here yet (I saw some people linking to it, but the word didn't get around I guess).

-----------------------------
Everyone,

Last night ExtremeTech, a member of our Beta Program wrote an article called "Driver Irregularities May Inflate nVidia Benchmarks" (http://www.extremetech.com/article2/0,3973,1086025,00.asp)

The findings in the article have also been confirmed by Beyond3D,which is also one of our Beta Members.
(http://www.beyond3d.com/forum/viewtopic.php?topic=5856&forum=9)

We've been notified last week about the same issues and as mentioned in the articles we have also tested these issues at our own labs. Based on our initial research the issues pointed out in the articles are real and should not exist as part of the 3DMark benchmark run.

We've also notified Nvidia about these issues a couple of days ago and are still waiting for a technical response. Please note that all Beta Members have an access to 3DMark version with freecam enabled. We'd hope that Nvidia would participate in our beta program so that we could work on this kind of issues together with them.

Right now we are gathering more information about the subject and may comment to this later.

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

That sums up the situation as things are for now. We've also requested technical explanation from Nvidia, but so far have only received the generic "it is most likely a driver bug" answer. We're still waiting for a more detailed technical explanation, until we decide how to proceed with the issue.

Cheers,

AJ

 

Vacs=Derek Perez

 

AJ,

Thx.

If FM hasn't come to a conclusion, that Nvidia has been informed a couple of days ago, and that the investigations are still under way, how come this issue has been made public? Why this special treatment?

Thx for the answer anyway!

 

If nVidia had been informed of the driver issue, how come they allowed 3DMark scores to be published by web sites using those drivers?

You might have a basis for a valid "concern" if no 3DMark benchmarks had been published using these drivers. The fact that they have been published (and also that the drivers are WHQL certified and available to the public), is valid reason for publicizing their behaviour with the benchmark.

Keep in mind that FM has not officially claimed nVidia is cheating or not. ExtremeTech used FM's tools (with their permission) to build a case for it.

Come on ED, I thought you'd be more interested in having the public aware of what's going on with published results and available drivers, rather than keep them in the dark.

 

When did the reviews have been made? And when did the information have been given to Nvidia?
Reviews: Up monday morning
Information to Nvidia: a couple of days ago means monday at best.

Don't you say, innocent proven guilty? It seems that some people goes otherway, guilty till proven innocent. BTW, you did see that FM has said they don't know if it's cheating or not.

Yes i have that in mind. But i have also in mind that the information we have shows us that Nvidia didn't have much time to respond, Futuremark did use some exceptionnal means and that the article comes out 3 days after the launch of a new product from NV. Futuremark still don't know, but FM accept the publication of the news. Odd at best i say.

Well, then perhaps should we have some information on each time there's some issues with drivers and 3DMark and no special cases.

I don't really want to rego in this, but yes there's some issues with NV drivers (it's most probabky cheating), no FM don't know if it's a bug or not, no NV/FM didn't have enough time to go further in the investigation, yes the article is well documented, yes the article has been writen before any conclusion from FM (in a rush?), yes it has been done with special authorisation from FM.