Photo's!!!!

[Ali]

Bugger, AltaVista only translates upto 10K of webpage, Im getting cut off half way through the specifications

Anybody with web space want to cut n paste the iXBT into several smaller bits, so we can translate the lot?

Ali

[Ali]

The parhelion/Of paryuelia-shchy2


. Information to the reflection .
17 instants of the spring
Position of the company Of matrokh on the market generally, and on the Russian in particular, sufficiently strange.Relation to this market even more is strange.Small example.company did not find time to report (sich!)to its distributors (with exception of very individual comrades) about the output/yield of new product.What already to speak about the press...But, as the saying goes, who searches for...We found and we desire company together with her distributorskimi channels of great successes...

Why we did make this today, but not on 14 May?Everything is very simple.we do not have any obligations before the gentlemen from Canada and their European associates.By them IS NOT INTERESTING Russia.Point.Let us place point and we went...

On 14 May 2002, the already forgotten by the enthusiasts ed of visualization company Of matrokh Of grapyuichs Of inch. announces the chip Of paryuelia-shchy2, which belongs to the new generation of the architectures of graphic accelerators.Most accurate will describe this generation as Generation Kh9.. it is possible to say that the new chip from Matrokh is the second representative of this generation.priority for eDlabs (with some reservations) of that already announced of its flexibly programmed BY VPU by the name Py0.

Before close passing to the new creation Of matrokh.the basic culprit of this article, let us discuss one additional sign hero:

Mister IKS, the ninth (DirechtKh 9)
Preserving characteristic of a good secret dossier style, let us give the key innovations (from the point of view of accelerators) of the future version popular play API.

Increased accuracy of data presentation
The new sizes of textures and buffers of sequence are introduced.Each component of four (RGBA) now can be represented in the form of 32 or 16 bit numbers with the floating point in the standard sizes IYEYEYE Of fe2 and Fy'. each point will occupy 128 or 64 bits.and this will significantly raise requirements for the capacity of local storage of graphic accelerators.However, to use the similar sizes vezde and entire/all as yet there is no need.Basic purpose of precise sizes.the realization of different effects and models of illumination (for example, storage of the tables of those utilized in piksel'nykh sheyderakh).Contents of the buffers of sequence in the floating size they cannot be brought out directly to the monitor through RAMDACH or the digital interface.they are used only for the internal targets/purposes.For the increased accuracy of the idea of the integral information of that concluded on the screen is introduced new size 10:10:10:2 (RGBA), where each color component is transferred with the accuracy of 10 bits.A similar accuracy with the surplus covers/coats the possibilities of contemporary display units and substantially enlarges dynamic range, and, therefore, and the quality of the resulting image.

Maps/charts/cards Of displacement (Displachement Of mapping)
The litsenzirovannaya in Matrokh Of grapyuichs Of inch. (and, it goes without saying, realized into Paryuelia-shchy2) technology Of displachement Of mapping makes it possible to substantially increase the degree of realism and detail of relief surfaces.In contrast to the customary already varieties of the relief, applied on the surface of triangles during the painting and which does not influence the visibility of various points (by map/chart/card of relief it is simulated only the illumination of point, and not its real attitude) of the map/chart/card of displacement they make it possible to create the geometrically correct reliefs, whose intersection in the space will not appear by ideal straight line.Actually this technology modifies the position of the apexes/vertexes of triangles, shifting/shearing them in the direction of normal to the surface to the value, proportional to value from the special texture (map/chart/card of displacement):




The map/chart/card of displacement answers after gruby1. relief, displacing the apexes/vertexes of triangles in the direction of the standard/normal, assigned for this apex/vertex.If necessary, together with the map/chart/card of displacement it is possible to use a traditional map/chart/card of relief (bump of map).for the creation thinner/finer popiksel6nyx. parts.

Of course the number of triangles must be sufficient for the transfer of all nuances of the rough relief, given by the map/chart/card of displacement.In order to create automatically these triangles (without driving in the head of programmers, the resource/lifetime of central processor and AGP busbar/tire) is used well familiar to us N -patci:




I.e. the technology of an increase in detailing models on the basis of additional tesselyatsii of triangles.The maps/charts/cards of displacement cannot be used without the apparatus support N -patce1.from the point of view DKH9 this technology only essential is addition to the algorithm N -patce1:




It is interesting that in DKH9 patchi underwent the essential improvements, for example the degree of detailing (number of partitions) now can be selected automatically depending on distance from the triangle to the observer.Thus, is achieved the more optimum, closer to the uniform (from the point of view of observer) partition of scene.all triangles have approximately identical visible size/dimension:




Furthermore, for the maps/charts/cards of displacement, just as for the usual textures, it is possible to use multiple filtration (mip-mapping):




In the sum obtaining the excellent result of bez of the special troubles:




Of course the maps/charts/cards of displacement can be used not only for the flat surfaces but also for the volumetric models:






In comparison with the traditional maps/charts/cards of relief (including popiksel'nyy relief on the basis of sheyderov), map/chart/card of displacement, undoubtedly, it is more than resursoyemki.joke whether to say, all parts of relief are transferred by triangles.But also are more realistic:




Of course in the real play applications/appendices, for obtaining the dynamic and realistic drawing it is necessary to reasonably combine these technologies.

