A computer components & hardware forum. HardwareBanter

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

Go Back   Home » HardwareBanter forum » Video Cards » Ati Videocards
Site Map Home Register Authors List Search Today's Posts Mark Forums Read Web Partners

Bad news for ATI: Nvidia to 'own' ATI at CeBit - no pixel shader 3.0 support in R420 (long)



 
 
Thread Tools Display Modes
  #1  
Old February 22nd 04, 06:48 PM
NV55
external usenet poster
 
Posts: n/a
Default Bad news for ATI: Nvidia to 'own' ATI at CeBit - no pixel shader 3.0 support in R420 (long)

Bad news for ATI: Nvidia to 'own' ATI at CeBit / no pixel shader 3.0
support in R420

It looks like R420, while potentially up to 3 times faster(in PS 2.0
and VS 2.0 shaders) than R360, will not be PS 3.0 compliant. If NVIDIA
notches up similar PS 2.0 performance, ATI is in trouble.




quote:
--------------------------------------------------------------------------------
Lastly there is word on the NV40 R420 front. Neither card was shown,
but several people who would be familiar with both have said Nvidia is
going to come back very strong soon. The word 'own' was bandied about,
as in Nvidia will own ATI. Not sure what it means, but CeBit should be
a very interesting show.
--------------------------------------------------------------------------------



Babelfish translation of the 'no PS2.0 shaders in R420' article:




quote:
--------------------------------------------------------------------------------
ATi R420, R423 & CO. Information
22. February 2004/from Leonidas /page 1 of 2



After we could give at the end of January an already rather complete
overview of the coming nVidia Grafikchips, naturally the question was
located in the area, which is to be expected 2004 now from ATi for the
spring. At the time of the nVidia article the information was however
still quite RSR sown to the coming ATi Grafikchips, which left itself
whole roughly in three paragraphs to summarize.

Remedy created here only an article on the part of AnandTech of last
week, in which the plans were stated in detail by ATi already quite.
We want in the following predominantly to rely on these information on
the part of AnandTech, enweaving thereby however still somewhat own
knowledge - and some substantial details have to in addition-carry for
the classification of the individual diagram chips, which do not admit
to date yet were and from which zumindestens one for more than
plentifully Diskussionstoff will provide: -).

First still briefly to the "class" of the following information: These
are - if not differently characterized - secure information. Both that
confirmed with AnandTech said and our additions us by one called from
first hand informed source and/or.



And thus to the coming main project of ATi, the R420 chip. That one
has - one believes the rumors to this case - a plentifully eventful
history behind itself. Originally the R400-Chip with Shader 3,0
support the chips of the R3x0-Serie should beerben still in the autumn
2003, but allegedly ATi was not really content in the last year with
the all first performance Simulationen of the R400. One can speculate
upon it that the Shader was possibly 3,0 performance correct, however
the Shader 2,0 performance to expectations here did not correspond.
Since however at present and in the near future with
DirectX9-Benchmarks the employment of Shader 3,0 takes time still, one
would have 3,0 with a beautiful technical toy for Shader, but without
the necessary basic performance in the reality with Shader 2,0 stood
there.

So anyhow a possible explanation, why ATi of the R400-Chip in a still
very early Design phase displeased so that one cancelte the project
completely. The useful parts of the R400-Designs for the R420
following the R500-Projekt are alleged were taken over - and for the
meantime, there one not simply after the R350 (Radeon 9800/pro)
nothing new to finally bring could, did not need one a solution which
can be realized fast, which above all the error of the R400 -
technically point, but with the performance less well - did not begin.
One decided in the long run to begin beside the R360-Chip (Radeon
9800XT ) as a further Refresh of the R300-Chips (Radeon 9700/pro ) for
temporal bypass, the R420-Projekt.

The R420, first also admits Loki "as project", should not thereby not
the large technological project like the R400 become, but the proven
technology of the preceding chips R300 (Radeon 9700/pro) , R350
(Radeon 9800/pro) and R360 (Radeon 9800XT) in substantially improved
way to resume, without beginning thereby however a generally new
Design as with the R400. Speak: Primarily clearly more performance,
less however new features.

Became already early for this the substantial point admits, in which
kind ATi wants to realize this: The use of the 130nm production
technology. The preceding chips of the R3x0-Linie all together still
in 150nm manufactured, in particular 412 MHz chip clock of the Radeon
9800XT (nevertheless 107 million transistors heavily) appear with it
as engineer-technical master achievement. By means of the smaller
production technology ATi might know the R420 with ease more highly
clocks, even if this more transistors bring along than the
predecessors from the R3x0-Linie.



The other substantial point resulted already indirectly from leading -
however naturally only on rumors are based - history over the
career/development of the R400 and R420-Chips. Because since the R420
descends generally seen still from the R3x0-Design, it was actually
improbable rather that the R420-Chip Shader 3,0 support could offer.
To a R3x0-Chip the support for Vertex and pixel Shader 3,0 simply "to
to-seal", is not really possible, but the Shader 3,0 to the Shadern
2.0 is too different.

Now, this was the theoretical consideration on basis of the rumor
Story around R400 and R420. In the future this opinion, the R420 did
not turn out could Shader 3,0 support have, however more and more into
oblivion - all world including even S3, XGI and Intel talked on its
Roadmaps finally about the Shader 3,0 support and each rumor kitchen
arranged the R420 generally than "Shader 3,0 chip", so that it became
ever more improbable that this like that is not. But far been missing
- the original theoretical consideration was nevertheless correct,
because: We can confirm free of doubts that the ATi R420-Chip will not
offer Shader 3,0 support.

Now the renouncement of this feature reaches alone still no
performance increase - apart from it that one saves for transistors
and so that only on 160 million with the R420 comes contrary to 175
million with nVidia the NV40. behind this renouncement it stands
nevertheless a clear intention: ATi says for this that the Shader 3,0
will not have such meaning, as the IT world assumes possibly at
present, to insert and that one concentrated therefore with the R420
rather purposefully on the Shader 2,0 performance, instead of with the
Shadern 3,0 a beautiful, but rather useless check list feature.

One does not refer here also to injustice to the Shader 2.X nVidia of
the DirectX9-Chips, which were used only extremely rarely by the play
programmers. Besides the compatibility of the R420 and its successors
endangered by the renouncement of Shader 3,0 is probably hardly,
because already million Shader 2,0 diagram chips were sold in form of
the present DirectX9-Beschleuniger, whose user can probably annoy no
play programmer in the next years by the requirement of Shader 3,0
capable diagram chips. Besides play programmers set gladly and often
on the smallest common denominator - and this reads now once with
DirectX9-Chips Shader 2.0.

Generally seen one must here thus quite give right to ATi: Presently
and also during lifetimes of the R420-Chips simple Shader 3,0 support
in plays could be counted probably still at a hand. Plays, which
Shader presupposes 3,0 directly, besides into the market will probably
only come if the R420 were degradiert long to the LowCost product.
Because straight come at present once the first plays with obligation
to DirectX8-Hardware into the market, plays with obligation to
DirectX9-Hardware (on Shader 2,0 basis) are not even announced our
knowledge to.

With difficulty it only to arrange this for the users and customers
had finally zeroed in itself more or less all on the Shader 3,0
support as the feature of the coming NEXT gene generation of ATi and
nVidia. Also the Shader makes itself 3,0 support surely outstanding as
check list feature in the OEM market, of the manufacturing firms then
gladly also falsely as "DirectX 9,1 support" described (which nonsense
is natural, as here already implemented) . Positive way is not
sufficient however the support of Shader 2,0 also, in order to come
with the Windows XP successor "Longhorn" into the highest possible
hardware class ("animal 2 ") to the representation of the new
3D-Oberflaeche von Longhorn, for this Shader 3,0 support is needed.

Surely the message, which equips ATi the R420 only with Shader 2,0
support, will provide for a violent discussion. ATi made the strategic
decision here to omit the Shader 3,0 support with the R420 in order to
concentrate on the performance under Shader 2,0 applications - these
are those applications, with which the R420 will be confronted during
its lifetimes. From the thing this consideration is correct, but ATi
offers thereby naturally also some attack region for nVidia, which
will evenly offer the Shader 3,0 support with the NV4x-Linie - and
this technological superiority surely also marketing-technically to
use will try.



