Inquirer spreading R420 info

[AlphaWolf]

Re: Hmmm, sounds slow to me?

Hmmm, although this is all idle speculation and I don't want to turn this into an argument I don't see how a clock for clock R420(XT) and NV40(Ultra) would show dramatically different FPS. I know that statement alone is ludricous but assuming from the initial previews the NV40 is a well balanced piece of hardware and already scroes the same as 2xR350 how can the R420 supposedly also based on 2xR350 cores produce anything miraculously higher?

Why would the comparison be clock for clock? You think the cards will actually be reviewed that way? Frames, triangles, textels per second seems to matter more to most people.

Performance surely will be down to a clock speed race? It used to be apparent that the PS/VS aspects of R3xx series and perhaps even it's 24bit precision helped against the NV30 and NV35. Unfortunately the limited testing so far appears to show the NV40s PS/VS units and 32 precision working just fine. Combined with a healthy PP boost and it's suddenly not so easy to see ATIs advatange? Thus for me, ATI need speed and plenty of it and from these current speculations I don't think it's enough. Knowing the ATI of old, they may not even go for 500Mhz on the XT if the PRO is set at 475Mhz. And as for that 400Mhz figure, boy do I hope that's "gobbledygook"
I suspect even a 475Mhz 12pipe will be shown a very clean pair of heals by any 16 pipe card clocked 75Mhz lower or not yet alone a 12 pipe card running at 400Mhz.

20% more clock speed vs 33% more pipes. That's not a huge difference and could certainly be made up by other advantages such as superior shader or VS performance. Most of the tests I have seen show the 6800U beating the 9800xt by 50-80% with a few over a 100% and a few closer to 20%. A 12 pipe 9800XT at 475 mhz (~60% more performance over 9800xt) seems like it would be a somewhat comparable part. Obviously that is way over simplified as I am sure there will be other changes to the r420 architecture, but I am inclined to believe they will tend more to be improvements.

 

[Bjorn]

Re: Hmmm, sounds slow to me?

Most of the tests I have seen show the 6800U beating the 9800xt by 50-80% with a few over a 100% and a few closer to 20%.

And a few closer to 150%

 

[Seiko]

Re: Hmmm, sounds slow to me?

Hmmm, although this is all idle speculation and I don't want to turn this into an argument I don't see how a clock for clock R420(XT) and NV40(Ultra) would show dramatically different FPS. I know that statement alone is ludricous but assuming from the initial previews the NV40 is a well balanced piece of hardware and already scroes the same as 2xR350 how can the R420 supposedly also based on 2xR350 cores produce anything miraculously higher?

Why would the comparison be clock for clock? You think the cards will actually be reviewed that way? Frames, triangles, textels per second seems to matter more to most people.

Performance surely will be down to a clock speed race? It used to be apparent that the PS/VS aspects of R3xx series and perhaps even it's 24bit precision helped against the NV30 and NV35. Unfortunately the limited testing so far appears to show the NV40s PS/VS units and 32 precision working just fine. Combined with a healthy PP boost and it's suddenly not so easy to see ATIs advatange? Thus for me, ATI need speed and plenty of it and from these current speculations I don't think it's enough. Knowing the ATI of old, they may not even go for 500Mhz on the XT if the PRO is set at 475Mhz. And as for that 400Mhz figure, boy do I hope that's "gobbledygook"
I suspect even a 475Mhz 12pipe will be shown a very clean pair of heals by any 16 pipe card clocked 75Mhz lower or not yet alone a 12 pipe card running at 400Mhz.

20% more clock speed vs 33% more pipes. That's not a huge difference and could certainly be made up by other advantages such as superior shader or VS performance. Most of the tests I have seen show the 6800U beating the 9800xt by 50-80% with a few over a 100% and a few closer to 20%. A 12 pipe 9800XT at 475 mhz (~60% more performance over 9800xt) seems like it would be a somewhat comparable part. Obviously that is way over simplified as I am sure there will be other changes to the r420 architecture, but I am inclined to believe they will tend more to be improvements.

I suggested clock for clock as it wouldn't surprise me if both the NV40(Ultra) and R420(PRO/XT) end up being 50-75Mhz within each other.

