Which path will NV40 use in Doom3?

[jimmyjames123]

Well duh, I am talking about those people with poor performing cards that needed their own special path. I'm talking abou those people that have now been orphaned if the NV3x special path has now been removed from the Nvidia killer app.

Obviously you aren't very open minded here. Have you thought for once that, if the NV3x path is really removed, that maybe, just maybe the NV3x current gen cards run well on the ARB2 path? You are not being logical here. You are basing all your comments and jumping to conclusions based on a comment stated a very long time ago, using raw drivers, using the ill-fated NV30 card.

To talk about the NV40 being better is irrelevent - which is why I'm not talking about it.

I certainly am not talking about the NV40, so what's your point?

I guess your attitude matches that of the company you love.

That's funny, I think that your attitude matches the company that you hate!

First let's pretend the NV30 never existed, then let's ignore all the people who bought NV3x and stop bothering to support them with special paths. Then tell them they should forget about the money they spent on NV3x cards and tell them to buy NV40 instead.

Looks like you are pulling these idiotic statements out of thin air. Where are you pulling this from? You seem so blinded that you don't even realize that there were major performance boosts just in going from NV30 to NV35. And with Ultrashadow technology, Doom 3 looks even more suitable for the NV cards. Yet all you do is stay grounded in the past.

And you have the gall to call me a fan boy

You are actually one of the worst types of fanb*ys, one who takes pleasure in constantly digging at a company like NVDA.

 

[digitalwanderer]

Obviously you aren't very open minded here. Have you thought for once that, if the NV3x path is really removed, that maybe, just maybe the NV3x current gen cards run well on the ARB2 path? You are not being logical here. You are basing all your comments and jumping to conclusions based on a comment stated a very long time ago, using raw drivers, using the ill-fated NV30 card.

Uhm, no. The nV3x hardware just has some basic problems with the ARB2 path. If it is running it well now than either Carmack has changed the ARB2 path to better suit the nV3x's architecture or else nVidia is doing some form of shader replacement for it.

IMHO, I think, mebbe and all that.

BTW-I really find it hilarous the way you constantly refer to nVidia by it's stock symbol "NVDA", it's quite telling.

 

[jimmyjames123]

Uhm, no. The nV3x hardware just has some basic problems with the ARB2 path. If it is running it well now than either Carmack has changed the ARB2 path to better suit the nV3x's architecture or else nVidia is doing some form of shader replacement for it.

Can you show some evidence to back this statement up? Doom 3 is not a PS 2.0 heavy game anyway. You also forget that Ultrashadow technology did not seem to be implemented when Carmack initially tested the NV30.

BTW-I really find it hilarous the way you constantly refer to nVidia by it's stock symbol "NVDA", it's quite telling.

What exactly does that tell you? I don't own a single share of NVDA stock. The only tie I have to NVDA is a Ti4200! However, that is an abbreviation that is commonly understood to be NVIDIA. Wow, you are really reading too much into things!

 

[Pete]

(Let me save you some typing: most ppl by now understand nV to mean nVidia.)

 

[karlotta]

what does PS2.0 have to do with the ARB2 path.
Maybe with ogl 1.5(1.3?) having the GLSL and becoming 2.0(?) has alowed
nVidia and id to "optimize" the nv30 class to run good enough for D3 engine games on the ARB2. Tuff to sell a lic. with lots of work for a "mainstream" card.

 

[Hellbinder]

I just dont see how this is possible.

First of all are there any Synthetic or benchmark tests where Nvidia is not getting to force FP16 or lower that they are competative with R3xx cards? I personally have not seen them. They are replacing shaders or Converting it all to fP16 or lower in every single game relased case i can think of.

Secondly..

While *somehow* it may be possible with like a four leaf clover or something to get the Nv30 to run the ARB2 path at full speed... How can that apply for the rest of the Nv3x line? I dont think it can. Which tells me that the mid and low range cards must be running in GF4 mode.