Off the surely flashing discussion over sense and nonsense of Shader
ATi with the renouncement of these in addition, a clear guideline for
the R420 gave 3,0: This chip is to accumulate under Shader 2,0 to the
absolute maximum form. The chip will have further 8 Rendering
pipelines like its predecessors, but these are to have been
substantially improved. Which hides itself exactly behind it, one did
not want to say us however yet. Several texture units per pipeline
could be (the R3x0-Chips had everything only one texture unit per
pipeline) , it could also internal improvements be, which will then
probably understand rather only the chip experts by us. Zumindestens
is to have energetically worked on the Rendering pipelines, in order
to carry the Shader 2,0 performance on a new level.

Furthermore the number of Vertex Shader of 4 to 6 was increased, which
could give an increase of the Vertex achievement to the R420 as well
as the clock increase and the internal improvements over up to 100 per
cent . In all other respects this increase of the number of Vertex
Shader (the number is not directly comparable in all other respects
with the nVidia chips, which a Vertex Shader array uses) is not enough
out, in order to explain the 50 million more transistors according to
our calculations - here might be in the R420 thus still place for or
other surprise, is it mentioned more texture units or however improved
anti- Aliasing and/or anisotropic filtering the already.

Unfortunately at present still no information is present, which
intends to really change and/or improve ATi with the R420 concerning
anti- Aliasing and anisotropic filter. Off the performance, which
enough is naturally never: -), is the quality of the anti- Aliasings
of the current ATi chips surely already quite optimally , while one
bilinearen however with the quality, trilinear and anisotropic filter
to surely still add can. From it thus momentarily the following can
refrained be said in summary about the R420-Chip:

ATi R420
generell basierend auf den R3x0-Chips, allerdings mit vielen
Verbesserungen an der Shader-Architektur
160 Millionen Transistoren, in 130nm bei TSMC gefertigt
8 Rendering-Pipelines, Anzahl der Textureneinheiten unbekannt
6 Vertex Shader Einheiten
DirectX 9.0 Architektur, unterstützt Shader 2.0
256 Bit Speicherinterface, unterstützt DDR1, GDDR2 und GDDR3
internes AGPx8 Interface
genaue Taktraten: unbekannt; laut AnandTech runde 500 MHz Chip- und
Speichertakt
Präsentation: vermutlich noch erstes Quartal 2004
Markteintritt: zweites Quartal 2004
Die aus dem Artikel von AnandTech stammenden Chip- und Speichertakte
hören sich erst einmal relativ niedrig an, insbesondere wenn man den
NV40 derzeit auf 500 bis 600 MHz Chiptakt und 600 bis 800 MHz
Speichertakt taxiert. Allerdings wären das gegenüber dem R360-Chip
(Radeon 9800XT) immerhin auch schon 21 Prozent mehr Chip- und ganze 37
Prozent mehr Speichertakt, während nVidia bei angenommenen 550/600 MHz
gegenüber dem NV38 (GeForceFX 5950 Ultra) takttechnisch "nur" um 16
bzw. 26 Prozent zulegt.

Zudem kann man davon ausgehen, daß beim Speichertakt definitiv noch
Spielraum nach oben ist, immerhin hat Samsung kürzlich die
Massenproduktion von GDDR3 mit bis zu 800 MHz Taktfrequenz begonnen.
Auch beim Chiptakt sollte aufgrund der 130nm Fertigung eigentlich noch
etwas mehr möglich sein - in diesem Punkt erscheinen uns die genannten
Taktraten doch etwas konservativ angegeben. Wir können uns für unseren
Teil also durchaus vorstellen, daß ATi eine mögliche Pro-Version des
R420 auch höher taktet als jene 500/500 MHz, welche AnandTech genannt
haben.


ATi R420, R423 & CO. Information
22. February 2004/from Leonidas /page 2 of 2



As evident from previous enumerating list, the R420 - like also the
NV40 - start still with an internal AGPx8 interface. In the coming age
of PCI express needs ATi naturally also a PCI express diagram chip
within the HighEnd range, which the R423 will place. That chip becomes
however, then one insured us, a native PCI express x16 interface to
have. This is in as much unusual as that is to appear the R423 in
principle in the same period as the R420 - in the second quarter. Two
160-Millionen-Transistoren-Chips however more or less in the same
period on the mass production to prepare, might be however not only
extremely expensive, but also to the personnel capacities of ATi push.

To that extent we can explain ourselves this at present only in this
kind, when the R420 and R423 are in principle the same chip. Either
the R420 is a R423 with settled AGP Bridge, which converts thus the
PCI express x16 of signals of the diagram chip into AGPx8 signal for
the AGP haven on the Main board, or however - hears themselves
improbably on, is however not not impossible - both internal AGPx8 and
internal PCI express x16 interface, which is then activated depending
upon need by BIOS, have R420 and R423. As this point dissolves in the
long run, remains being waiting however.

Abseites its is considered against the fact as safe that the R423
represents really only a pure PCI express modification of the R420:
There are no technical changes and also not more clock. Only the R480
chip will represent a genuine Refresh to the R420/R423, which will
throw however - after the experiences with the last Refresh chips of
ATi - probably only more clock into the scale pan, possibly already in
the 110nm production technology. The R480 is to begin sometime in the
second half-year 2004, in the third quarter is probably thus
introduced and in the fourth quarter delivered.



And thus we come off the HighEnd chips to the candidates for the
Mainstream and LowCost range. AnandTech have to offer here already
beautiful technical information, we can besides again the appropriate
classification of the chips offer. In the second quarter it will give
only once the RV380 to chip, which one can charaktersisieren briefly
in such a way:

ATi RV380
direct descent of RV350/RV360, invariably internally in principle
75 million transistors, in 130nm with TSMC manufactured
4 Rendering pipelines with ever a texture unit
2 Vertex Shader units
DirectX 9,0 architecture, supports Shader 2.0
, DDR1, GDDR2 and GDDR3 supports 128 bits memory interface
internal PCI express interface
exact clock rates: unknown; according to AnandTech round 500 MHz chip
and 400 MHz storing act
Presentation: probably still first quarter 2004
Market entrance: second quarter 2004
It is not correct that AnandTech indicate the same improvements on
architecture here for the RV380 as for the R420 have. The RV380 is our
knowledge after only a further Refresh of the original RV350-Chips
(Radeon 9600/pro/cSe) , which already found a first Refresh in the
RV360-Chip (Radeon 9600XT). With the RV380 apparent at the chip clock
one does not turn, which however with 500 MHz enough is already
sufficiently high, but alone the storing act raised, which in view of
the unfavorable relationship between chip and storing act with the
direct predecessor Radeon 9600XT (500/300 MHz) might be also the best
way to give to this chip more performance.

Internally the RV380-Chip thereby might be perfectly unchanged to the
RV350 and/or RV360, however the native PCI express interface would be
new. ATi goes here not the way of nVidia with the PCI express Bridge
chip, but presents the chip specially for PCI express again. The
reason for this lies in the universe in Wonder maps and their TV
abilities: For the use and treatment of in America momentarily heavily
in coming HDTV present the backward channel is needed by PCI express
x16, which nevertheless 4 GB/sec carry out can - and this
independently of the Hinkanal, which likewise 4 GB/ses width exhibits.
Because ATi for its universe in Wonder maps wants to use this
potential, one needs native PCI express interface and therefore to
Bridge chips, how she begins nVidia, a general refusal gave.

It will really give one of the R420 to descending Mainstream chip only
with the RV410 in the second half-year 2004 (the indication of
AnandTech "Q2'04" is surely a write error) . This then also the
improvements on the architecture of the R420 into the Mainstream
market to carry, wommoeglich opposite the RV380 again more clock bring
along and completely possibly already in 110nm manufactured its -
which applies to however naturally wait for it.



