Brian Burke is a SNOTRAG.

Rel. 50 is the best driver we've ever built - it includes significant optimizations for the highly-programmable GeForce FX architecture...

Interpretation: Lots more cheats coming your way.

...and includes feature and performance benefits for over 100 million NVIDIA GPU customers.

Interpretation: "Screenshot fraud... It's not a bug, it's a FEATURE."

Pending detailed information from Valve, we are only aware one bug with Rel. 50 and the version of Half Life 2 that we currently have - this is the fog issue that Gabe referred to in his presentation. It is not a cheat or an over optimization.

Interpretation: ...and? And that's his entire case, a denial.

Since we know that obtaining the best pixel shader performance from the GeForce FX GPUs currently requires some specialized work, our developer technology team works very closely with game developers.

Interpretation: We dissassemble every major game and benchmark so we can cheat as much as possible.

Part of this is understanding that in many cases promoting PS 1.4 (DirectX 8) to PS 2.0 (DirectX 9) provides no image quality benefit.

Interpretation: We're know we suck at DX9, so we're going to commit outright fraud by telling you that this is DX9 card on the box, but we're only going to enable DX8 in the drivers

Our goal is to provide our consumers the best experience possible, and that means games must both look and run great.

Interpretation: ...even if that means running at lower image quality, committing fraud when a screenshot is detected, disabling entire image aspects of games, and of course, running at DX8 (DX7 if we can get away with it.)

In addition to the developer efforts, our driver team has developed a next-generation automatic shader optimizer

Interpretation: Our guys are tired of trying to find every last way to cheat on benchmarks by hand, so we've built an automatic cheat bot.

The fruits of these efforts will be seen in our Rel.50 driver release

Interpretation: You think we cheat? You ain't seen nuthin yet.

remember all those image quality tests? remember nvidia came out on top and faster with AA/AF? WHO HAS THE BEST IMAGE QUALITY?
nvidia were leading the world in 3d graphics but 6 months ago - wot the hell happened? nvidia are going to destroy themselves at this rate - im with 5150 "Oh Gawd! nVidia, you're breaking my heart! Keep up with the pace and shut the hell up."
AND ON THE SUBJECT OF 3DFX
now - i may have missed something somewhere but i seem to remember that EVERY SINGLE ONE OF US owe our latest gen 3d cards TO 3DFX - 3dfx invented the damn 3d accelerator! between 3dfx and nvidia we should be seeing cards that can wipe the floor with anything ATi can show - so why the hell arnt we?

Steg

this is a new fucking low....even for $nivida

http://www.anandtech.com/news/shownews.html?i=20525

this is exaclty what i dont want happening to the industry and why shit like this will kill it....sing a deal with nvida and get your as rapped by lawers

summary "nvida says, hmm we run better as dx8, but look like shit.....hmmm ati runs better at dx9 and looks better to....call edios and threatne thme to pull the demo patch....we cant have this negativre publicity"

maxx

Ok, Gerbil, what are you saying? It seems like your posts pretty much oscillate evenly between bashing nVidia and bashing ATI. How could multiple personalities have enough of the same memory to use the same login and password, are is it all a magical effect of the little password wizard?

Well, to correct myself in the light of a new day, the NV3x doesn't process fp16 instructions faster is more accurate. It can of course be said to "run" faster if it can avoid the stumbling a bit longer when using it. The issue to remember is that it still stumbles remarkably soon with fp16.

The challenge nVidia faces with their claims about Det 50 is picking a new (longer) path without the things it finds as obstacles to stumble over. But the more the shader is trying to accomplish, the longer and more convoluted the detours required to reach the goal, which still leads to large percentage differences in performance comparison to cards that take all of these obstacles in stride. Though outside of the specialized nVidia paths in the likes of Doom 3, HL 2, and TR: AOD, this should show some benefit.

Oh, let me see if this provides a clickable link without the parentheses: http://www.3dcenter.de/artikel/cinefx/index_e.php

The NV3x doesn't run fp16 faster. As far as established overall impact, the benefit in fp16 is that allows the NV3x to use more temporary registers (coding "variables") before choking severely. If you use fp32 without exceeding the (very very restrictive) limitations, it would be as fast as using fp16. However, the limitations on the registers is so restrictive, that it begins to progressively choke on shaders that exceed about DX ps1.3 complexity (ps1.4 is more functionally divergent from 1.3 than the minor increment in version number indicates) in any case.

Using values at fp16 allows slightly more complex shaders before the nv3x's specific and significant problems come into dramatic effect. However, the performance deficit to the R3xx exists even outside of that. For example, if you've read this 3dcenter article (http://www.3dcenter.de/artikel/cinefx/index_e.php) on the NV3x architecture, realize that its main body of discussion of the NV3x problems ignores both this very significant register limitation issue and the advantages the R3xx brings to floating point processing that allows it to double processing instruction throughput (besides the independent texture instruction processing discussed in the article) in some instances.

For something like the Doom 3 "full feature" shader and its very specific environment and main feature (lighting/shadow model) used to efficiently implement an advanced non-shader conceived effect (it was conceived for DX 7 cards and that era's bump mapping), that article shows what the competing architectures bring to the table. Stretching variety or complexity beyond that at all (which, btw, Doom 3 does in the ARB2 path), things tend to have opportunities to improve for the R3xx, and very drastically worsen for the current NV3x cards, based on these register and double throughput factors.

And there are other issues nVidia is having problems with, like floating point render targets, that are also important (the implementation for the HDR HL2 video you might have seen doesn't work on nVidia hardware at the moment for this reason). In this case, there is some functionality in the hardware to overcome this, but nVidia seems to have decided not to due to speed issues. Not knowing what the exact problem is, I actually think this is one thing that might conceivably be fixed by the NV38, but that is the epitome of unfounded speculation.

If this is news to you: Discussions on these issues stretch back to the end of last year for the early indications, and the very beginning of this year for very specific discoveries related to this, including comments by John Carmack in his .plan files. The HL2 discussions just brought it to everyone's attention dramatically and definitevly, because the "big" sites were silent in the interim.

The state of 3d benchmarking and independent sources of consumer information were sacrificed for this silence. :( I think that situation is what prompted Newell to speak out.

oh man, that's not someone's name.......

http://www.tweaktown.com/document.php?dType=article&dId=517&dPage=1

Truth is out Nvidia. You can't lie anymore........

"NVIDIA's Det50's don't quite grab the high-needed performance boost we had hoped for - granted they still have some time to improve upon them. However, we can now conclude as to why NVIDIA wanted Valve to use these in the recent testing we did in Half-Life 2 - they make the image quality look like crap as a way to boost performance.

To us, the choice for which graphics card is quite clear. Do you want to buy a video card that uses DX9 “Smoke and mirrors” to emulate performance, or one that provides best-of-class performance and image quality in one package. You’d truly have to be a NVIDIA fan-boy to buy one of their cards given the current state of affairs.

We’re told that both ATI and NVIDIA have new products coming out relatively soon; however, both are loosely based off of current generation GPUs, which means – all things holding true to form – that NVIDIA will still be taking in the shorts for the majority of DX9 / “Next-gen” games.

We do understand that our build of the Det50's are in beta, and there are some things that could - and obviously need to change - for NVIDIA to keep a running in the race. For starters, they need to crank up the image quality again and make sure, first and foremost, things look good. Secondly, find a way to optimize their shaders so games run much better. It's unfortunate that NVIDIA's way of optimizing games is to drop the precision down to 12 or 16-bit.

It's in our opinion that both end-users and OEM's out there should be highly cautioned into buying any of NVIDIA's current hardware under the assumption that it'll offer the best gaming experience for DirectX9-class games. Until NV40 rolls out, it appears for now that NVIDIA will be sitting on the side-lines. "

My GeForce4400 wallops ati's best.
word.

I will tell you all what's wrong with nvidia,,,3dfx they destroyed themselves with their stupid ideas of creating their own chips and parrallel proccesors and now their wacko engineers have infultrated nvidia,, the 5900's are a joke they remind me of the voodoo 6000's Nvidia wake up fire those idiot 3dfx dolts before it's to late... Last best nvidia card ti4200

Welcome to a fun new world kiddies. GPUs are now very VERY much like CPUs. If your gpu isn't targeted specifically for optimization, it'll suck. Very much like a fun running debate for a long time amongst PC users, AMD vs Intel. If you base your code path on 3DNow!, AMD looks good, base it on SSE2, Intel looks good. In this case it's probalbly more accurate to describe this as follows...

MS for god knows why decided to come up with a new programming language and compiler. ATi signed on knowing the restrictions/consessions they'd have to make (patents/etc being handed over to MS, etc) and got to see from inception what this ISA was going to look like. They then turned around and built a CPU that was hand designed to work well with this new language and compiler. NVidia on the other hand, didn't feel like bowing to MS strongarming, and said up yours. They got to design in the dark while guessing the most likely design of the new language. In the end, ATi built a P4, NVidia built a G5. While the G5 can have the same code compiled for it, it's not a good fit and does poorly, the P4 on the otherhand was the natural target for the lang/compiler and flies. Many will argue the G5 has more horsepower, but we'll never see it till a new spec is introduced that targets the G5's capibilities better.

Ok, as always, analogies suck, and that one did too. But hopefully you get the idea. As GPUs move away from fixed function units into dynamic, programable, cpus you're going to see more and more of this. For one round, one mfg's GPU is going to be the rocket as its the closest match to the API. Next time around, it could switch.

The other fun aspect is going to be the ubiquitous IQ debate. As of right now there is no specification saying I stuff x,y,z into a DX 9 compliant renderer, a,b,c should come out. Untill there is a refrence rederer that covers all corner cases, that can be used as a basis, you will see lots of variations. This is what'll really piss people off... none of those variations are inherently WRONG either.

This round, Nvidia has some potentially powerful hardware, you won't see it perform with DX 9. So sad, move on. Who gives a flying fsck about the pr manuvering/bs/etc the corps play. When you get HL2 you'll either buy hardware to make it play better or not. Make up your own mind based on your own research, and you'll be fine.

I see so much ATI FUD flowing through this board that it makes me want to puke.

Instead of all this speculating, how about everybody judges by the products they see when they're released, instead of what they think is going to happen?

In other words, Shut the fu(k up, you retards.

This is indeed funny. For a year now we've seen raves on ATI's hardware, complete whines on Nvidia tactics and drivers, and both play current games fine, if not just a tad overpowered. Now the game of the year isn't even out yet, and we have people accusing manufacturers of this and that.

Hilarious. I think we'll look back at '02-'03 as the teen angst years of 3D-hardware.

You can seriously just wait the first few months on HL2's release and make your decision then. Droopy, you need to calm down a tad. NONE of this is life or death, but a stage of a very strange play.

I for one am quite surprised at all the benches. All the cards seem to play a little underpowered at those higher resolutions. HL2 I thought was going to be more friendly towards hardware.

So both NVIDIA's hardware and drivers suck now?

...

Hehehe, not for me they don't! Hurray for "denial" !

This is not happening, this is not happening...

http://techreport.com/reviews/2002q3/ti4600s/index.x?pg=7

In terms of performance, stability, and compatibility, you can't go wrong with nearly any flavor of GeForce4 Ti 4600 (...) you'll have the absolute fastest 3D graphics chip anywhere

Yayy!! Take that ATI !