This just smacks again why Carmack and Sweeney really irritate the hell out of me. How can they (in this case carmack) State that the nv30 is running the full percision path at full speed when he *KNOWS* they have to be changing things at the driver level to do the same kinds of changes he was forcing before with the game code? Its flatly Dishonest becuase he knows they are in essense *CHEATING* from a Marketing perspective that is Sure to come. Yet has nothing but good thigs to say, still claims Nvidias drivers are *better* even though i know PERSONALLY from about a year ago he had not submitted as SINGLE Driver issue in well over 6 months (at that time). I could go on but most of you already know how i feel.

 

[anaqer]

"You're the f4nb0y!" "No, you're the one!"
Stop with this bloody crap already.
It's neither constructive nor funny.
Or has this been said before? :?


( And just how ridiculous it is that even the term has been filtered out because of such childishness? )

 

[Sxotty]

This just smacks again why Carmack and Sweeney really irritate the hell out of me. How can they (in this case carmack) State that the nv30 is running the full percision path at full speed when he *KNOWS* they have to be changing things at the driver level to do the same kinds of changes he was forcing before with the game code? Its flatly Dishonest becuase he knows they are in essense *CHEATING* from a Marketing perspective that is Sure to come. Yet has nothing but good thigs to say, still claims Nvidias drivers are *better* even though i know PERSONALLY from about a year ago he had not submitted as SINGLE Driver issue in well over 6 months (at that time). I could go on but most of you already know how i feel.

First it "really irritates the hell out of me. How (you in this case hellbinder) state that " Carmack said something you have absolutley no proof he said, and which furthermore is not even what the person said he said, so you are going on a second hand report and misquoting it at that. Jeez man grow up and quit attacking people just b/c you are jealous that ATI isn't bequething you with tons of money for being their cute little attack dog.


Here is what Carmack ACTUALLY SAID

The NV30 runs the ARB2 path MUCH slower than the NV30 path.
Half the speed at the moment. This is unfortunate, because when you do an
exact, apples-to-apples comparison using exactly the same API, the R300 looks
twice as fast, but when you use the vendor-specific paths, the NV30 wins.

Here is what we were told he emailed to somebody but have no clue if it is true.

I dunno if anyone bothered emailing him, but this is what John Carmack had to say:

[Doom 3 on the NV40 will use] ARB2. The NV40 is incredibly fast, far and away the fastest thing
you can
get right now.

I have removed the NV30 path, because Nvidia now has the ARB2 path
running
well on the earlier hardware.

John Carmack

He doesn't say it is running at the same precision, even if he really wrote this email it doesn't imply he thinks they are using fp32, and not fp16. Nv very likely is replacing all arb2 calls with fp16 instead of fp32, but it doesn't really matter in D3 and that is what his company is concerned with, not some other situation or some other game.

People like you make me very angry b/c you make it so that people like J.C. don't want to answer any email b/c some moron will claim he said things he did not and that he is somehow evil. This means instead of replying they have to take forever and a day to make sure it cannot be misconstrued and instead they don't bother to reply.. and yes I have emailed and got a reply from jc before so I realize he actually does still respond although it would be in his best interest not to.

 

[Hellbinder]

This just smacks again why Carmack and Sweeney really irritate the hell out of me. How can they (in this case carmack) State that the nv30 is running the full percision path at full speed when he *KNOWS* they have to be changing things at the driver level to do the same kinds of changes he was forcing before with the game code? Its flatly Dishonest becuase he knows they are in essense *CHEATING* from a Marketing perspective that is Sure to come. Yet has nothing but good thigs to say, still claims Nvidias drivers are *better* even though i know PERSONALLY from about a year ago he had not submitted as SINGLE Driver issue in well over 6 months (at that time). I could go on but most of you already know how i feel.