For the LowCost market ATi for the second quarter positioned the RV370
chip, which will be according to AnandTech the first diagram chip in
110nm production technology and can be characterized as follows:

ATi RV370
complete (?) new development for the LowCost market
unknown number of transistors, in 110nm with TSMC manufactured
4 Rendering pipelines, number of texture units unknown (probably per
one)
2 Vertex Shader units
DirectX 9,0 architecture, supports Shader 2.0
64/128 bit memory interface, supports DDR1, GDDR2 and GDDR3
internal PCI express interface
exact clock rates: unknown; according to AnandTech round 300 MHz
storing act
Presentation: probably still first quarter 2004
Market entrance: second quarter 2004
The objective of ATi with the RV370-Chip is clear thereby: Finally a
competitive chip for the DirectX9 OEM market. So far nVidia in this
rough sells quantities of GeForceFX 5200 chips, partly also with only
64 bits memory interface, at prices, with which ATi only with the own
DirectX8-Grafikchips Paroli offer can. Problematic way is however
calculated this OEM market a market, where it depends much on check
list features like "DirectX9 support", with which ATi cannot score in
the long run with a Radeon 9200 despite possibly better performance
opposite a GeForceFX 5200.

Differently around the RV350-Chip is even in its lowest form, the
Radeon 9600SE diagram map not to get so cheaply than the fact that ATi
the prices, which nVidia for the GeForceFX 5200 makes, could reach.
The reason lies here by the majority in the transistor quantity and/or
the origin of the chips: The RV350 was now once for the Mainstream
segment intended the NV34-Chip (GeForceFX 5200/Ultra & GeForceFX 5500)
and weighs nevertheless 75 million transistors, during nVidia
specially only for the LowCost segment developed and by various
reductions the number of transistors on 45 million to lower could.

ATi will hit here in our view same way with the RV370. Surely one
could manufacture also simply a RV350 in 110nm, but one will achieve
most piercing success with a clear lowering of the number of
transistors. Momentarily this point is only, safe of speculation is
however the 110nm manufacturing and the adjustment as GeForceFX
5200/5500 competitor to down into the absolute LowCost market. For
genuine players such diagram maps should be naturally no view worth,
but for ATi as nVidia they also represent the lion's share of the
business - and also this field, where in the long run on basis of the
high numbers of items set off there profits are erwirtschaftbar.



Thus still another one point remains: As already mentions, ATi gave
all solutions with Bridge chips a general refusal, since one for the
own universe in Wonder maps native PCI express interface necessarily
and there temporary solutions with Bridge chips do not lead to the
desired result. This refusal refers however also to the past AGPx8
maps of ATi, speaks Radeon 9800/Pro/XT, Radeon 9600/Pro/XT/SE and
Radeon 9200/pro/cSe. This does not want to offer ATi in accordance
with current planning on PCI to express - in contrast to it nVidia
well-known-measured its complete present line places at diagram maps
still on PCI express over, before the new chips come.

With ATi this may be connected surely also with the fact that one
wants to bring new diagram chips in the second quarter in all partial
markets into the market, while with nVidia the new Mainstream is to be
expected as also LowCost solutions rather only in the third quarter.
Nothing the despite surprises this step of ATi something, because the
need at the predecessor hardware does not even shrink according to
experience over night with the introduction on the market of new
diagram maps. Speak: It will quite give users, which for a new PCI
express Main board possibly a Radeon 9800 express wish themselves pro
on PCI (which could be realized by means of a Bridge chip
problem-free) . It remains hoping that the diagram board manufacturers
will fill to substitute accordingly here and these by ATi somewhat
left open market gap.

There itself in the article with nVidia the NV40, NV41, NV45 & CO.
Information still another (naturally prospective) nVidia Roadmap
finds, wants we also this article with such to the coming ATi
Grafikchips to lock. It would pay attention that the individual chips
are arranged after their market entrance like already with the nVidia
Roadmap after their supposed presentation time and not:





After the R4x0/RV4x0-Linie it will continue with ATi with the R500
chip in the year 2005. This will support DirectX10, which covers among
other things the support of the Shader 4,0, ATi jumps over the Shader
3,0 thus completely. More details on this chip, which already is in
development, is not well-known, just as little relative to the further
plans of ATi however yet from R500-Abkoemmlingen for the Mainstream
such as LowCost market.
  #2  
Old February 22nd 04, 11:49 PM
Dark Avenger
external usenet poster
 
Posts: n/a
Default

* allot of snip snip *

Hehehe, nvidia would just WISH!

WE shall se.. we shall see!
  #3  
Old February 23rd 04, 12:56 AM
K
external usenet poster
 
Posts: n/a
Default


"Dark Avenger" wrote in message
om...
* allot of snip snip *

Hehehe, nvidia would just WISH!

WE shall se.. we shall see!


Well as an ATI owner, if Nvidia can produce cards that are the best value
for money with great performance I would switch back to them like a shot.
ATI's drivers have come a long way but they still have a number of
annoyances, especially in games that are not so well known.

Also Nvidia is still the king of OpenGL and Linux. ATI's Linux drivers
*really* suck.

K


  #4  
Old February 23rd 04, 12:59 AM
NightSky 421
external usenet poster
 
Posts: n/a
Default

"NV55" wrote in message
om...
Bad news for ATI: Nvidia to 'own' ATI at CeBit / no pixel shader 3.0
support in R420

It looks like R420, while potentially up to 3 times faster(in PS 2.0
and VS 2.0 shaders) than R360, will not be PS 3.0 compliant. If NVIDIA
notches up similar PS 2.0 performance, ATI is in trouble.



If what you are saying turns out to be true, I'm still not particularly
concerned. I know that my current 9800 Pro card will carry me over nicely
until the second half of 2005. At that time, I'll see what video card
technology in general is like and how the driver and public relations
departments of both companies are behaving and base my upgrade decision on
those factors.


  #5  
Old February 23rd 04, 12:59 AM
John Reynolds
external usenet poster
 
Posts: n/a
Default

"NV55" wrote in message
om...
Bad news for ATI: Nvidia to 'own' ATI at CeBit / no pixel shader 3.0
support in R420

It looks like R420, while potentially up to 3 times faster(in PS 2.0
and VS 2.0 shaders) than R360, will not be PS 3.0 compliant. If NVIDIA
notches up similar PS 2.0 performance, ATI is in trouble.


Allow me to suggest something. One of the main reasons cited for NV30's
lackluster performance was because of NVIDIA's prior focus on the X-box. .
..the engineering resources required for that sapped them for their next
generation PC part. ATI's movement of R400 from late last year to the X-Box
2/R500 could have been to pre-empt the fact that they are going to have to
divert resources to meet the X-Box deadlines, so rather than take the hit
later they are doing it now. In other words, release the next part (R420)
with a lower specification that's build primarily upon their existing
architecture and therefore requires less engineering time and resources to
finish, and where the key inflection point isn't going to be architectural
abilities (i.e. Shader 3.0 is an interim spec) but a relatively small change
to PCI-Express, while ensuring that the technology isn't lacking after the
X-Box is released. In short, the hit that NVIDIA took post-X-Box, ATI are
trying to minimize by taking it pre-X-box2.

Just a thought. And I don't know NV40 or R420 specs.

John


  #6  
Old February 23rd 04, 04:19 AM
Andrew MacPherson
external usenet poster
 
Posts: n/a
Default

In article ,
(K) wrote:

Well as an ATI owner, if Nvidia can produce cards that are the best
value for money with great performance I would switch back to them
like a shot.


Agreed, but what I'm hoping for more than raw speed is that they've
finally used those 3dfx engineers to generate FSAA to match or beat the
V5. I moved to ATI for a combination of speed *with* superior FSAA (which
makes a huge difference to flight and driving sims IMO). But I have no
loyalty to any hardware supplier... all I want is good kit at a fair
price.

Andrew McP

  #7  
Old February 23rd 04, 08:07 AM
Minotaur
external usenet poster
 
Posts: n/a
Default

LOL, But nothing shall use it! PS3. As everyone shall be working on DX10
and PS4 as you have mentioned. Also, ATI shall own nVidia, because they
are supplying the GFX for the XBox2 and developers shall be focusing on
what it can do, not what nVidia could have done. Why is the XBox2
important? Because most games shall be ported between the XBox2 and PC
as it is happening now with the XBox.

Just like how Halo was a crap port to the PC, for ATI owners because it
was tweaked for nVidia.
In the future games ported from the XBox2 to the PC, shall be optimised
for ATI.