Now, im sure most of you have read Gabes recent comments regarding the detonator 51.75s, and Nvidia's offical response but I really do have to say, having seen this first hand it confirms to both myself and Veridian that the new detonators are far from a high quality IQ set. Alot of negative publicity is currently surrounding Nvidia, and here at driverheaven we like to remain as impartial and open minded as we possibly can, but after reading all the articles recently such as here coming from good sources and experiencing this ourselves first hand, I can no longer recommend an nvidia card to anyone. Ill be speaking with nvidia about this over the coming days and if I can make anything public I will.

http://www.driverheaven.net/articles/aquamark3/index3.htm

How many more sites will be saying things like this. And you wonder why Valve did not want to use those drivers........

Look at all these posts.. arguing/posting like your life depended on it.. NV3x is a mess, no doubt. Even nvidia guys don't know how that thing works. Still, people should lay off this "guilty until proven innocent" mentality. I mean most of people here haven't got a clue about how NV3x works and why it sucks, yet it seems some have decided that NV3x performance can't be improved at all with det50s or that det50s is just a bunch of cheats "optimized" for benchmarks. Sure, det50s probably won't suddenly make NV3x outperform R3xx, but it will certainly help NV3x. Besides, nvidia knows that it will be much scrutinized, so they would be careful not to go overboard with "optimizations" if they are not plain dumb.

170+ comments and only three comprehensible pages needed to display...threaded layout = 73h r0x0rz

What's really scary is the obvious fact that the 9800pro is $100-$150 cheaper then the 5900 Ultra. Now THAT is a major issue for Nvidia with these turn of events.....

Question: Does this mean nVidia is *partly* at fault for the delays in HL2 and D3?

Quote from Anand's article:

The most obvious is through updated drivers; NVIDIA does have a new driver release on the horizon, the Detonator 50 series of drivers, however Valve instructed us not to use these drivers as they do not render fog in Half-Life 2.

I hope that's just because they're in beta.

So here is the big question. Next round if Nvidia turns around and releases a card a lot faster (without cheating in any way of course) then ATI's would you guys buy it?

Why don't all you twits realize nVidia isn't bullshitting.

DirextX 9 doesn't state how a given effect should be implemented, only that it is. nVidia's GFFX cards run certain shader effects in a much higher quality than ATI's, which causes them to be slower. ATI, on the other hand, uses their lower bitrates to immprove performance while using tricks to keep the quality up. Next Gen, unreleased, cards will be different, but this is the case with current top offerings.

Of course, I'm sure all the majority of the poster's here have are knowing that PS2.0 is a bigger number than PS1.4 and wouldn't know how to do the most basic of operations like calculating the reflection of a vector against a surface.

Or just shove your head in your asses and keep talking moronic shit.

Anyone see the coincedence?

The NV3* series of cards, were the first that the 3Dfx Engineers worked on . . .

AG #141 it a bigger issue than you think. Most LAN parties dont have a internet connection, so that pretty much kills HL2 at LAN parties. This will also ring true for CS 1.6. There will not be a standalone version and you can only get it via STEAM with a valid internet connection, even if you are only playing on a LAN. Oops kill another one for LAN parties. This is just one example of the flawed STEAM design.

http://www.anandtech.com/video/showdoc.html?i=1863

#130 Did you even read the mod article. It said 9800 non pro, NON SE.

OK its after twelve so I guess here is the anand performance review

http://www.anandtech.com/video/showdoc.html?i=1863

More benchies:
http://www.anandtech.com/video/showdoc.html?i=1863&p=1

Edit: But dammit why didn't they bench with the Det50s...

Edit: ME FIRST :-D

Edit: Ok there are a few typos there already from a cursory reading... I'm not sure what to make of it in that light.

nVidia has been circulating its Det50 driver to analysts in hopes that we would use it for our Half-Life 2 benchmarking. The driver contains application-specific optimizations that will likely improve nVidia's overall performance picture, however Valve's Gabe Newell expressed concerns that some of nVidia's optimizations may go too far. Doug Lombardi of Valve has explicitly asked that beta versions of the Det50 drivers not be used for benchmarking.

Why do i get this feeling we are looking at Nvidia style 3dmarks all over again..........