Anand posts Initial HL2 results with new Dets

[Hellbinder]

Here it is...

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

If those numbers hold true then things definitely look better than from Half Life 2 day, but we'll reserve judgement until we get the benchmark in house. I just thought you'd like to see what we're seeing, I wouldn't draw any conclusions based on this data yet, just wanted to share :)

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.

[extreme_dB]

I'd like to see if Valve has anything to say about Nvidia's performance gains when the drivers are official. Would they be prepared to publically criticize Nvidia for any IQ degradation? Or would they accept or even praise Nvidia's efforts as a beneficial compromise for their customers? We could be in for some high drama. :)

[ChrisW]

I like the way they keep posting benchmarks that they are not sure of and from "other sources". This has a bad smell to it.

How could they have doubled performance? They already had a special code path that reduced precision. I think the strategy here is to get these benchmarks leaked now using these hacked drivers knowing Valve will disable the cheats in the final version of the game. Then nVidia will claim Valve is disabling performance on GFFX cards just to punish nVidia and their customers (what they did with 3DMark03) and these benchmarks will be their "proof".

[Bopple]

OMG, I am so sick of this "Neverending Story".
When will this madness ever stop? Oh, please...

[rth]

I think I know what they did....

Valve treated FX's as DX8 hardware to make the game runable.

Now nVidia is going one better and treating it as DX7 hardware to make their cards competitive.

?

[serAph]

ya - I sure would like to know wtf is going on here - especially before I go and run out and buy a 9800... :mad:

[Sazar]

Originally posted by rth
I think I know what they did....

Valve treated FX's as DX8 hardware to make the game runable.

Now nVidia is going one better and treating it as DX7 hardware to make their cards competitive.

?

its partial precision for nvidia compared to dx9 codepath used by ati...

[jAkUp]

its still lower iq, and probably even lower iq since these cheating drivers, not to mention it still gets owned by 30fps in techdemo 5 owned:

when is punkbuster gonna detect driver cheats??:D

[nrdstrm]

We won't know for sure until the DAMN HL2 Bench and Det 52.10's come out. Of course if freaking people could keep thier words (Valve/Vivendi/Nvidia) the we would alread have our answers (as Valve had said that HL2 & HL2 Bench would be out 9/30 and Nvidia said Det 50's would be out BEFORE that). I personally think there is some serious scrambling to get this thing playable on Nvidia cards (Valve and Nvidia working on optimizations). I don't care WHAT Gabe said, he can't alienate a company that represents such a large part of the games's customers. Hope it comes out soon though...at least the bench. I wanna see what my crappy 5900U can pull of...lol.
Nrdstrm

[mikechai]

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.

What is disturbing you? (since you won't buy a FX card)

[serAph]

Originally posted by jAkUp
its still lower iq, and probably even lower iq since these cheating drivers, not to mention it still gets owned by 30fps in techdemo 5 owned:

when is punkbuster gonna detect driver cheats??:D

no offense, but you sound sorta stirred up by that benchmark comparison. Are you lettin it get personal?

Kudos to nVidia for pullin themselves outta the rut they were in, thats for sure, but Im still buyin this 9800 pro when it comes out next week to replace my 5900

http://www.ocprices.com/index.php?rev_id=183&page=4&action=reviews

[Hellbinder]

Originally posted by mikechai
What is disturbing you? (since you won't buy a FX card)
The fact that Nvidia in their official response to valve a couple weeks ago stated plainly that they would replace PS2.0 with PS 1.4 wherever they felt it was "Ok".

Meaning I gurantee you that even this "Mixed mode" test Nvidia is doing Specific per level application detection and completely replacing the games code with DX8 stuff. Meaning even replacing Valves own Shaders that they customized for Nvidia themselves.

It means that it is flatly not DX9 Vs DX9 at all. It is DX8 being Reported as DX9 "mixed mode". True "Mixed mode" is still DX9 it just uses FP16. This is clearly not whats happening here.

This my Friends is outright Fraud.

[mikechai]

Originally posted by Hellbinder
The fact that Nvidia in their official response to valve a couple weeks ago stated plainly that they would replace PS2.0 with PS 1.4 wherever they felt it was "Ok".

Meaning I gurantee you that even this "Mixed mode" test Nvidia is doing Specific per level application detection and completely replacing the games code with DX8 stuff. Meaning even replacing Valves own Shaders that they customized for Nvidia themselves.

It means that it is flatly not DX9 Vs DX9 at all. It is DX8 being Reported as DX9 "mixed mode". True "Mixed mode" is still DX9 it just uses FP16. This is clearly not whats happening here.

This my Friends is outright Fraud.

Agree with that part. It will definitely not an apple to apple comparison anymore.
I think ATI will have an edge on image quality no matter how good the "automatic shader replacement" is.
We just have to put that in mind when we read any reviews.

But if the new Dets able to put up that kind of score even in mixed mode or partial precision or whatever, they have already done a great job.
Then FX card owners will still have a great gaming experience with HL2.

[Ruined]

Let's wait and see the final product before going on a witchhunt. These are very positive numbers, however.