First it "really irritates the hell out of me. How (you in this case hellbinder) state that " Carmack said something you have absolutley no proof he said, and which furthermore is not even what the person said he said, so you are going on a second hand report and misquoting it at that. Jeez man grow up and quit attacking people just b/c you are jealous that ATI isn't bequething you with tons of money for being their cute little attack dog.


Here is what Carmack ACTUALLY SAID

The NV30 runs the ARB2 path MUCH slower than the NV30 path.
Half the speed at the moment. This is unfortunate, because when you do an
exact, apples-to-apples comparison using exactly the same API, the R300 looks
twice as fast, but when you use the vendor-specific paths, the NV30 wins.

Here is waht we were told he emailed to somebody but have no clue if it is true.

I dunno if anyone bothered emailing him, but this is what John Carmack had to say:

[Doom 3 on the NV40 will use] ARB2. The NV40 is incredibly fast, far and away the fastest thing
you can
get right now.

I have removed the NV30 path, because Nvidia now has the ARB2 path
running
well on the earlier hardware.

John Carmack

He doesn't say it is running at the same precision, even if he really wrote this email it doesn't imply he thinks they are using fp32, and not fp16.

People like you make me very angry b/c you make it so that people like J.C. don't want to answer any email b/c some moron will claim he said things he did not and that he is somehow evil. This means instead of replying they have to take forever and a day to make sure it cannot be misconstrued and instead they don't bother to reply.. and yes I have emailed and got a reply from jc before so I realize he actually does still respond although it would be in his best interest not to.

Yawn...

If the Nvidia card is now running ARB2 then it can only be running in high Percision. Unless he changed the ARB2 path to use low percision and didnt tell anyone.

I could make a post about how people who dont know what they are talking about really irritate me as well... But i digress...

 

[Sxotty]

I realize this should be taken to pm's but if you want to make comments that have nothing to do with my post try this.


There is not reason to quote a bunch of meaningless information and clutter up the boards.

 

[Hellbinder]

hehe Ill keep that in mind next time. I am in a bit of a hurry at the moment 8)

(edited for personal brain cell failure issues)

 

[Richard]

I think many are flying off the handle for no good reason. Assuming that email conversation is real, JC says several things:

NV30 path is gone because pre-nv40 hardware runs well in the ARB2 path. He doesn't say it's running as fast as <insert your favourite card here>. It simply implies JC is happy enough with the performance he's getting with pre-nv40 hardware in ARB2 that makes him remove the nv30 path. Having said that I agree with whoever said throwing your work out the window is odd.

Also, JC never said all nv3x based cards would run on the nv30 path. It's more than likely the 5200/5600 would have to run in the nv20 path. Many people are assuming that mail means "all gffx will run well on the arb2 path". Just like many assumed that "recommended for DOOM 3" sticker on "GeForce FX" cards included the 5200/5600. The email does not say this.

About how nvidia could have improved their speed. Again, the last JC .plan is only about the nv30, and I think everyone agrees the nv35 and variant cores are a clear improvement over the nv30. It's possible that what JC calls pre-nv40 hardware means 5700/5900/5950. If so, this is highly unfortunate.

Are they cheating to get that speed? I hope not, but claiming they are based on that mail is reckless at best. If we don't get confirmation until then, we'll see it in the image quality comparisons. Either way, it's premature to speculate on these matters.

About UltraShadow, nvidia.com mention ultrashadow 1 is present in the 5900/5700/Go5700 chips.

 

[{Sniping}Waste]

Im afraid that Carmak will treat ATI like he treats AMD
This is sad that I have to say this about a Garland TX man.

 

[ChrisRay]

Since we are on the issue of doom 3. Has anyone considered the Quality Differences between the R200 path and the Nv20 pathway?

8500 users and Geforce 3/4 users specifically? My presumption was that they were similar in Quality but the 8500 does its shader routines in less passes.

Tho there is the integer 16 and integer 12 argument for IQ. I'm Almost certain that it would provide nothing tangible in concerns for Image Quality in this game.

Thoughts?