Apical sheydery of the second generation (Vertekh Syuader 2.0)
Let us briefly describe the possibilities of the new apical programs:

256 assembler commands
As the minimum of 256 constants
12 registers of the general purpose
Control of performance.condition, cycles, passages, subprogram. The substantial limitations :control is possible only on the basis of the values of constants.Conditional jumps only forward.Cycles only back.The calls of subprograms cannot be inserted.
1 register with the value of the parameter of the cycle
16 registers with the constants for the task of the parameters of the cycles
16 registers with the logical constants for the passages
Possibility, when desired, to write down sheydery on S a similar high level language, compiled into the assembler code automatically
Piksel'nye sheydery of the second generation (Pikhel Of syuader 2.0)
Let us briefly describe the possibilities of the new piksel'nykh programs:

All calculations only with real numbers (with the floating point) from the beginning also to the end of sheydera.If it is necessary, result is converted to the integral size only during the writing into the memory.
Possibility, when desired, to write down sheydery on S a similar high level language, automatically compiled into the assembler code.
To 8 textural iteratorov (i.e. the computable for each point of triangle corresponding to it textural coordinates).
To 16 samples from the textures, including dependent samples on the preliminarily calculated in sheydere coordinates.Degree of the nesting of such samples.not more than 4.
Sheyder can include to 32 assembler commands of the addressing of textures and to 64 assembler arithmetical commands in the arbitrary order.
Appeared connected/bonded with the floating calculations the commands (yekhp, ravine, rsts so forth.)
12 temporary/time vector registers of the general purpose
32 constants
Gamma-korrekqi4 values.It makes it possible to increase the realism of the transfer of thin/fine color nuances, especially in the shadows of the synthesized images.Is conducted straight line Gamma-korrekqi4 during the record of the results of sheydera into the buffer of sequence and reverse correction upon the entering of initial data from the buffer of sequence or texture, it is possible to govern/control the value of gamma/range.
Furthermore, appeared so urgent/actual a size of the compression of volumetric textures.DKHVCH, litsenzirovannyy in NVIDIA.

Let us total .Essential rate is made (as in the case Of openGL 2.0 ) to the programmability of accelerator.An appearance of a high level language and a considerable increase in the flexibility of the programmed blocks will play their positive role.Dream of many companies, which develop/process graphic chips.to shift the majority of complexities (algorithms) of iron on the arms of programmers is personified by seven-mile steps/pitches.Now, inside the conveyors of accelerator, all data are transferred and are processed with the high accuracy, ensured by the floating size.

The appearance of the stabilized version MS DKH9 is expected at the end of the summer, and its official output/yield, judging by everything, will be christmas gift.

But now make it possible to present to you in entire krase new chip from Matrokh Of grapyuichs:

Paryuelia-shchy2 THE GPU

[Helevitia]

check out the other threads, all the info is there.

Hele

[ravalox]

Seems Matrox has pissed off the Russians...

[Novdid]

What I really liked to see is there, and that's 10bit RGB.

[Dr Mordrid]

That certainly applies to the Chinese page, but the Russian one is a bit more problematic.

Lets not forget that there are rampant reports of Matrox giving alpha Parhelias to companies for review & testing without NDA's.

If the Russian info originated from a leak at one of those sites it's Matrox's own fault.

Dr. Mordrid

[UncleSam]

Here official english version of iXBT.com in-depth technology review for Parhelia-512 chip

http://www.digit-life.com/articles/matroxparhelia512/index.html

Please Note: iXBT.com have _no _NDA _signed with Matrox and free to post all the details whenever they want.

[Maniac]

Thanks finally I can understand


P.S. why are all these threads being placed in the cystal ball forum?
None of this is speculation anymore, this is actual Matrox Hardware now!

[Genom]

A couple days late...

Dammit. And here I was so proud this weekend cause I got to play with one of those buggers and I wanted to let ya'll know how cool it was. (I dont have an NDA). ANyways, a local company to me has their engineering samples in and I was playing with it and all I can say is DAMN! I am glad I have been waiting for a new Matrox card. Only bad thing about it is the price on the high end here in the US is gonna be a pain in the ass. But it's definitly worth it ya'll.

[Pace]

Say Matrox wanted Parhelia to have 512MB and QuadHead (3 VDUs for surround gaming plus a score VDU ). They put this under NDA in case they can't realise their plans.

Now some website hears about this, tell everyone Matrox will have QuadHead/512MB....then Matrox announce TriHead/256MB. Maybe Displacement Mapping wont work right out of the box, so they don't say they have that. If any features are lessened, then people are disappointed with the new card.

Maybe NVIDIA will annouce a QuadView capable card now to take the wind out of Matrox' announcement tomorrow.

There are plenty of things that can cause harm. And I think NVIDIA/ATi will have plans like this somewhere - so they could've announced something like that today.

P.

(PS: I know QuadHead ain't that good for surround gaming, but I wouldn't mind a 4th display showing something else other than the FOV )

[Chrono_Wanderer]

Good old days: found this really really

*****MAJOR BUMP*****

Originally posted by TDB
that would require more than 256mb video-ram, so don´t count on it.

unless matrox found a way to make fsaa in some fancy new way.

Originally posted by Joel
Quote: unless matrox found a way to make fsaa in some fancy new way.
Joel:

Joel

Originally posted by Ant
Quote: unless matrox found a way to make fsaa in some fancy new way.
Ant:

and also the fact that almost all ppl's sigs have a parhelia in them

[TdB]

aahhh! the good old days.

[borat]

what i like is that when no body could benchmark the card or new its clock speed they weer all blown away then when the benchies emerged - wham. also dont understand why the compition have not developed faa now, its got to be way cheaper than adding horsepower to the card.(i know its not perfect but the option would be usefull on all cards).

[chaoliang]

It was sooo exciting. But the dream did not become true. What a pity.

[Topha]

Soooo... whats next