As we have expected, yesterday Matrox announced its new Parhelia-512 graphics chip (parhelion is an optical effect) also known as G1000.

Parhelion optical effect

The major chips specification of the chip reported by our friends from ReactorCritical web-site proved absolutely correct. Parhelia-512 chip is monster featuring around 80-90 million transistors, 4 rendering pipelines with 4 TMUs each, 512bit internal structure, 256bit memory access, AGP 4x/8x support, and support of TripleHead technology (simultaneous output to three displays) and SurroundGaming (3D image similar to the one used in IMAX theaters). The chip also allows using up to 256MB DDR SDRAM.

The official press-release also added the whole bunch of interesting details: 10bit GigaColor technology (supports 10bit colors in R/G/B channel), two 400MHz RAMDAC chips, Glyph Antialiasing (hardware fonts anti-aliasing), 16x (!)Fragment Antialiasing, 8-sample anisotropic filtering, Quad DirectX 9 Vertex Shader Array, Quad Texturing, HDM (Hardware Displacement Mapping), “virtual” resolution up to 3840x1024x32bpp in case of three displays, DVD playback with PC-Theater quality (10bit decoding, filtering and scaling), and so on, and so forth.

The chip will be produced by UMC with 0.15micron technology. So far, there are no official details about the working frequencies, pricing and shipping schedule (for the chips and the cards based on these chips). Our sources claim that the card built on new Matrox parhelia-512 will cost at least $450 (the picture of the first Parhelia based card is available here).

Many web-sites, which have already reported something about the fresh Parhelia-512 were nearly unanimous when calling it a revolutionary product and predict it to be a performance king of the graphics market (even for a short period of time). In some other situation I could have agreed with them, but not now. I will not deny that the chip boasts a number of really interesting features and might turn out quite a tidbit in terms of rich set of functions and (maybe) great image quality. But!

Let’s cast a bit more critical glance at Parhelia-512. The company claims that its new product is "designed to surpass the expectations of the most demanding professional users and PC enthusiasts". Let’s see what kind of users belong to these two target groups. We would like to start with the latter:

• The PC enthusiasts are not so numerous in general, and those of them who pay attention not to the fps rate in Quake3, but to the rich additional features and image quality in the first place are very few.
• As for professional application, which I considered the most suitable for parhelia-512 once I saw the specs, everything will depend on the software quality, on the drivers for professional applications (such as ELSA drivers for Quadro, for instance). And Matrox has never been very successful at writing drivers. However, everything may change this time, though I will not believe it until I see the tests results. Of course, there is also a group of professional users requiring excellent 2D quality (such as editors, for instance), but do they really need such a powerful 3D then?

So, Matrox may have got beyond the expectations of these particular users, but do the users really need it? In fact, a solution like Parhelia-512 seems a very unprofitable product from a marketing point of view, as it is too complicated and stuffed with various functions. Take, for example, 80 million transistors! There will appear some problems with the high working frequencies, and with the production cost of the chip. Even for a start in the professional market they could have designed something not so complicated, not to mention the enthusiasts, of course.

Well, you may call me an inveterate skeptic, but I do not believe in immense popularity of the new Matrox Parhelia-512 in any of the market segments, and hence cannot regard this product as a revolution in the full meaning of this word. At least until I see some details test results.