NV55 wrote:

Bad news for ATI: Nvidia to 'own' ATI at CeBit / no pixel shader 3.0
support in R420

It looks like R420, while potentially up to 3 times faster(in PS 2.0
and VS 2.0 shaders) than R360, will not be PS 3.0 compliant. If NVIDIA
notches up similar PS 2.0 performance, ATI is in trouble.




quote:
--------------------------------------------------------------------------------
Lastly there is word on the NV40 R420 front. Neither card was shown,
but several people who would be familiar with both have said Nvidia is
going to come back very strong soon. The word 'own' was bandied about,
as in Nvidia will own ATI. Not sure what it means, but CeBit should be
a very interesting show.
--------------------------------------------------------------------------------



Babelfish translation of the 'no PS2.0 shaders in R420' article:




quote:
--------------------------------------------------------------------------------
ATi R420, R423 & CO. Information
22. February 2004/from Leonidas /page 1 of 2



After we could give at the end of January an already rather complete
overview of the coming nVidia Grafikchips, naturally the question was
located in the area, which is to be expected 2004 now from ATi for the
spring. At the time of the nVidia article the information was however
still quite RSR sown to the coming ATi Grafikchips, which left itself
whole roughly in three paragraphs to summarize.

Remedy created here only an article on the part of AnandTech of last
week, in which the plans were stated in detail by ATi already quite.
We want in the following predominantly to rely on these information on
the part of AnandTech, enweaving thereby however still somewhat own
knowledge - and some substantial details have to in addition-carry for
the classification of the individual diagram chips, which do not admit
to date yet were and from which zumindestens one for more than
plentifully Diskussionstoff will provide: -).

First still briefly to the "class" of the following information: These
are - if not differently characterized - secure information. Both that
confirmed with AnandTech said and our additions us by one called from
first hand informed source and/or.



And thus to the coming main project of ATi, the R420 chip. That one
has - one believes the rumors to this case - a plentifully eventful
history behind itself. Originally the R400-Chip with Shader 3,0
support the chips of the R3x0-Serie should beerben still in the autumn
2003, but allegedly ATi was not really content in the last year with
the all first performance Simulationen of the R400. One can speculate
upon it that the Shader was possibly 3,0 performance correct, however
the Shader 2,0 performance to expectations here did not correspond.
Since however at present and in the near future with
DirectX9-Benchmarks the employment of Shader 3,0 takes time still, one
would have 3,0 with a beautiful technical toy for Shader, but without
the necessary basic performance in the reality with Shader 2,0 stood
there.

So anyhow a possible explanation, why ATi of the R400-Chip in a still
very early Design phase displeased so that one cancelte the project
completely. The useful parts of the R400-Designs for the R420
following the R500-Projekt are alleged were taken over - and for the
meantime, there one not simply after the R350 (Radeon 9800/pro)
nothing new to finally bring could, did not need one a solution which
can be realized fast, which above all the error of the R400 -
technically point, but with the performance less well - did not begin.
One decided in the long run to begin beside the R360-Chip (Radeon
9800XT ) as a further Refresh of the R300-Chips (Radeon 9700/pro ) for
temporal bypass, the R420-Projekt.

The R420, first also admits Loki "as project", should not thereby not
the large technological project like the R400 become, but the proven
technology of the preceding chips R300 (Radeon 9700/pro) , R350
(Radeon 9800/pro) and R360 (Radeon 9800XT) in substantially improved
way to resume, without beginning thereby however a generally new
Design as with the R400. Speak: Primarily clearly more performance,
less however new features.

Became already early for this the substantial point admits, in which
kind ATi wants to realize this: The use of the 130nm production
technology. The preceding chips of the R3x0-Linie all together still
in 150nm manufactured, in particular 412 MHz chip clock of the Radeon
9800XT (nevertheless 107 million transistors heavily) appear with it
as engineer-technical master achievement. By means of the smaller
production technology ATi might know the R420 with ease more highly
clocks, even if this more transistors bring along than the
predecessors from the R3x0-Linie.



The other substantial point resulted already indirectly from leading -
however naturally only on rumors are based - history over the
career/development of the R400 and R420-Chips. Because since the R420
descends generally seen still from the R3x0-Design, it was actually
improbable rather that the R420-Chip Shader 3,0 support could offer.
To a R3x0-Chip the support for Vertex and pixel Shader 3,0 simply "to
to-seal", is not really possible, but the Shader 3,0 to the Shadern
2.0 is too different.

Now, this was the theoretical consideration on basis of the rumor
Story around R400 and R420. In the future this opinion, the R420 did
not turn out could Shader 3,0 support have, however more and more into
oblivion - all world including even S3, XGI and Intel talked on its
Roadmaps finally about the Shader 3,0 support and each rumor kitchen
arranged the R420 generally than "Shader 3,0 chip", so that it became
ever more improbable that this like that is not. But far been missing
- the original theoretical consideration was nevertheless correct,
because: We can confirm free of doubts that the ATi R420-Chip will not
offer Shader 3,0 support.

Now the renouncement of this feature reaches alone still no
performance increase - apart from it that one saves for transistors
and so that only on 160 million with the R420 comes contrary to 175
million with nVidia the NV40. behind this renouncement it stands
nevertheless a clear intention: ATi says for this that the Shader 3,0
will not have such meaning, as the IT world assumes possibly at
present, to insert and that one concentrated therefore with the R420
rather purposefully on the Shader 2,0 performance, instead of with the
Shadern 3,0 a beautiful, but rather useless check list feature.

One does not refer here also to injustice to the Shader 2.X nVidia of
the DirectX9-Chips, which were used only extremely rarely by the play
programmers. Besides the compatibility of the R420 and its successors
endangered by the renouncement of Shader 3,0 is probably hardly,
because already million Shader 2,0 diagram chips were sold in form of
the present DirectX9-Beschleuniger, whose user can probably annoy no
play programmer in the next years by the requirement of Shader 3,0
capable diagram chips. Besides play programmers set gladly and often
on the smallest common denominator - and this reads now once with
DirectX9-Chips Shader 2.0.

Generally seen one must here thus quite give right to ATi: Presently
and also during lifetimes of the R420-Chips simple Shader 3,0 support
in plays could be counted probably still at a hand. Plays, which
Shader presupposes 3,0 directly, besides into the market will probably
only come if the R420 were degradiert long to the LowCost product.
Because straight come at present once the first plays with obligation
to DirectX8-Hardware into the market, plays with obligation to
DirectX9-Hardware (on Shader 2,0 basis) are not even announced our
knowledge to.

With difficulty it only to arrange this for the users and customers
had finally zeroed in itself more or less all on the Shader 3,0
support as the feature of the coming NEXT gene generation of ATi and
nVidia. Also the Shader makes itself 3,0 support surely outstanding as
check list feature in the OEM market, of the manufacturing firms then
gladly also falsely as "DirectX 9,1 support" described (which nonsense
is natural, as here already implemented) . Positive way is not
sufficient however the support of Shader 2,0 also, in order to come
with the Windows XP successor "Longhorn" into the highest possible
hardware class ("animal 2 ") to the representation of the new
3D-Oberflaeche von Longhorn, for this Shader 3,0 support is needed.

Surely the message, which equips ATi the R420 only with Shader 2,0
support, will provide for a violent discussion. ATi made the strategic
decision here to omit the Shader 3,0 support with the R420 in order to
concentrate on the performance under Shader 2,0 applications - these
are those applications, with which the R420 will be confronted during
its lifetimes. From the thing this consideration is correct, but ATi
offers thereby naturally also some attack region for nVidia, which
will evenly offer the Shader 3,0 support with the NV4x-Linie - and
this technological superiority surely also marketing-technically to
use will try.



Off the surely flashing discussion over sense and nonsense of Shader
ATi with the renouncement of these in addition, a clear guideline for
the R420 gave 3,0: This chip is to accumulate under Shader 2,0 to the
absolute maximum form. The chip will have further 8 Rendering
pipelines like its predecessors, but these are to have been
substantially improved. Which hides itself exactly behind it, one did
not want to say us however yet. Several texture units per pipeline
could be (the R3x0-Chips had everything only one texture unit per
pipeline) , it could also internal improvements be, which will then
probably understand rather only the chip experts by us. Zumindestens
is to have energetically worked on the Rendering pipelines, in order
to carry the Shader 2,0 performance on a new level.

Furthermore the number of Vertex Shader of 4 to 6 was increased, which
could give an increase of the Vertex achievement to the R420 as well
as the clock increase and the internal improvements over up to 100 per
cent . In all other respects this increase of the number of Vertex
Shader (the number is not directly comparable in all other respects
with the nVidia chips, which a Vertex Shader array uses) is not enough
out, in order to explain the 50 million more transistors according to
our calculations - here might be in the R420 thus still place for or
other surprise, is it mentioned more texture units or however improved
anti- Aliasing and/or anisotropic filtering the already.

Unfortunately at present still no information is present, which
intends to really change and/or improve ATi with the R420 concerning
anti- Aliasing and anisotropic filter. Off the performance, which
enough is naturally never: -), is the quality of the anti- Aliasings
of the current ATi chips surely already quite optimally , while one
bilinearen however with the quality, trilinear and anisotropic filter
to surely still add can. From it thus momentarily the following can
refrained be said in summary about the R420-Chip:

ATi R420
generell basierend auf den R3x0-Chips, allerdings mit vielen
Verbesserungen an der Shader-Architektur
160 Millionen Transistoren, in 130nm bei TSMC gefertigt
8 Rendering-Pipelines, Anzahl der Textureneinheiten unbekannt
6 Vertex Shader Einheiten
DirectX 9.0 Architektur, unterstützt Shader 2.0
256 Bit Speicherinterface, unterstützt DDR1, GDDR2 und GDDR3
internes AGPx8 Interface
genaue Taktraten: unbekannt; laut AnandTech runde 500 MHz Chip- und
Speichertakt
Präsentation: vermutlich noch erstes Quartal 2004
Markteintritt: zweites Quartal 2004
Die aus dem Artikel von AnandTech stammenden Chip- und Speichertakte
hören sich erst einmal relativ niedrig an, insbesondere wenn man den
NV40 derzeit auf 500 bis 600 MHz Chiptakt und 600 bis 800 MHz
Speichertakt taxiert. Allerdings wären das gegenüber dem R360-Chip
(Radeon 9800XT) immerhin auch schon 21 Prozent mehr Chip- und ganze 37
Prozent mehr Speichertakt, während nVidia bei angenommenen 550/600 MHz
gegenüber dem NV38 (GeForceFX 5950 Ultra) takttechnisch "nur" um 16
bzw. 26 Prozent zulegt.

Zudem kann man davon ausgehen, daß beim Speichertakt definitiv noch
Spielraum nach oben ist, immerhin hat Samsung kürzlich die
Massenproduktion von GDDR3 mit bis zu 800 MHz Taktfrequenz begonnen.
Auch beim Chiptakt sollte aufgrund der 130nm Fertigung eigentlich noch
etwas mehr möglich sein - in diesem Punkt erscheinen uns die genannten
Taktraten doch etwas konservativ angegeben. Wir können uns für unseren
Teil also durchaus vorstellen, daß ATi eine mögliche Pro-Version des
R420 auch höher taktet als jene 500/500 MHz, welche AnandTech genannt
haben.


ATi R420, R423 & CO. Information
22. February 2004/from Leonidas /page 2 of 2



As evident from previous enumerating list, the R420 - like also the
NV40 - start still with an internal AGPx8 interface. In the coming age
of PCI express needs ATi naturally also a PCI express diagram chip
within the HighEnd range, which the R423 will place. That chip becomes
however, then one insured us, a native PCI express x16 interface to
have. This is in as much unusual as that is to appear the R423 in
principle in the same period as the R420 - in the second quarter. Two
160-Millionen-Transistoren-Chips however more or less in the same
period on the mass production to prepare, might be however not only
extremely expensive, but also to the personnel capacities of ATi push.

To that extent we can explain ourselves this at present only in this
kind, when the R420 and R423 are in principle the same chip. Either
the R420 is a R423 with settled AGP Bridge, which converts thus the
PCI express x16 of signals of the diagram chip into AGPx8 signal for
the AGP haven on the Main board, or however - hears themselves
improbably on, is however not not impossible - both internal AGPx8 and
internal PCI express x16 interface, which is then activated depending
upon need by BIOS, have R420 and R423. As this point dissolves in the
long run, remains being waiting however.

Abseites its is considered against the fact as safe that the R423
represents really only a pure PCI express modification of the R420:
There are no technical changes and also not more clock. Only the R480
chip will represent a genuine Refresh to the R420/R423, which will
throw however - after the experiences with the last Refresh chips of
ATi - probably only more clock into the scale pan, possibly already in
the 110nm production technology. The R480 is to begin sometime in the
second half-year 2004, in the third quarter is probably thus
introduced and in the fourth quarter delivered.



And thus we come off the HighEnd chips to the candidates for the
Mainstream and LowCost range. AnandTech have to offer here already
beautiful technical information, we can besides again the appropriate
classification of the chips offer. In the second quarter it will give
only once the RV380 to chip, which one can charaktersisieren briefly
in such a way:

ATi RV380
direct descent of RV350/RV360, invariably internally in principle
75 million transistors, in 130nm with TSMC manufactured
4 Rendering pipelines with ever a texture unit
2 Vertex Shader units
DirectX 9,0 architecture, supports Shader 2.0
, DDR1, GDDR2 and GDDR3 supports 128 bits memory interface
internal PCI express interface
exact clock rates: unknown; according to AnandTech round 500 MHz chip
and 400 MHz storing act
Presentation: probably still first quarter 2004
Market entrance: second quarter 2004
It is not correct that AnandTech indicate the same improvements on
architecture here for the RV380 as for the R420 have. The RV380 is our
knowledge after only a further Refresh of the original RV350-Chips
(Radeon 9600/pro/cSe) , which already found a first Refresh in the
RV360-Chip (Radeon 9600XT). With the RV380 apparent at the chip clock
one does not turn, which however with 500 MHz enough is already
sufficiently high, but alone the storing act raised, which in view of
the unfavorable relationship between chip and storing act with the
direct predecessor Radeon 9600XT (500/300 MHz) might be also the best
way to give to this chip more performance.

Internally the RV380-Chip thereby might be perfectly unchanged to the
RV350 and/or RV360, however the native PCI express interface would be
new. ATi goes here not the way of nVidia with the PCI express Bridge
chip, but presents the chip specially for PCI express again. The
reason for this lies in the universe in Wonder maps and their TV
abilities: For the use and treatment of in America momentarily heavily
in coming HDTV present the backward channel is needed by PCI express
x16, which nevertheless 4 GB/sec carry out can - and this
independently of the Hinkanal, which likewise 4 GB/ses width exhibits.
Because ATi for its universe in Wonder maps wants to use this
potential, one needs native PCI express interface and therefore to
Bridge chips, how she begins nVidia, a general refusal gave.

It will really give one of the R420 to descending Mainstream chip only
with the RV410 in the second half-year 2004 (the indication of
AnandTech "Q2'04" is surely a write error) . This then also the
improvements on the architecture of the R420 into the Mainstream
market to carry, wommoeglich opposite the RV380 again more clock bring
along and completely possibly already in 110nm manufactured its -
which applies to however naturally wait for it.



For the LowCost market ATi for the second quarter positioned the RV370
chip, which will be according to AnandTech the first diagram chip in
110nm production technology and can be characterized as follows:

ATi RV370
complete (?) new development for the LowCost market
unknown number of transistors, in 110nm with TSMC manufactured
4 Rendering pipelines, number of texture units unknown (probably per
one)
2 Vertex Shader units
DirectX 9,0 architecture, supports Shader 2.0
64/128 bit memory interface, supports DDR1, GDDR2 and GDDR3
internal PCI express interface
exact clock rates: unknown; according to AnandTech round 300 MHz
storing act
Presentation: probably still first quarter 2004
Market entrance: second quarter 2004
The objective of ATi with the RV370-Chip is clear thereby: Finally a
competitive chip for the DirectX9 OEM market. So far nVidia in this
rough sells quantities of GeForceFX 5200 chips, partly also with only
64 bits memory interface, at prices, with which ATi only with the own
DirectX8-Grafikchips Paroli offer can. Problematic way is however
calculated this OEM market a market, where it depends much on check
list features like "DirectX9 support", with which ATi cannot score in
the long run with a Radeon 9200 despite possibly better performance
opposite a GeForceFX 5200.

Differently around the RV350-Chip is even in its lowest form, the
Radeon 9600SE diagram map not to get so cheaply than the fact that ATi
the prices, which nVidia for the GeForceFX 5200 makes, could reach.
The reason lies here by the majority in the transistor quantity and/or
the origin of the chips: The RV350 was now once for the Mainstream
segment intended the NV34-Chip (GeForceFX 5200/Ultra & GeForceFX 5500)
and weighs nevertheless 75 million transistors, during nVidia
specially only for the LowCost segment developed and by various
reductions the number of transistors on 45 million to lower could.