[AthlonXP1800]

Originally posted by jAkUp
its still lower iq, and probably even lower iq since these cheating drivers, not to mention it still gets owned by 30fps in techdemo 5 owned:

when is punkbuster gonna detect driver cheats??:D

It did not mentioned which driver it used, I think it probably use the latest 52.14 driver, it has better IQ than before in beta 51.75 with lower IQ.

[Hellbinder]

Originally posted by mikechai
Agree with that part. It will definitely not an apple to apple comparison anymore.
I think ATI will have an edge on image quality no matter how good the "automatic shader replacement" is.
We just have to put that in mind when we read any reviews.

But if the new Dets able to put up that kind of score even in mixed mode or partial precision or whatever, they have already done a great job.
Then FX card owners will still have a great gaming experience with HL2.
I dont think you understand. We are not talkng about an "Edge in IQ" here. This means no HDR support, and many of the effects simply wont even be close to the same.

I have to be honest it really bugs me that people simply do not understand the nature of whats going on. Because it means that Nvidia is likely to get away with it. This is not a true representation of "Mixed-Mode". Mixed mode is FP16/FP32 and is fully DX9. Valves shaders were already completely optomized in this way for the entire FX line. This is obviously a complete Replacement of the games code as Nvidia sees fit to get their numbers up.

What if ATi Suddenly stopped rendering half the effects and reduced some Quality and gained 30FPS over what they have now. Would you guys all sit around here and Applaud ATi.. patting them on the back and Cheering for what an amazing job the Driver team did???

Yet the line of people are already forming to congradulate Nvidia as heros for doing Exactly That.

[Hellbinder]

Originally posted by AthlonXP1800
It did not mentioned which driver it used, I think it probably use the latest 52.14 driver, it has better IQ than before in beta 51.75 with lower IQ.
Anand makes around 4 comments in his review regarding image Quality "issues". which he refused to go into. But in each case it was a benchmark where Nvidia had gained considerable ground.

[StealthHawk]

Wow, I applaud NVIDIA's clairvoyant driver team. After the drivers are optimized NVIDIA's scores in benchmarks are always at the same level as ATI's. Amazing, isn't it? It's just one big coincidence after coincidence :p

[ChrisW]

Originally posted by StealthHawk
Wow, I applaud NVIDIA's clairvoyant driver team. After the drivers are optimized NVIDIA's scores in benchmarks are always at the same level as ATI's. Amazing, isn't it? It's just one big coincidence after coincidence :p
LOL! Why stop there? Why not make their drivers give twice the frame rate of ATI's cards? That would be something!

[StealthHawk]

Originally posted by ChrisW
LOL! Why stop there? Why not make their drivers give twice the frame rate of ATI's cards? That would be something!

Then that would make it even more obvious! Actually, no, it wouldn't. It's obvious enough already :p

[jimbob0i0]

hehe I can't wait till the det 50's are released... hopefully they will coincide with the HL2 benchmark and cat 3.8's

/side note... I personally think that HL2 benchmark will be made avaliable for download/steam on October 8th to coincide with cat 3.8's and the first set of people receiving their 9800XT's/

I want to see IQ comparisons.... but more importantly I want the det 50's to be run in 'full DX9 mode' 'mixed mode DX9' and 'DX8 mode' to see what the IQ and speed differences are on the NV3X with the drivers over all three modes.... if the modes are practically identical with these drivers in IQ/effects/speed I think it would then be safe to say that NV has then committed fraud by declaring a DX9 setting but only using DX8 level technology...

Also I can't wait for October 31st and see what Futuremark with their new rules make of these drivers - does anyone think that NV will actually have drivers that can be allowed on that date considering the sheer effort that appears to be going into cheating right now?

[r2d2d3d4d5]

When was HL2 benchmarked on a 5950U? Perhaps this is the card that will turn it all around.

My guess is that this benchmarks have surfaced to spoil the XT/Cat3.8 launch and will probably disappear again in a few weeks.

[nForceMan]

Originally posted by Hellbinder


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



Well, well. NV38 flexes his muscle AGAIN! :rofl

That awesome NV driver develompent team. I love those guys! :super:

[Richthofen]

Originally posted by Hellbinder
The fact that Nvidia in their official response to valve a couple weeks ago stated plainly that they would replace PS2.0 with PS 1.4 wherever they felt it was "Ok".

Meaning I gurantee you that even this "Mixed mode" test Nvidia is doing Specific per level application detection and completely replacing the games code with DX8 stuff. Meaning even replacing Valves own Shaders that they customized for Nvidia themselves.

It means that it is flatly not DX9 Vs DX9 at all. It is DX8 being Reported as DX9 "mixed mode". True "Mixed mode" is still DX9 it just uses FP16. This is clearly not whats happening here.

This my Friends is outright Fraud.

as long as i don't see a huge difference in quality i won't care what Nvidia replaces in that game.
ATI payed VALVE to code that game the way they wan't to have it.
If Nvidia now replaces some things for their cards. Fine...
Pretty funny. Nice 8 Mio dollar investment :)

[ChrisW]

Originally posted by nForceMan
Well, well. NV38 flexes his muscle AGAIN! :rofl

That awesome NV driver develompent team. I love those guys! :super:
LOL! What makes you think they NV38 is so great? It's benchmarks show only a marginal improvement over the NV35.