...NV40 is 16 pipelines part, sometimes....

[silence]

NV40 is 16 pipelines part, sometimes.

this is taken from latest INQ article (i know, i know, it's INQ), but...still.....
wasnt it confirmed that it's 16 pipes?....
unless they think 32x0 here.
any ideas?

[mikechai]

NV40 is 16 pipelines part, sometimes.

this is taken from latest INQ article (i know, i know, it's INQ), but...still.....
wasnt it confirmed that it's 16 pipes?....
unless they think 32x0 here.
any ideas?

If The INQ is right, then NV40 is not 16 pipelines after all.
But the Baron and Uttar had confirmed the 16 pipes thing, WTF?

[harl]

http://www.theinquirer.net/?article=14878

INQ says it's 8 pipes hard + 8 pipes via shaders :balloon:

[Demirug]

Maybe someday Fuad will learn how a GPU works.

If you do not understand the technic behind something you should not try to play the "INVESTIGATIVE JOURNALISM" game.

[MUYA]

Can it get more confusing? first 6X2, then 16X1 now 8X2 but works as 16X1 with some VS3 tricks??? oh man.... :wtf:

[mikechai]

Can it get more confusing? first 6X2, then 16X1 now 8X2 but works as 16X1 with some VS3 tricks??? oh man.... :wtf:

INQ said 16x0 ... WTF?

[MUYA]

Here is the confusion for me...ok the b3d thread (dunno if its legit) said that the single texture fillrate was 7 GP/s and 7.2 GP/s for multi texture fillrate in 3dmark2k1. Now, wouldn't a 8X2 deliver more of a fillrate in multi texturing?

any takers on an explantion?

:wtf:

[AthlonXP1800]

However 3D Mark 2001SE only test PS 1.1 and 1.4. The fillrate test confirmed single texture fillrate of 7,010 Gigapixels/sec and multi texture fillrate of 7,234 Gigapixels/sec.

NV40 does have 16x1, I have a feel after read the inquirer, I think NV40 do have 16x1/8x2 hybrid because of multi texture fillrate result.

[zakelwe]

This is like flying into an airport in bad weather, at one point the clouds clear and you can see the runway and then the clouds come back in again and you are flying blind.

From demirug's and others statements I tend to dismiss this rumour from Fuad as being in error . I guess demirug cannot say due to NDA but I think he has indicated that it is not 8x2 using virtual pipes to go to 16x .

Regards

Andy

[Demirug]

If the single texture fillrate and the multi texturing fillrate they same (or nearly the same) they chip can work as ?x1. If you want to know if it is a full ?x1 desgin you have to test with 3,5,7 texturlayer, too.

btw: the fillrate here should MTexel/s and not MPixel/s.

[MUYA]

Perhaps Demirug could sorta enlighten us with the quad talk. Maybe Demirug, you could discuss what your predictiosn will be as pertaining to the amount of quads the pipelines of the nv40 or r420 can process at any time...under ideal circumstances. or sumat.... (pb)

[Brick_Top]

Uttar wrote:
NV40 is 16x1/32x0.
Unless NVIDIA never really taped-out anything, and they plan on making tons of money on pre-orders then changing their company name.

Uttar


http://www.beyond3d.com/forum/viewtopic.php?t=10990

:surrender

[Brick_Top]

Hanners wrote:
Supposedly not, 16 'proper' pipelines.


http://www.rage3d.com/board/showthread.php?threadid=33747967

:surrender

[MUYA]

http://www.beyond3d.com/forum/viewtopic.php?t=10990

:surrender


who really knows the real story at this point....

I sure hope for a card that is theorectically 16X1 :D

[Brick_Top]

If this is true I sense another year of "Driver hacking"

:banghead: :surrender

[{Sniping}Waste]

Looks like its a 8X2 in games and a 16X0 in some test to inflat scores. :mad: :banghead:

This reminds me of the NV30 paper launch with 4X2/8X0. :( :angry:

[Demirug]

Perhaps Demirug could sorta enlighten us with the quad talk. Maybe Demirug, you could discuss what your predictiosn will be as pertaining to the amount of quads the pipelines of the nv40 or r420 can process at any time...under ideal circumstances. or sumat.... (pb)

IMHO we should not replace counting piplines with simply counting quads.

If you anybody want to do this I will tell him that the P10 from 3dlabs is working on a tilebase (8*8 Pixel). It work on 4 of this tiles at the same time. Each of this Tileprocessors have one TMU. With this values it should run like hell but it did not because the number and sie of the working units (pixel, quad, tile) is only the half truth. You need to know how much work can be done on one working unit per clock. If you did not have this number you never can calculate how fast a chip is.

[MUYA]

IMHO we should not replace counting piplines with simply counting quads.

If you anybody want to do this I will tell him that the P10 from 3dlabs is working on a tilebase (8*8 Pixel). It work on 4 of this tiles at the same time. Each of this Tileprocessors have one TMU. With this values it should run like hell but it did not because the number and sie of the working units (pixel, quad, tile) is only the half truth. You need to know how much work can be done on one working unit per clock. If you did not have this number you never can calculate how fast a chip is.
Yes you are right...I was trying to say get a quote from u for the new cards and how in theory in ideal circumstances they can work just because you are someone with loads of knowledge on this matter. I think NV or ATI engineers are better in terms of getting their PS units to work on the quads with whatever functions that have been called for by a PS code....hmm i can't understand what i said either.