 

[Frank]

It is quite possible that the ARB2 path has increased in preformance on NV3x hardware by quite a bit. As has been discussed in this thread, OGLSL is much better to optimize by the driver (compiler) than DX9. The hardware can really shine when using OGLSL and improvements of the driver can be "honest" improvements, instead of replacing shaders like you need to do in a DX9 driver.

 

[Doomtrooper]

Depends on what Doom 3 uses for its featureset and if there is diffuse bump mapping used, there is image quality differences with the extra range that PS 1.4 offers, like this old example:

 

[AlphaWolf]

Actually he did, but that was over a year ago (Jan 29. 2003)

quote from his .plan:

The NV30 runs the ARB2 path MUCH slower than the NV30 path.
Half the speed at the moment. This is unfortunate, because when you do an
exact, apples-to-apples comparison using exactly the same API, the R300 looks
twice as fast, but when you use the vendor-specific paths, the NV30 wins.

Half the speed of nv30 path =! half the speed of r300 running arb2 path. Which is what I thought people were suggesting.

 

[Zeross]

It is quite possible that the ARB2 path has increased in preformance on NV3x hardware by quite a bit. As has been discussed in this thread, OGLSL is much better to optimize by the driver (compiler) than DX9. The hardware can really shine when using OGLSL and improvements of the driver can be "honest" improvements, instead of replacing shaders like you need to do in a DX9 driver.

ARB2 code path does not use GLSL, it's just straight assembly (ARB_VP & ARB_FP)

Concerning NV20 vs R200 precision Carmack already stated that it shouldn't be an issue :

Higher precision rendering. It appears that the GF3/GF4Ti clamps the results (including intermediate ones) when some part of the calculations goes over 1.0. The Radeon 8500, with up to 8.0 higher internal ranges, can keep higher numbers in the registers when combining, which allows for better lighting dynamics. How much will this have an impact in DOOM3's graphics?

At the moment, it has no impact. The DOOM engine performs some pre modulation and post scaling to support arbitrarily bright light values without clamping at the expense of dynamically tossing low order precision bits, but so far, the level designers aren't taking much advantage of this. If they do (and it is a good feature!), I can allow the ATI to do this internally without losing the precision bits, as well as saving a tiny bit of speed.

http://www.beyond3d.com/interviews/carmackdoom3/

 

[Frank]

Depends on what Doom 3 uses for its featureset and if there is diffuse bump mapping used, there is image quality differences with the extra range that PS 1.4 offers, like this old example:

Yes, it depends on the features used, but the pictures you use are misleading.

Like any high-level language, a lot of features used in the shader language have no direct assembly replacement. As long as the compiler produces a shader that delivers the result the developer wants to accomplish, the feature is useable. If it is fast enough, of course. What the compiler and hardware actually do to produce that effect is irrelevant. Also, when the compiler sees that it can use fp16 for some instructions without altering the resulting image in any way, that's a good optimization.

Only when the result is not guaranteed to be what the designer wanted to accomplish (disregarding programming bugs), can we say that the driver "cheated", like when replacing all those shaders from all those games with hand-tuned ones that have less precision. Which is something the hardware vendor absolutely doesn't want to spend all that time and money on in the first place!

And it was to be expected, that nVidia would spend time and money in creating a large performance gain, as they could use the boost. That is, if there is a large performance gain in the first place. I wouldn't be surprised, but this is the first time I heard about it.

 

[Frank]

It is quite possible that the ARB2 path has increased in preformance on NV3x hardware by quite a bit. As has been discussed in this thread, OGLSL is much better to optimize by the driver (compiler) than DX9. The hardware can really shine when using OGLSL and improvements of the driver can be "honest" improvements, instead of replacing shaders like you need to do in a DX9 driver.

ARB2 code path does not use GLSL, it's just straight assembly (ARB_VP & ARB_FP)

Ah, ok. In that case, I guess I'm talking nonsense. I should start writing shader programs, that way I know what I'm talking about.