ATi unveil R420

[kemosabe]

Yes, I followed MuFu's link to the other thread and made the connection.

I wonder if we've finally pinned down the number of functional pipes in the X800 Pro. Assuming that it has 12 pipes, it stands to reason it will bow to the 6800 Ultra unless ATI pulls some rabbit out its posterior. The XT, meanwhile, should be another story and the one which hardware sites get to talk about in the near future.

Question for Dave: without revealing anything specific about either card, can you at least confirm that ATI will be sending reviewers samples of the fastest version of R420 ? :|

 

[Colourless]

The card in the new pic seems to have more caps too

 

[MuFu]

The card in the new pic seems to have more caps too

It's a photoshopped 9800XT pic. The X800 Pro doesn't have RAMsinks and the layout is a bit different (as seen in the one, legit photograph currently available on the web).

I'd expect the HS/fan on the X800 XT to be more substantial. RAMsinks and dual Molex as well, perhaps?

MuFu.

 

[kemosabe]

Well the latest talk is that it isn't a (fake) photoshopped 9800XT. The guys at Clubic have apparently confirmed that this was an official photo released to the press by ATI. The speculation is that ATI modified the 9800XT pic itself, but that this doesn't quite represent the final product, although the expectation is that the resemblence will ultimately be significant.

http://www.presence-pc.com/news/n3698.html

Edit: Never mind, I see it's being discussed on the other thread.

 

[Joe DeFuria]

The yellow connector has nothing to do with power.

I'm guessing: USB 2.0 port?

http://www.ati.com/companyinfo/press/2004/4748.html

 

[Sxotty]

2) nVidia has not been alleged to overclock their memory on many of their reference boards...it's a fact.

Hey joe didn't they supposedly underclock some of it as well though? I thought I read it was a mix of 500 and 600 all clocked to 550? Was that correct?

 

[Joe DeFuria]

Hey joe didn't they supposedly underclock some of it as well though? I thought I read it was a mix of 500 and 600 all clocked to 550? Was that correct?

That is correct.

Though it's actually common to release retail products where the memory is not actually clocked as high as the spec.

 

[Evildeus]

So what's the problem

 

[Sxotty]

The inq which is full of lies and deceit, is saying that the R420XT flavor will be clocked at 525

but actually it learned the Nvidia numbers from online reviews and decided to increase the frequency a little bit more just to match Nvidia's Geforce 6800 Ultra threat.

Knowledgeable friends tell us that Nvidia wins 3dmark 2003 and loses but in most of the other game test ATI wins over Nvidia. Still, bear in mind that ATI is only marginally faster when it's faster and we are not talking about quantum leaps here. Whoever wins in the game tests, whether it's Nvidia or ATI, wins by a small margin.

Now I am slightly confused. Last generation we all know NV borked things, but ATI had much more efficient pipes, now the X800XT supposedly has 16 pipes, and is clocked higher and is still fp24 right? All of which seems to indicate that it will be faster even if they are equally efficient, yet we are led to believe that ATI will be more efficient still. In addition the 3Dc compression or whatever should increase efficiency even further, so how is NV going to be competitive at all? I am honestly curious it seems as though the rumors about ATI are dichotomous or something...

 

[anaqer]

Now I am slightly confused. [...]

Yep, even with 525MHz I'd expect the XT to have a more significant performance delta compared to NV40U. Unless they are having supply troubles with those uberfast Micron chips thus R420 ends up severely bandwith limited, I don't know what's going on. Even IF NV40 and R420 are about equally efficient, 30% more clockspeed should get the XT more than just a tie with the NV40U.

 

[digitalwanderer]

Even IF NV40 and R420 are about equally efficient, 30% more clockspeed should get the XT more than just a tie with the NV40U.

To the nV40u that was previewed all over the web, yes....but how it will stack up against the REAL ultra is what the big question is.

 

[hstewarth]

Even IF NV40 and R420 are about equally efficient, 30% more clockspeed should get the XT more than just a tie with the NV40U.

To the nV40u that was previewed all over the web, yes....but how it will stack up against the REAL ultra is what the big question is.