ATi will hit here in our view same way with the RV370. Surely one
could manufacture also simply a RV350 in 110nm, but one will achieve
most piercing success with a clear lowering of the number of
transistors. Momentarily this point is only, safe of speculation is
however the 110nm manufacturing and the adjustment as GeForceFX
5200/5500 competitor to down into the absolute LowCost market. For
genuine players such diagram maps should be naturally no view worth,
but for ATi as nVidia they also represent the lion's share of the
business - and also this field, where in the long run on basis of the
high numbers of items set off there profits are erwirtschaftbar.



Thus still another one point remains: As already mentions, ATi gave
all solutions with Bridge chips a general refusal, since one for the
own universe in Wonder maps native PCI express interface necessarily
and there temporary solutions with Bridge chips do not lead to the
desired result. This refusal refers however also to the past AGPx8
maps of ATi, speaks Radeon 9800/Pro/XT, Radeon 9600/Pro/XT/SE and
Radeon 9200/pro/cSe. This does not want to offer ATi in accordance
with current planning on PCI to express - in contrast to it nVidia
well-known-measured its complete present line places at diagram maps
still on PCI express over, before the new chips come.

With ATi this may be connected surely also with the fact that one
wants to bring new diagram chips in the second quarter in all partial
markets into the market, while with nVidia the new Mainstream is to be
expected as also LowCost solutions rather only in the third quarter.
Nothing the despite surprises this step of ATi something, because the
need at the predecessor hardware does not even shrink according to
experience over night with the introduction on the market of new
diagram maps. Speak: It will quite give users, which for a new PCI
express Main board possibly a Radeon 9800 express wish themselves pro
on PCI (which could be realized by means of a Bridge chip
problem-free) . It remains hoping that the diagram board manufacturers
will fill to substitute accordingly here and these by ATi somewhat
left open market gap.

There itself in the article with nVidia the NV40, NV41, NV45 & CO.
Information still another (naturally prospective) nVidia Roadmap
finds, wants we also this article with such to the coming ATi
Grafikchips to lock. It would pay attention that the individual chips
are arranged after their market entrance like already with the nVidia
Roadmap after their supposed presentation time and not:





After the R4x0/RV4x0-Linie it will continue with ATi with the R500
chip in the year 2005. This will support DirectX10, which covers among
other things the support of the Shader 4,0, ATi jumps over the Shader
3,0 thus completely. More details on this chip, which already is in
development, is not well-known, just as little relative to the further
plans of ATi however yet from R500-Abkoemmlingen for the Mainstream
such as LowCost market.

  #8  
Old February 23rd 04, 01:50 PM
DD
external usenet poster
 
Posts: n/a
Default


"Minotaur" wrote in message
...
LOL, But nothing shall use it! PS3. As everyone shall be working on DX10
and PS4 as you have mentioned. Also, ATI shall own nVidia, because they
are supplying the GFX for the XBox2 and developers shall be focusing on
what it can do, not what nVidia could have done. Why is the XBox2
important? Because most games shall be ported between the XBox2 and PC
as it is happening now with the XBox.


LMAO your posts get funnier and funner with each one I read.

Ports notwithstanding, console gaming and PC gaming are two completely
different experiences, and consoles will never beat PCs for screen
resolution. As for X-Box, Microsoft's little overpriced half-computer,
half-console, it was and still is the biggest joke in electronic
entertainment since the Intellivision. It only has decent market share in
the USA, the global market spits on it, and will be happy to see Microsoft
give up on it when the new machine fails to corner the console market once
again.

Just like how Halo was a crap port to the PC, for ATI owners because it
was tweaked for nVidia.
In the future games ported from the XBox2 to the PC, shall be optimised
for ATI.


Ported games suck, plain and simple, consoles games and PC games just don't
make the transition well, optimised or not.

DD
\


  #9  
Old February 23rd 04, 02:39 PM
JTS
external usenet poster
 
Posts: n/a
Default

Good news for consumers!!

If Nvidia makes a better care, hopefully it will force ATI to top Nvidia.
Nothing like good old fashioned competition that will ultimately benefit the
consumer.

Todd

"NV55" wrote in message
om...
Bad news for ATI: Nvidia to 'own' ATI at CeBit / no pixel shader 3.0
support in R420

It looks like R420, while potentially up to 3 times faster(in PS 2.0
and VS 2.0 shaders) than R360, will not be PS 3.0 compliant. If NVIDIA
notches up similar PS 2.0 performance, ATI is in trouble.




quote:
--------------------------------------------------------------------------

------
Lastly there is word on the NV40 R420 front. Neither card was shown,
but several people who would be familiar with both have said Nvidia is
going to come back very strong soon. The word 'own' was bandied about,
as in Nvidia will own ATI. Not sure what it means, but CeBit should be
a very interesting show.
--------------------------------------------------------------------------

------



Babelfish translation of the 'no PS2.0 shaders in R420' article:




quote:
--------------------------------------------------------------------------

------
ATi R420, R423 & CO. Information
22. February 2004/from Leonidas /page 1 of 2



After we could give at the end of January an already rather complete
overview of the coming nVidia Grafikchips, naturally the question was
located in the area, which is to be expected 2004 now from ATi for the
spring. At the time of the nVidia article the information was however
still quite RSR sown to the coming ATi Grafikchips, which left itself
whole roughly in three paragraphs to summarize.

Remedy created here only an article on the part of AnandTech of last
week, in which the plans were stated in detail by ATi already quite.
We want in the following predominantly to rely on these information on
the part of AnandTech, enweaving thereby however still somewhat own
knowledge - and some substantial details have to in addition-carry for
the classification of the individual diagram chips, which do not admit
to date yet were and from which zumindestens one for more than
plentifully Diskussionstoff will provide: -).

First still briefly to the "class" of the following information: These
are - if not differently characterized - secure information. Both that
confirmed with AnandTech said and our additions us by one called from
first hand informed source and/or.



And thus to the coming main project of ATi, the R420 chip. That one
has - one believes the rumors to this case - a plentifully eventful
history behind itself. Originally the R400-Chip with Shader 3,0
support the chips of the R3x0-Serie should beerben still in the autumn
2003, but allegedly ATi was not really content in the last year with
the all first performance Simulationen of the R400. One can speculate
upon it that the Shader was possibly 3,0 performance correct, however
the Shader 2,0 performance to expectations here did not correspond.
Since however at present and in the near future with
DirectX9-Benchmarks the employment of Shader 3,0 takes time still, one
would have 3,0 with a beautiful technical toy for Shader, but without
the necessary basic performance in the reality with Shader 2,0 stood
there.

So anyhow a possible explanation, why ATi of the R400-Chip in a still
very early Design phase displeased so that one cancelte the project
completely. The useful parts of the R400-Designs for the R420
following the R500-Projekt are alleged were taken over - and for the
meantime, there one not simply after the R350 (Radeon 9800/pro)
nothing new to finally bring could, did not need one a solution which
can be realized fast, which above all the error of the R400 -
technically point, but with the performance less well - did not begin.
One decided in the long run to begin beside the R360-Chip (Radeon
9800XT ) as a further Refresh of the R300-Chips (Radeon 9700/pro ) for
temporal bypass, the R420-Projekt.

The R420, first also admits Loki "as project", should not thereby not
the large technological project like the R400 become, but the proven
technology of the preceding chips R300 (Radeon 9700/pro) , R350
(Radeon 9800/pro) and R360 (Radeon 9800XT) in substantially improved
way to resume, without beginning thereby however a generally new
Design as with the R400. Speak: Primarily clearly more performance,
less however new features.

Became already early for this the substantial point admits, in which
kind ATi wants to realize this: The use of the 130nm production
technology. The preceding chips of the R3x0-Linie all together still
in 150nm manufactured, in particular 412 MHz chip clock of the Radeon
9800XT (nevertheless 107 million transistors heavily) appear with it
as engineer-technical master achievement. By means of the smaller
production technology ATi might know the R420 with ease more highly
clocks, even if this more transistors bring along than the
predecessors from the R3x0-Linie.