:D

I know it has been said that the NV3X has lots of stalls in its pipeline and that is why PS performance is bad. I hope at least t his time round NV has fixed or optimized its PS engine to get less stalls (or bubbles as DaveB puts it).

[Brick_Top]

:banghead: :banghead: :banghead:

http://www.beyond3d.com/forum/viewtopic.php?t=10993

At this rate we are only going to have a next-generation product by the time nv40.567 r420.566 arrive :surrender

"rumours" competing with "rumours" :angry:

[Demirug]

Yes you are right...I was trying to say get a quote from u for the new cards and how in theory in ideal circumstances they can work just because you are someone with loads of knowledge on this matter. I think NV or ATI engineers are better in terms of getting their PS units to work on the quads with whatever functions that have been called for by a PS code....hmm i can't understand what i said either.

:D

Sure I can calculate they performance of both Chips based on some assumption about the internal desgin of the pixelshader units. But it can be absolute wrong if my assumption is not correct.

I know it has been said that the NV3X has lots of stalls in its pipeline and that is why PS performance is bad. I hope at least t his time round NV has fixed or optimized its PS engine to get less stalls (or bubbles as DaveB puts it).

bubbles are more correct in this context. stall means the unit is complete blocked. This can happen too if you are run out of memery bandwith. Bubble means that they chip have to insert a dummy quad in the pipeline because it was not possible to insert a real one. At the end the dummy quad is throw away but the calculation power it has used is lost. But even if they chip did not insert bubbles there are still too many situations were the power is wasted.

[MUYA]

Further down in the B3D, someone translated an article by a german site Within Games. Here's that info.

Better translation by madshi (at b3d). Within Games (www.withingames.net).

"Finishing this year's CeBIT report we will tell you about some interesting rumors, which we were hearing during those 4 days:

- The NV40 architecture (16x1 for textures / color values, 32x0 for Z values) was confirmed by the x-th source. That cip is going to be gigantic.

- The NV40 will be faster than the R420. It can be expected that the R420 is between 15% and 5% slower than its NVidia competitor.

- There will be two models of the R420: A "Pro" card with 500Mhz chip and 475Mhz memory clock and a card called „Nitro“ with 600Mhz chip and 550Mhz memory clock.

- It's very probable, that the R423 will really have 16 pipelines, while the R420 comes with 12. Possibly those 2 chips are identical, with 4 pipelines (one quad block) getting disabled depending on the quality.

- The R420 (as another R300 refresh) is known to be a temporary solution. So ATi will not rely on the architecture too long anymore and will soon bring the R500, whose development state is reportedly quite far. Additionally, when developing the R500 ATI can rely on Microsoft support, cause the technology is planned to be used for the Xbox2 in one form or another."


Already posted here ;

http://www.nvnews.net/vbulletin/showthread.php?t=26385

;)

[-=DVS=-]

Nvidia's NV40 mimics a 16x1-pipeline architecture with just eight pipelines
The Source Inq (http://www.theinquirer.net/?article=14878)

Well well i told you so :D not real 16 pipe but 8 pipe , just with trick to act like 16 in some cases :rolleyes: i heard that before
There is mutch more explained at inq then i quoted , they say it will incure image quality becouse virtual 8 pipes can't use Bilinear, Trilinear or Anisotropic filtering :rant:


We understand that the NV40 will actually only have eight physical pipelines, but these will appear act like 16 in certain games and, indeed, in 3DMark 2001 Nvidia was telling people how 3DMark01 is a very nice benchmark since they can render 16 textures per pass in it, using only eight pipelines. What Nvidia is using is the ability of the Vertex Shader model 3.0 (known as PS 3.0 or VS 3.0) where the Shader can actually render textures as a virtual pipeline but can only render them without filtering information.

But, in modern games you always apply Bilinear, Trilinear or Anisotropic filtering to textures that you render, otherwise they will have some visual malformations. This is why, in normal games, eight pipelines will be able to render eight textures in single texturing pass and possibly even 16 if you render both textures in same pass using an 8x2 architecture


ATI's R420 offering remains at 12x1 architecture which makes this games-running business very complicated, since you cannot bring any final judgment about who will win the chip speed race.

However, we expect NV40 to deliver 16x1-like performance in Doom 3 (or should we say 16x0, which is how people call Nvidia's approach) where they render all the information except Z or colour features. This will be used by Doom III and games based on this engine.



Also note Inq is saying NV40 will get only 6500 Mtexels thats is 406.25 core , if R420 will get 500 Mhz core will have 6000Mtexels in everything not just conditional :smoking: looks like its same story again just like we have right now , one card will do superb in some games other will do in other were NV40 can use only 8 pipes.

[Demirug]

Fuad is changing his opinion on a nearly daily basis. I cannot believe that there are still people who take his article as fact.

[GlowStick]

Wow,

alot of new info, however i was almost ready to accept Nv40 as a 16pipe card, but good thing i held off!

[-=DVS=-]

Funny indeed the closer to release date we get the more rumors fly arround :D about NV40 that is there is hardly any about R420 :eek2:
Same happen in NV30/R300 period wonder why :wtf: