first r420 review leak

[jvd]

Ok, he benched with:
Breed_________TWIMTBP
FarCry________TWIMTBP
Call of Duty____TWIMTBP
Aquamark______benchmark of a TWIMTBP game
X2 the Threat___TWIMTBP
Colin McRae____TWIMTBP
Nascar Thunder 2004__TWIMTBP
Unreal Tournament 2004__TWIMTBP

The only game that I'm not sure of is "Halo-combat evolved". If it's not part of "TWIMTBP" program, I would be very surprised as Halo was a big game for nVidia.

well if you want to look for bias halo was an xbox game which most likely means they worked closer with nvida than ati .

 

[Vegetto-eX]

Most lame thread ever

 

[Johnny Rotten]

I must be missing something - are there no Far Cry AA + AF charts?

 

[Bitpower]

ok - everything is mirrored on the 3dc board if the computerbase is going down for some reason
(btw: where do i get a x800xt?)

Well, i think our server can go much more.

http://www.computerbase.de/forum/showthread.php?t=70949&page=2&pp=20

or direct:

http://www.computerbase.de/forum/showpost.php?p=646488&postcount=28

 

[Sanctusx2]

I must be missing something - are there no Far Cry AA + AF charts?

Check out images 61 and 62. Or if you're looking at the thread with allt he images, go to next page(5). ^^

 

[Laa-Yosh]

It's not just benches, there seem to be some architectural details too.

No sign of any extreme things here...

X800 seems to have 4 ALU ops per clock + a texture op in the pixel shader...?

No vertex tecturing here, so it seems no support for SM 3.0 at all.

 

[DemoCoder]

Yes its good on dx 8 games . But it sucks on dx 9 games. WHile the r300 is great on all the games .

So obviously based on the one example we have right now a smaller number of pipelines but higher clocks does not have just as good performance as a card with more pipes but lower clocks .

Neither card is the overall winner on all games, yet you interpret this as proving that higher clocks with less pipelines can't have just as good performance? Are you going to eat your words tommorow when the X800Pro proves it can have good performance compared to the 6800U?

The issues that torpedoed the NV3x in PS2.0 had alot more to do with register file architecture, legacy baggage, etc. An NV3x with 2 ALUs per pipe, 4-8 temporary registers without penalty, and similar clocks would have given a 9700 a run for its money. (e.g. take an NV40 with 4 pipelines @ 500Mhz, and 2 TMUs)

So if you can get off your nvidia is great kick you would see that the nv30 was crap and that the r300 was a much better designed card

I have never said either. Find a post written by me since the launch of the NV3x that claimed such. Otherwise, shut up and step off. I'm sick of pointing out hypocrisy and getting it turned into an NV3x "defense"

The f*nb*ys are looking for one card to completely dominate this round in everytest, and it just ain't gonna happen. It will be much more subjective this time, no overall winner, some cards winning in some games, losing in others. People buying for features vs IQ vs noise vs price vs overclockability vs favorite game performance, etc

 

[kipper67]

Ok, he benched with:
Breed_________TWIMTBP
FarCry________TWIMTBP
Call of Duty____TWIMTBP
Aquamark______benchmark of a TWIMTBP game
X2 the Threat___TWIMTBP
Colin McRae____TWIMTBP
Nascar Thunder 2004__TWIMTBP
Unreal Tournament 2004__TWIMTBP

The only game that I'm not sure of is "Halo-combat evolved". If it's not part of "TWIMTBP" program, I would be very surprised as Halo was a big game for nVidia.

well if you want to look for bias halo was an xbox game which most likely means they worked closer with nvida than ati .

Didn't Nvidia devrel actually write some of the shaders for gearbox in HALO - at the Nvidia editor's day Randy Pichford said something along the lines that they did a good job as the shaders ran as well on Ati hardware as their own.

Yeah - here you go http://www.bjorn3d.com/_preview.php?articleID=363&pageID=529&41507

Mark

Edited a couple of typos

 

[Johnny Rotten]

I must be missing something - are there no Far Cry AA + AF charts?

Check out images 61 and 62. Or if you're looking at the thread with allt he images, go to next page(5). ^^

Aah, thnx, nm, he just put the AA/AF details at the top and worded them differently so my eye didnt pick them out.

 

[Pete]

I must be missing something - are there no Far Cry AA + AF charts?

There are: "FSAA High, AF 4x." Both look good for ATi, but this is just one review.

Meh, I'm waiting for official reviews tomorrow. This leak left me as unsatisfied as the [H] NV40 one (maybe because I need the text to go along with the charts ). OTOH, [H]'s X800XT review format should be interesting now that they have both cards. And I'm really curious how reviewers will incorporate Temporal AA into their reviews and benchmark methodology.

Interesting how ATi is still much faster in AF, despite nV having adopted ATi's angle dependency.

 

[loekf2]

ok - everything is mirrored on the 3dc board if the computerbase is going down for some reason

http://www.forum-3dcenter.org/vbulletin/showthread.php?s=&postid=1776546#post1776546

(btw: where do i get a x800xt?)

Okay... finally pictures of the R420's architecture...

With my "John Doe" hat on, I see six vertex shaders, but what should I make of the pixelpipes ? I see four quads, sixteen identical pixelpipes, what happened to the 8 extreme ones ? 128 bits, FP32, looks there.