The other substantial point resulted already indirectly from leading -
however naturally only on rumors are based - history over the
career/development of the R400 and R420-Chips. Because since the R420
descends generally seen still from the R3x0-Design, it was actually
improbable rather that the R420-Chip Shader 3,0 support could offer.
To a R3x0-Chip the support for Vertex and pixel Shader 3,0 simply "to
to-seal", is not really possible, but the Shader 3,0 to the Shadern
2.0 is too different.

Now, this was the theoretical consideration on basis of the rumor
Story around R400 and R420. In the future this opinion, the R420 did
not turn out could Shader 3,0 support have, however more and more into
oblivion - all world including even S3, XGI and Intel talked on its
Roadmaps finally about the Shader 3,0 support and each rumor kitchen
arranged the R420 generally than "Shader 3,0 chip", so that it became
ever more improbable that this like that is not. But far been missing
- the original theoretical consideration was nevertheless correct,
because: We can confirm free of doubts that the ATi R420-Chip will not
offer Shader 3,0 support.

Now the renouncement of this feature reaches alone still no
performance increase - apart from it that one saves for transistors
and so that only on 160 million with the R420 comes contrary to 175
million with nVidia the NV40. behind this renouncement it stands
nevertheless a clear intention: ATi says for this that the Shader 3,0
will not have such meaning, as the IT world assumes possibly at
present, to insert and that one concentrated therefore with the R420
rather purposefully on the Shader 2,0 performance, instead of with the
Shadern 3,0 a beautiful, but rather useless check list feature.

One does not refer here also to injustice to the Shader 2.X nVidia of
the DirectX9-Chips, which were used only extremely rarely by the play
programmers. Besides the compatibility of the R420 and its successors
endangered by the renouncement of Shader 3,0 is probably hardly,
because already million Shader 2,0 diagram chips were sold in form of
the present DirectX9-Beschleuniger, whose user can probably annoy no
play programmer in the next years by the requirement of Shader 3,0
capable diagram chips. Besides play programmers set gladly and often
on the smallest common denominator - and this reads now once with
DirectX9-Chips Shader 2.0.

Generally seen one must here thus quite give right to ATi: Presently
and also during lifetimes of the R420-Chips simple Shader 3,0 support
in plays could be counted probably still at a hand. Plays, which
Shader presupposes 3,0 directly, besides into the market will probably
only come if the R420 were degradiert long to the LowCost product.
Because straight come at present once the first plays with obligation
to DirectX8-Hardware into the market, plays with obligation to
DirectX9-Hardware (on Shader 2,0 basis) are not even announced our
knowledge to.

With difficulty it only to arrange this for the users and customers
had finally zeroed in itself more or less all on the Shader 3,0
support as the feature of the coming NEXT gene generation of ATi and
nVidia. Also the Shader makes itself 3,0 support surely outstanding as
check list feature in the OEM market, of the manufacturing firms then
gladly also falsely as "DirectX 9,1 support" described (which nonsense
is natural, as here already implemented) . Positive way is not
sufficient however the support of Shader 2,0 also, in order to come
with the Windows XP successor "Longhorn" into the highest possible
hardware class ("animal 2 ") to the representation of the new
3D-Oberflaeche von Longhorn, for this Shader 3,0 support is needed.

Surely the message, which equips ATi the R420 only with Shader 2,0
support, will provide for a violent discussion. ATi made the strategic
decision here to omit the Shader 3,0 support with the R420 in order to
concentrate on the performance under Shader 2,0 applications - these
are those applications, with which the R420 will be confronted during
its lifetimes. From the thing this consideration is correct, but ATi
offers thereby naturally also some attack region for nVidia, which
will evenly offer the Shader 3,0 support with the NV4x-Linie - and
this technological superiority surely also marketing-technically to
use will try.



Off the surely flashing discussion over sense and nonsense of Shader
ATi with the renouncement of these in addition, a clear guideline for
the R420 gave 3,0: This chip is to accumulate under Shader 2,0 to the
absolute maximum form. The chip will have further 8 Rendering
pipelines like its predecessors, but these are to have been
substantially improved. Which hides itself exactly behind it, one did
not want to say us however yet. Several texture units per pipeline
could be (the R3x0-Chips had everything only one texture unit per
pipeline) , it could also internal improvements be, which will then
probably understand rather only the chip experts by us. Zumindestens
is to have energetically worked on the Rendering pipelines, in order
to carry the Shader 2,0 performance on a new level.

Furthermore the number of Vertex Shader of 4 to 6 was increased, which
could give an increase of the Vertex achievement to the R420 as well
as the clock increase and the internal improvements over up to 100 per
cent . In all other respects this increase of the number of Vertex
Shader (the number is not directly comparable in all other respects
with the nVidia chips, which a Vertex Shader array uses) is not enough
out, in order to explain the 50 million more transistors according to
our calculations - here might be in the R420 thus still place for or
other surprise, is it mentioned more texture units or however improved
anti- Aliasing and/or anisotropic filtering the already.

Unfortunately at present still no information is present, which
intends to really change and/or improve ATi with the R420 concerning
anti- Aliasing and anisotropic filter. Off the performance, which
enough is naturally never: -), is the quality of the anti- Aliasings
of the current ATi chips surely already quite optimally , while one
bilinearen however with the quality, trilinear and anisotropic filter
to surely still add can. From it thus momentarily the following can
refrained be said in summary about the R420-Chip:

ATi R420
generell basierend auf den R3x0-Chips, allerdings mit vielen
Verbesserungen an der Shader-Architektur
160 Millionen Transistoren, in 130nm bei TSMC gefertigt
8 Rendering-Pipelines, Anzahl der Textureneinheiten unbekannt
6 Vertex Shader Einheiten
DirectX 9.0 Architektur, unterstützt Shader 2.0
256 Bit Speicherinterface, unterstützt DDR1, GDDR2 und GDDR3
internes AGPx8 Interface
genaue Taktraten: unbekannt; laut AnandTech runde 500 MHz Chip- und
Speichertakt
Präsentation: vermutlich noch erstes Quartal 2004
Markteintritt: zweites Quartal 2004
Die aus dem Artikel von AnandTech stammenden Chip- und Speichertakte
hören sich erst einmal relativ niedrig an, insbesondere wenn man den
NV40 derzeit auf 500 bis 600 MHz Chiptakt und 600 bis 800 MHz
Speichertakt taxiert. Allerdings wären das gegenüber dem R360-Chip
(Radeon 9800XT) immerhin auch schon 21 Prozent mehr Chip- und ganze 37
Prozent mehr Speichertakt, während nVidia bei angenommenen 550/600 MHz
gegenüber dem NV38 (GeForceFX 5950 Ultra) takttechnisch "nur" um 16
bzw. 26 Prozent zulegt.

Zudem kann man davon ausgehen, daß beim Speichertakt definitiv noch
Spielraum nach oben ist, immerhin hat Samsung kürzlich die
Massenproduktion von GDDR3 mit bis zu 800 MHz Taktfrequenz begonnen.
Auch beim Chiptakt sollte aufgrund der 130nm Fertigung eigentlich noch
etwas mehr möglich sein - in diesem Punkt erscheinen uns die genannten
Taktraten doch etwas konservativ angegeben. Wir können uns für unseren
Teil also durchaus vorstellen, daß ATi eine mögliche Pro-Version des
R420 auch höher taktet als jene 500/500 MHz, welche AnandTech genannt
haben.


ATi R420, R423 & CO. Information
22. February 2004/from Leonidas /page 2 of 2



As evident from previous enumerating list, the R420 - like also the
NV40 - start still with an internal AGPx8 interface. In the coming age
of PCI express needs ATi naturally also a PCI express diagram chip
within the HighEnd range, which the R423 will place. That chip becomes
however, then one insured us, a native PCI express x16 interface to
have. This is in as much unusual as that is to appear the R423 in
principle in the same period as the R420 - in the second quarter. Two
160-Millionen-Transistoren-Chips however more or less in the same
period on the mass production to prepare, might be however not only
extremely expensive, but also to the personnel capacities of ATi push.