As for clocks Vs Pipes. Pipes every time please. Maybe it's my naitivity but take a look at the 9500Pro/9600Pro. Maybe other factors played a larger part but from my limited understanding the 8 pipes of the 9500Pro simply gave it too much of ahead start and the 9600Pro just couldn't get in reach no mater what clock speed advantage it had. In fact the 9600 for me at least was a method for ATI to improve yields and margins not for performance.

 

[AlphaWolf]

Re: Hmmm, sounds slow to me?

I suggested clock for clock as it wouldn't surprise me if both the NV40(Ultra) and R420(PRO/XT) end up being 50-75Mhz within each other.

Well 75mhz is almost 20% isn't it.

As for clocks Vs Pipes. Pipes every time please. Maybe it's my naitivity but take a look at the 9500Pro/9600Pro. Maybe other factors played a larger part but from my limited understanding the 8 pipes of the 9500Pro simply gave it too much of ahead start and the 9600Pro just couldn't get in reach no mater what clock speed advantage it had. In fact the 9600 for me at least was a method for ATI to improve yields and margins not for performance.

Oh I agree the 9600 pro was a margin improvement not performance. You are also talking about 100% more pipes not 33% more.

 

[AlphaWolf]

Re: Hmmm, sounds slow to me?

Most of the tests I have seen show the 6800U beating the 9800xt by 50-80% with a few over a 100% and a few closer to 20%.

And a few closer to 150%

I have seen a couple with the 9800XT ahead also, but I am more inclined to believe the mean type results.

 

[Hellbinder]

If the 12 pipeline card is able to diliver more than one pixel or operation per clock. Then it may well be faster than a 16 pipe card. There are more things to consider than just the raw number of pipelines.

Just a thought.

 

[surfhurleydude]

Re: Hmmm, sounds slow to me?

Most of the tests I have seen show the 6800U beating the 9800xt by 50-80% with a few over a 100% and a few closer to 20%.

And a few closer to 150%

I have seen a couple with the 9800XT ahead also, but I am more inclined to believe the mean type results.

Um...there are absolutely no benchmark results that put the 9800 XT ahead of the 6800 Ultra EXCEPT in CPU limited situations where the 6800's immature drivers fall a little bit short. And if you can recall correctly, the same thing happened with the 9700 Pro vs. the Ti4600 as well.

 

[PaulS]

Re: Hmmm, sounds slow to me?

IQ could be the ace in the hole but I'd assumed FSAA would remain as is. Have you heard of anything on this?

Why do you make that assumption?

 

[Hellbinder]

Re: Hmmm, sounds slow to me?

IQ could be the ace in the hole but I'd assumed FSAA would remain as is. Have you heard of anything on this?

Why do you make that assumption?

hehehehe... 8)

 

[McDusty]

Re: Hmmm, sounds slow to me?

Um...there are absolutely no benchmark results that put the 9800 XT ahead of the 6800 Ultra EXCEPT in CPU limited situations where the 6800's immature drivers fall a little bit short. And if you can recall correctly, the same thing happened with the 9700 Pro vs. the Ti4600 as well.

Its all in the FSAA:

http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD03NDcmdXJsX3BhZ2U9MjI=

http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD03NDcmdXJsX3BhZ2U9MjM=

 

[PaulS]

Re: Hmmm, sounds slow to me?

Its all in the FSAA

Both of those links show scores below 9800XT only when using 6x and 8x AA. We've already established the performance is terrible in those modes. 4x is what it's designed for.

 

[Hellbinder]

Re: Hmmm, sounds slow to me?

Its all in the FSAA

Both of those links show scores below 9800XT only when using 6x and 8x AA. We've already established the performance is terrible in those modes. 4x is what it's designed for.

Does that somehow excuse it? When i can play game X at a usable resolution with much HIGHER FSAA+AF and get MUCH HIGHER and very playable scores.

Why should Nvidia get a pass on this? If you play those games with eye candy turned up then even the 9800XT is a far better choice. Thats just simple math.

 

[Ardrid]

Isn't NVIDIA's current 8x mode broken right now and thus not indicative of the performance shown?

 

[Seiko]

Re: Hmmm, sounds slow to me?

IQ could be the ace in the hole but I'd assumed FSAA would remain as is. Have you heard of anything on this?

Why do you make that assumption?