 

[anaqer]

Bit of an OT : could anyone please explain to me what the point in making the heatsink have it's upper right corner "chopped off" is?
It's not like it would interfere with the power connector or capacitors (like it did on the 9800XT) or something.
I thought it was common sense to maximize heatsink surface...


Also : "video processing engine"... someone will like this fo sho' 8)

 

[dan2097]

Ok, he benched with:
Breed_________TWIMTBP
FarCry________TWIMTBP
Call of Duty____TWIMTBP
Aquamark______benchmark of a TWIMTBP game
X2 the Threat___TWIMTBP
Colin McRae____TWIMTBP
Nascar Thunder 2004__TWIMTBP
Unreal Tournament 2004__TWIMTBP

The only game that I'm not sure of is "Halo-combat evolved". If it's not part of "TWIMTBP" program, I would be very surprised as Halo was a big game for nVidia.

well if you want to look for bias halo was an xbox game which most likely means they worked closer with nvida than ati .

Didn't Nvidia devrel actually write some of the shaders for gearbox in HALO - at the Nvidia edotors day Randy Pichford said something along the lines that they did a good job as the shaders ran as well on Ati hardware as their own.

Mark

Doesnt that imply they did an awful job

Ont hat computerbase link is basically the pictures from the whole review. The X800XT seems to lead especially with af/fsaa. Have to wait for more reviews, Id especially like to see hardocps and some synthetics, for example results from aquamark 3 and halo's built in time demo (which we can assume is what toms uses) are potentially valueless due to detection

 

[Johnny Rotten]

And I'm really curious how reviewers will incorporate Temporal AA into their reviews and benchmark methodology.

Interesting how ATi is still much faster in AF, despite nV having adopted ATi's angle dependency.

Temporal AA is going to be a nasty ball of wax since the 'artifacting' is going to be such a variable and subjective thing.

 

[ChrisW]

Ok, he benched with:
Breed_________TWIMTBP
FarCry________TWIMTBP
Call of Duty____TWIMTBP
Aquamark______benchmark of a TWIMTBP game
X2 the Threat___TWIMTBP
Colin McRae____TWIMTBP
Nascar Thunder 2004__TWIMTBP
Unreal Tournament 2004__TWIMTBP

The only game that I'm not sure of is "Halo-combat evolved". If it's not part of "TWIMTBP" program, I would be very surprised as Halo was a big game for nVidia.

well if you want to look for bias halo was an xbox game which most likely means they worked closer with nvida than ati .

Didn't Nvidia devrel actually write some of the shaders for gearbox in HALO - at the Nvidia edotors day Randy Pichford said something along the lines that they did a good job as the shaders ran as well on Ati hardware as their own.

Yeah - here you go http://www.bjorn3d.com/_preview.php?articleID=363&pageID=529&41507

Mark

Good find. I guess this makes it official. 100% of the games benchmarked by Tom's Hardware are optimized for nVidia hardware. Even Halo is using hand written shaders from nVidia themselves. This is why it's important to take Tom's Hardware with a grain of salt and wait for some more credible sites to release their reviews before making any kind of judgement here.

 

[trinibwoy]

Interesting ATI Pro 475/900/12 matching NV Ultra 400/1100/16 shows a superior architecture for ATI but ATI XT 525/1120/16 not much better?

Superior? How come when NVidia was going for high clocks but ATI was doing well with low clocks and high-parallelism (4pipe NV vs 8pipe ATI), it was ATI with the superior architecture (wide shallow pipes), and now that ATI has gone more for high clocks again it signals the superiority of ATI?

Is AMD with its good per-clock performance vs Intel P4 better? Performance and cost is what matters, not the architecture that was used to achieve it.

LOL. Damn Calm down dude. I'm not 'claiming' ATI has a superior architecture. I leave that to guys like you, Chalnoth, Hellbinder etc. And yes I agree with you that in a 400/475 16 pipeline contest the NV40 would appear more efficient than ATI. but this is 400*16 = 6400 vs 475*12 = 5700 raw theoretical fillrate. Of course there is a lot more at hand here but wouldnt you say that points to ATI being more efficient than the other way around?

 

[AlphaWolf]

I must be missing something - are there no Far Cry AA + AF charts?

The last few shots are farcry AA/AF. I guess they used in game settings, it says AA high, aniso 4.

 

[Bitpower]

Here is the original link again:
http://www.computerbase.de/forum/showthread.php?t=70949&page=2

 

[Ailuros]

And I'm really curious how reviewers will incorporate Temporal AA into their reviews and benchmark methodology.

Interesting how ATi is still much faster in AF, despite nV having adopted ATi's angle dependency.

Temporal AA is going to be a nasty ball of wax since the 'artifacting' is going to be such a variable and subjective thing.

I've been twisting my fingers over it for days now and I can't seem to be able to find any real usability for it. If I have performance to spare I'd rather invest it in a higher sampling pattern and/or in higher resolutions.

***oooops edit:

On the so far leaked results: from what I'm seeing so far ATI is holding itself more than just well. If the X800PRO truly should come at just 299$, then it has an excellent price/performance ratio.

 

[Eronarn]

Any chance of there being tests with R420 running the NV40 path and NV40 running the R420 path, just to check to see what the deal is there?