To that extent we can explain ourselves this at present only in this
kind, when the R420 and R423 are in principle the same chip. Either
the R420 is a R423 with settled AGP Bridge, which converts thus the
PCI express x16 of signals of the diagram chip into AGPx8 signal for
the AGP haven on the Main board, or however - hears themselves
improbably on, is however not not impossible - both internal AGPx8 and
internal PCI express x16 interface, which is then activated depending
upon need by BIOS, have R420 and R423. As this point dissolves in the
long run, remains being waiting however.

Abseites its is considered against the fact as safe that the R423
represents really only a pure PCI express modification of the R420:
There are no technical changes and also not more clock. Only the R480
chip will represent a genuine Refresh to the R420/R423, which will
throw however - after the experiences with the last Refresh chips of
ATi - probably only more clock into the scale pan, possibly already in
the 110nm production technology. The R480 is to begin sometime in the
second half-year 2004, in the third quarter is probably thus
introduced and in the fourth quarter delivered.



And thus we come off the HighEnd chips to the candidates for the
Mainstream and LowCost range. AnandTech have to offer here already
beautiful technical information, we can besides again the appropriate
classification of the chips offer. In the second quarter it will give
only once the RV380 to chip, which one can charaktersisieren briefly
in such a way:

ATi RV380
direct descent of RV350/RV360, invariably internally in principle
75 million transistors, in 130nm with TSMC manufactured
4 Rendering pipelines with ever a texture unit
2 Vertex Shader units
DirectX 9,0 architecture, supports Shader 2.0
, DDR1, GDDR2 and GDDR3 supports 128 bits memory interface
internal PCI express interface
exact clock rates: unknown; according to AnandTech round 500 MHz chip
and 400 MHz storing act
Presentation: probably still first quarter 2004
Market entrance: second quarter 2004
It is not correct that AnandTech indicate the same improvements on
architecture here for the RV380 as for the R420 have. The RV380 is our
knowledge after only a further Refresh of the original RV350-Chips
(Radeon 9600/pro/cSe) , which already found a first Refresh in the
RV360-Chip (Radeon 9600XT). With the RV380 apparent at the chip clock
one does not turn, which however with 500 MHz enough is already
sufficiently high, but alone the storing act raised, which in view of
the unfavorable relationship between chip and storing act with the
direct predecessor Radeon 9600XT (500/300 MHz) might be also the best
way to give to this chip more performance.

Internally the RV380-Chip thereby might be perfectly unchanged to the
RV350 and/or RV360, however the native PCI express interface would be
new. ATi goes here not the way of nVidia with the PCI express Bridge
chip, but presents the chip specially for PCI express again. The
reason for this lies in the universe in Wonder maps and their TV
abilities: For the use and treatment of in America momentarily heavily
in coming HDTV present the backward channel is needed by PCI express
x16, which nevertheless 4 GB/sec carry out can - and this
independently of the Hinkanal, which likewise 4 GB/ses width exhibits.
Because ATi for its universe in Wonder maps wants to use this
potential, one needs native PCI express interface and therefore to
Bridge chips, how she begins nVidia, a general refusal gave.

It will really give one of the R420 to descending Mainstream chip only
with the RV410 in the second half-year 2004 (the indication of
AnandTech "Q2'04" is surely a write error) . This then also the
improvements on the architecture of the R420 into the Mainstream
market to carry, wommoeglich opposite the RV380 again more clock bring
along and completely possibly already in 110nm manufactured its -
which applies to however naturally wait for it.



For the LowCost market ATi for the second quarter positioned the RV370
chip, which will be according to AnandTech the first diagram chip in
110nm production technology and can be characterized as follows:

ATi RV370
complete (?) new development for the LowCost market
unknown number of transistors, in 110nm with TSMC manufactured
4 Rendering pipelines, number of texture units unknown (probably per
one)
2 Vertex Shader units
DirectX 9,0 architecture, supports Shader 2.0
64/128 bit memory interface, supports DDR1, GDDR2 and GDDR3
internal PCI express interface
exact clock rates: unknown; according to AnandTech round 300 MHz
storing act
Presentation: probably still first quarter 2004
Market entrance: second quarter 2004
The objective of ATi with the RV370-Chip is clear thereby: Finally a
competitive chip for the DirectX9 OEM market. So far nVidia in this
rough sells quantities of GeForceFX 5200 chips, partly also with only
64 bits memory interface, at prices, with which ATi only with the own
DirectX8-Grafikchips Paroli offer can. Problematic way is however
calculated this OEM market a market, where it depends much on check
list features like "DirectX9 support", with which ATi cannot score in
the long run with a Radeon 9200 despite possibly better performance
opposite a GeForceFX 5200.

Differently around the RV350-Chip is even in its lowest form, the
Radeon 9600SE diagram map not to get so cheaply than the fact that ATi
the prices, which nVidia for the GeForceFX 5200 makes, could reach.
The reason lies here by the majority in the transistor quantity and/or
the origin of the chips: The RV350 was now once for the Mainstream
segment intended the NV34-Chip (GeForceFX 5200/Ultra & GeForceFX 5500)
and weighs nevertheless 75 million transistors, during nVidia
specially only for the LowCost segment developed and by various
reductions the number of transistors on 45 million to lower could.

ATi will hit here in our view same way with the RV370. Surely one
could manufacture also simply a RV350 in 110nm, but one will achieve
most piercing success with a clear lowering of the number of
transistors. Momentarily this point is only, safe of speculation is
however the 110nm manufacturing and the adjustment as GeForceFX
5200/5500 competitor to down into the absolute LowCost market. For
genuine players such diagram maps should be naturally no view worth,
but for ATi as nVidia they also represent the lion's share of the
business - and also this field, where in the long run on basis of the
high numbers of items set off there profits are erwirtschaftbar.



Thus still another one point remains: As already mentions, ATi gave
all solutions with Bridge chips a general refusal, since one for the
own universe in Wonder maps native PCI express interface necessarily
and there temporary solutions with Bridge chips do not lead to the
desired result. This refusal refers however also to the past AGPx8
maps of ATi, speaks Radeon 9800/Pro/XT, Radeon 9600/Pro/XT/SE and
Radeon 9200/pro/cSe. This does not want to offer ATi in accordance
with current planning on PCI to express - in contrast to it nVidia
well-known-measured its complete present line places at diagram maps
still on PCI express over, before the new chips come.

With ATi this may be connected surely also with the fact that one
wants to bring new diagram chips in the second quarter in all partial
markets into the market, while with nVidia the new Mainstream is to be
expected as also LowCost solutions rather only in the third quarter.
Nothing the despite surprises this step of ATi something, because the
need at the predecessor hardware does not even shrink according to
experience over night with the introduction on the market of new
diagram maps. Speak: It will quite give users, which for a new PCI
express Main board possibly a Radeon 9800 express wish themselves pro
on PCI (which could be realized by means of a Bridge chip
problem-free) . It remains hoping that the diagram board manufacturers
will fill to substitute accordingly here and these by ATi somewhat
left open market gap.

There itself in the article with nVidia the NV40, NV41, NV45 & CO.
Information still another (naturally prospective) nVidia Roadmap
finds, wants we also this article with such to the coming ATi
Grafikchips to lock. It would pay attention that the individual chips
are arranged after their market entrance like already with the nVidia
Roadmap after their supposed presentation time and not:





After the R4x0/RV4x0-Linie it will continue with ATi with the R500
chip in the year 2005. This will support DirectX10, which covers among
other things the support of the Shader 4,0, ATi jumps over the Shader
3,0 thus completely. More details on this chip, which already is in
development, is not well-known, just as little relative to the further
plans of ATi however yet from R500-Abkoemmlingen for the Mainstream
such as LowCost market.



 




Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
pc problems after g card upgrade + sp2 ben reed Homebuilt PC's 9 November 30th 04 02:04 AM
Response by Nvidia concerning HL2 *warning, lengthy post, strong opinion content, some bad langwidge* NC-17 rating administered... Dave Ati Videocards 28 September 14th 03 05:51 PM
I dont see that nvidia is "finished"... Steven C \(Doktersteve\) Ati Videocards 17 September 13th 03 09:00 PM
Tomb Raider AOD benches: Bad news for Nvidia who be dat? Ati Videocards 33 September 4th 03 10:35 AM
Kyle Bennett (HardOCP) blasts NVIDIA Radeon350 Ati Videocards 12 August 13th 03 09:19 PM


All times are GMT +1. The time now is 09:46 PM.


Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 HardwareBanter.
The comments are property of their posters.