I'm combining a number of things (all from rumours mind)

1) The original R420 was scrapped
2) The idea of 2xR350 cores in a 175Million package comes in instead.
3) The R350 had very good IQ already and is still in front of the competition.
4) No new features have apparently been added.
5) ATI themselves said they would concentrate on speed.
6) The Xbox2 contract has been won and will need resource diverted
7) ATI have shown over the last 18 months they are prepared to call speedbinned and very minor tweaked chips new products.

This leads me to believe that we'll not get any better IQ from ATI this time around but please believe me when I say I really hope they make me eat my words!

 

[AlphaWolf]

Re: Hmmm, sounds slow to me?

Most of the tests I have seen show the 6800U beating the 9800xt by 50-80% with a few over a 100% and a few closer to 20%.

And a few closer to 150%

I have seen a couple with the 9800XT ahead also, but I am more inclined to believe the mean type results.

Um...there are absolutely no benchmark results that put the 9800 XT ahead of the 6800 Ultra EXCEPT in CPU limited situations where the 6800's immature drivers fall a little bit short. And if you can recall correctly, the same thing happened with the 9700 Pro vs. the Ti4600 as well.

I didn't say why, and you don't know why, but the fact is that there are cases where the 6800U has benched slower than the 9800XT. Check out Mafia on this page.

As I said I am inclined to ignore abhorrent results. Both ways.

Note that it is not an 8xAA mode bench.

 

[Evildeus]

Isn't NVIDIA's current 8x mode broken right now and thus not indicative of the performance shown?

I think so.

 

[PaulS]

Re: Hmmm, sounds slow to me?

IQ could be the ace in the hole but I'd assumed FSAA would remain as is. Have you heard of anything on this?

Why do you make that assumption?

1) The original R420 was scrapped
2) The idea of 2xR350 cores in a 175Million package comes in instead.
4) No new features have apparently been added.

All inaccurate

3) The R350 had very good IQ already and is still in front of the competition.

Just about at 4x AA, but it's not the kind of gap it once was. Ever heard the term "Moving Target"? Just because it would suit NVIDIA for ATi to keep the same IQ, it doesn't mean it will work out like that.

5) ATI themselves said they would concentrate on speed.

Speed and improved IQ aren't mutually exclusive.

6) The Xbox2 contract has been won and will need resource diverted

Extra engineers were hired as a direct result of the console deals.

7) ATI have shown over the last 18 months they are prepared to call speedbinned and very minor tweaked chips new products.

Refresh products (and this isn't one) are almost always "tweaks" of existing chips. And the same things could be levelled at NV, yet they improved their own IQ (admittedly not past ATi level) with their new product.

I didn't say why, and you don't know why, but the fact is that there are cases where the 6800U has benched slower than the 9800XT. Check out Mafia on this page.

Hmm, thanks for that link. Interesting result, which seems at odds with other DX8 titles at 4xAA. May be driver related (or not. Difficult to tell right now).

 

[Seiko]

Re: Hmmm, sounds slow to me?

IQ could be the ace in the hole but I'd assumed FSAA would remain as is. Have you heard of anything on this?

Why do you make that assumption?

1) The original R420 was scrapped
2) The idea of 2xR350 cores in a 175Million package comes in instead.
4) No new features have apparently been added.

All inaccurate

Sorry that should have been the R400 was scrapped. As for the others being inaccurate perhaps but can you explain why you feel they are?

 

[DW Fan!!!!!]

I love it when people bring up the fact that developing for the XBOX2 can hurt development because it hurt Nvidia, but the thing is everyone seems to forget at that very same time ATI were developing the graphics for nintendo's gamecube, the difference between these is that as far as I am aware, Nvidia wanted to make the parts themselves of there designs (microsoft did not like this) and ATI created there GPU and let nintendo arrange to have it manufactured themselves giving ATI a cut.

This round ATI are developing for both the XBOX2 and the GAMECUBE2 but once again they are creating the hardware of which they will then pass on to Microsoft and nintendo to arrange production themselves.

This excuse should not be used as a reason for nvidia bringing out a crap card while ATI brought out a good one and definatly should not be used as a reason why ATI might fail this round when no one has seen anything from them.

Peace people

 

[PaulS]

Re: Hmmm, sounds slow to me?

Sorry that should have been the R400 was scrapped. As for the others being inaccurate perhaps but can you explain why you feel they are?

The rumour you're talking about with your second point was actually 3 or 4 RV360, not 2 R350. And even then, it's not literally 4 separate cores in there.

As for the last one - what do you expect me to say?