Also according to interview with CEO of makers of FarCry and tech specs of SM 3.0, SM 3.0 will signficantly increase the performance of application over SM 2.0. Who knows what drivers changes are coming and what NVidia has for the final release of hardware.

One thing for sure, this is not going to be a one-sided fight. Which is good for the customers.

 

[Joe DeFuria]

Also according to interview with CEO of makers of FarCry and tech specs of SM 3.0, SM 3.0 will signficantly increase the performance of application over SM 2.0.

1) Where did he say "significantly"
2) Performance on Architecture A's PS 3.0 implementation vs their own 2.0 implemention, doesn't say much about Architecture B's PS .20 implementation.

 

[anaqer]

Who knows what drivers changes are coming and what NVidia has for the final release of hardware.

Yes, yes, who knows. But you DO realize I was referring to the Inq findings, right? Using (at max) PS2.0 titles? With current drivers? With current hardware?

One thing for sure, this is not going to be a one-sided fight. Which is good for the customers.

Agreed.

To the nV40u that was previewed all over the web, yes....but how it will stack up against the REAL ultra is what the big question is

I thought we were done with the "475MHz" and "thisisnotanultra" rumours by now?

 

[digitalwanderer]

I thought we were done with the "475MHz" and "thisisnotanultra" rumours by now?

Well, I'm done with it in I don't think it's a rumor so much as a fact that we just haven't gotten confirmation on yet...

 

[T2k]

I thought we were done with the "475MHz" and "thisisnotanultra" rumours by now?

Well, I'm done with it in I don't think it's a rumor so much as a fact that we just haven't gotten confirmation on yet...

Well said...

 

[anaqer]

I'm surprised, but it's definitely good to hear... although I kinda wonder how easily the cooling setup we've seen so far will handle the extra load...?

 

[digitalwanderer]

I'm surprised, but it's definitely good to hear... although I kinda wonder how easily the cooling setup we've seen so far will handle the extra load...?

The only thing I'm sure on right now with regards to whatever-the-hell-they-call-their-actual-top-end-model-whenever-it-comes-out is that no one knows anything about what it's going to finally end up being right now....not even nVidia.

They're waiting for ATi to show their hand before they figure out what they'll need to beat it, and they're going to try their balls off to whatever the cost. 8)

 

[Sxotty]

They're waiting for ATi to show their hand before they figure out what they'll need to beat it, and they're going to try their balls off to whatever the cost. 8)

I hope this is true b/c it will make even more competition. Considering how many cards were released last year if the trend continues there will be tons to choose from.

However how do you conclude that ATI is not going to do the exact same thing? I mean their supposed specs keep going up and up Woo faster

 

[hstewarth]

Also according to interview with CEO of makers of FarCry and tech specs of SM 3.0, SM 3.0 will signficantly increase the performance of application over SM 2.0.

1) Where did he say "significantly"
2) Performance on Architecture A's PS 3.0 implementation vs their own 2.0 implemention, doesn't say much about Architecture B's PS .20 implementation.

Significantly is based on my interperations of what the ceo states and tech specs of Direct X 9 are on.

On the performance of differennt architextures, it could mean that there is a performance increase because of changes of specations. It does not rule out that if Architextuie's A 2.0 implementation is the same performance as Architextures B 2.0 implemention that Architexture's 3.0 implemention would be faster than Architexture 2.0 implemention. So if two competitor chips are on the level for 2.0, then I believe that the 3.0 implemention will be fastest of any implemention.

In reality, it going to be hard to judge for a lot of reasons.. some that come to mind are..

1. Game programers will likely not program the same way for each card.

2. Game programmers will take advantage of 2.0 and 3.0 differently - maybe by personal perference, by company ties...what hardware they are running personally...

3. Not likely that both sets of hardware will be equal.


I also believe that a true benchmark can not be made between the two.. One will stated not using 3.0 that is unfair and other states that both should be on 2.0 because that is what is common between them.

There could be many factors that would determined which card is better.. but likely this time there is going to be no clear winner. Ones that like A's will like A's and ones that like B's will like B's... Also likely that some A's that switch to B will switch back to A and vice-versa..

Should be interesting...