Anand posts Initial HL2 results with new Dets

[Bopple]

Originally posted by WarheadMM
Stop thinking the new nvidia drivers instantly cheat.
Currently, it is very rational to think so, unless there are some hidden/untapped shader units in FX hardware.
But i don't remember nvidia claimed that. Or did they?

[Skuzzy]

You want a developers perspective?

I have been struggling for the last month to get this 5900U to display what I intend to be displayed. The damn drivers keep changing some of my shaders (and other things as well) and impacting what I am trying to get accomplished.
So, as a result, I end up having to go back and rewrite code to get around these damnable 'optimizations'.
In some cases it appears to just ignore what I send down the pipeline, but it does not return an error at all. Quite the contrary, it returns good status all the time, even though it is apparent it did not do what I told it to.

I have a full understanding of Gabe's frustration now. I'm no fanboy. I just want the damn thing to do what I tell it to do. I have been up all night working on this one effect, which the card/drivers claim it can do, but it doesn't.
So, I am left with having to accomplish it on the CPU side and just chalking it up as something the card/drivers is pretty much lieing about. Raising the frustration level, the effect works great on a Ti4600!

On the good side, I am keeping all this code handy with the intent on releasing a utility to show the problems. I am tired and fed up with this crap.

[digitalwanderer]

Originally posted by Skuzzy
On the good side, I am keeping all this code handy with the intent on releasing a utility to show the problems. I am tired and fed up with this crap.
Woo-hoo!! Another new tool for the toolbox! :D

[Sazar]

Originally posted by cthellis

The only other "ATi cheating" issue I know of was the one that simultaneously came out with nVidia's in regards to 3DMark03, in which they gained 4% (to nVidia's monolithically huge gains), and when the issue was brought to light and complained about, they said "terribly sorry, guys. We didn't change the nature of the test, but we know you don't to see this anyway so it's gone." at which point they removed it in the next Catalyst update.

overall increase was 1.9%... all of it pretty much coming from optimization of 2 shaders in gt4 where that increase was around 9% or so... removed from next drivers per statements by ati... acknowledgement of optimizations... :)

[jimmyjames123]

I have been struggling for the last month to get this 5900U to display what I intend to be displayed. The damn drivers keep changing some of my shaders (and other things as well) and impacting what I am trying to get accomplished.

By chance, have you posed these questions/concerns directly to anyone at NVIDIA? What is their response, if any? They claim that they have had some success when working directly with the developers in optimizing for their new hardware.

[gmontem]

Originally posted by jimmyjames123
By chance, have you posed these questions/concerns directly to anyone at NVIDIA? What is their response, if any? They claim that they have had some success when working directly with the developers in optimizing for their new hardware.
No developer should go through that kind of crap. If f(10) has been documented to return a value of 5 then it should do so. Something must be wrong if NVIDIA's returns a value of 7, don't you think? ;)

[StealthHawk]

Originally posted by cthellis
The only other "ATi cheating" issue I know of was the one that simultaneously came out with nVidia's in regards to 3DMark03, in which they gained 4% (to nVidia's monolithically huge gains), and when the issue was brought to light and complained about, they said "terribly sorry, guys. We didn't change the nature of the test, but we know you don't to see this anyway so it's gone." at which point they removed it in the next Catalyst update.

ATI cheats in 3dmark2001 GT4. NVIDIA does it in GT4 and other tests in 3dmark2001 too.

[Hellbinder]

Originally posted by StealthHawk
ATI cheats in 3dmark2001 GT4. NVIDIA does it in GT4 and other tests in 3dmark2001 too.
True Dat.

[AlphaWolf]

They claim that they have had some success when working directly with the developers in optimizing for their new hardware.

Developer X contacts nvidia with regards to making DX9 functions run at acceptable speeds.

...nvidia rep hands some money to developer x.

Developer X... DX 9 functions are overrated anyway.

[digitalwanderer]

Originally posted by AlphaWolf
Developer X contacts nvidia with regards to making DX9 functions run at acceptable speeds.

...nvidia rep hands some money to developer x.

Developer X... DX 9 functions are overrated anyway.
Well, that IS the definition of success for nVidia lately... :bleh:

[gmontem]

Hey guys let's start our own company to make a DX9 game and ask NVIDIA for assistance. We'll close shop as soon as NVIDIA forks over the cash to us. ;)

[digitalwanderer]

Originally posted by gmontem
Hey guys let's start our own company to make a DX9 game and ask NVIDIA for assistance. We'll close shop as soon as NVIDIA forks over the cash to us. ;)
How about an industry defining benchmark, it seems a tad easier... :lol:

[theultimo]

Originally posted by digitalwanderer
How about an industry defining benchmark, it seems a tad easier... :lol:

Hmm...i was thinking Dx9 CAD Software :)

[RipItup]

Originally posted by Hellbinder
Here it is...

http://www.anandtech.com/weblog/index.html

Of course you all know what i have to say about this.. so ill let you guys have first crack at it.

One word from me...

Disturbing.


Disturbing ? what, you mean the bit where it is summized :-

". So if you’re looking to buy a video card right now, ATI is the way to go."

Unless reviews say nVidia is worse than herpes you'll never be happy will you?

[digitalwanderer]

Originally posted by RipItup
Disturbing ?
Yup, deeply disturbing. That graph that shows the 5900Ultra magically neck-n-neck with the 9800 Pro in HL2 from "a reliable source" but NO other info?

Yeah, that's pretty stinking disturbing...it's basically implying that the magic det 5s will cure all of FX's woes and there is NO WAY that is possible without some severe image quality loss.

VERY disturbing. :(

[RipItup]

Originally posted by digitalwanderer
Yup, deeply disturbing. That graph that shows the 5900Ultra magically neck-n-neck with the 9800 Pro in HL2 from "a reliable source" but NO other info?

Yeah, that's pretty stinking disturbing...it's basically implying that the magic det 5s will cure all of FX's woes and there is NO WAY that is possible without some severe image quality loss.

VERY disturbing. :(

So he advised that everyone should buy nVidia cards then on his HL2 neck and neck benchmark ? So everyone is rushing out and buying them on a false premis ?

Clue: No and No

[digitalwanderer]

Originally posted by RipItup
So he advised that everyone should buy nVidia cards then on his HL2 neck and neck benchmark ? So everyone is rushing out and buying them on a false premis ?

Clue: No and No
No, but he is saying that we should expect the 5900Ultra & the 9800 Pro to be giving the same HL2 gameplay experience...which is damned hard to believe in right now with all evidence indicating the contrary.

[RipItup]

Originally posted by digitalwanderer
No, but he is saying that we should expect the 5900Ultra & the 9800 Pro to be giving the same HL2 gameplay experience...which is damned hard to believe in right now with all evidence indicating the contrary.

Part 2 .... to be continued . This is a never ending story mate, hence your 50000 posts :D Just like F1 cars

I'm going to take my kicked in nuts off to bed to feck the girlfriend. To much talk, not enough bonking.

I think you ought to pray before you have your cocoa and go to bed, beg forgiveness for calling into question Saint Anand. :D

[digitalwanderer]

Originally posted by RipItup
I think you ought to pray before you have your cocoa and go to bed, beg forgiveness for calling into question Saint Anand. :D
I save my respects for people who earn it, but I'm weird that way. :)

[theultimo]

Well, I'm waiting when the FX displays gourad shaded polygons in DX9, as nVidia's been "optimizing" way too much lately IMO.